?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
5445 Add more visibility via arcstats; specifically arc_state_t stats and differentia...
[unleashed.git] / usr / src / cmd / mdb / common / modules / zfs / zfs.c
blob225cf3dee18e110149ce6c7180fa50a3928893f9
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) 2011, 2014 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 enum spa_flags {
60 SPA_FLAG_CONFIG = 1 << 0,
61 SPA_FLAG_VDEVS = 1 << 1,
62 SPA_FLAG_ERRORS = 1 << 2,
63 SPA_FLAG_METASLAB_GROUPS = 1 << 3,
64 SPA_FLAG_METASLABS = 1 << 4,
65 SPA_FLAG_HISTOGRAMS = 1 << 5
66 };
67
68 #define SPA_FLAG_ALL_VDEV \
69 (SPA_FLAG_VDEVS | SPA_FLAG_ERRORS | SPA_FLAG_METASLAB_GROUPS | \
70 SPA_FLAG_METASLABS | SPA_FLAG_HISTOGRAMS)
71
72 static int
73 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp,
74 const char *member, int len, void *buf)
75 {
76 mdb_ctf_id_t id;
77 ulong_t off;
78 char name[64];
79
80 if (idp == NULL) {
81 if (mdb_ctf_lookup_by_name(type, &id) == -1) {
82 mdb_warn("couldn't find type %s", type);
83 return (DCMD_ERR);
84 }
85 idp = &id;
86 } else {
87 type = name;
88 mdb_ctf_type_name(*idp, name, sizeof (name));
89 }
90
91 if (mdb_ctf_offsetof(*idp, member, &off) == -1) {
92 mdb_warn("couldn't find member %s of type %s\n", member, type);
93 return (DCMD_ERR);
94 }
95 if (off % 8 != 0) {
96 mdb_warn("member %s of type %s is unsupported bitfield",
97 member, type);
98 return (DCMD_ERR);
99 }
100 off /= 8;
101
102 if (mdb_vread(buf, len, addr + off) == -1) {
103 mdb_warn("failed to read %s from %s at %p",
104 member, type, addr + off);
105 return (DCMD_ERR);
106 }
107 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */
108
109 return (0);
110 }
111
112 #define GETMEMB(addr, structname, member, dest) \
113 getmember(addr, ZFS_STRUCT structname, NULL, #member, \
114 sizeof (dest), &(dest))
115
116 #define GETMEMBID(addr, ctfid, member, dest) \
117 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest))
118
119 static boolean_t
120 strisprint(const char *cp)
121 {
122 for (; *cp; cp++) {
123 if (!isprint(*cp))
124 return (B_FALSE);
125 }
126 return (B_TRUE);
127 }
128
129 #define NICENUM_BUFLEN 6
130
131 static int
132 snprintfrac(char *buf, int len,
133 uint64_t numerator, uint64_t denom, int frac_digits)
134 {
135 int mul = 1;
136 int whole, frac, i;
137
138 for (i = frac_digits; i; i--)
139 mul *= 10;
140 whole = numerator / denom;
141 frac = mul * numerator / denom - mul * whole;
142 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac));
143 }
144
145 static void
146 mdb_nicenum(uint64_t num, char *buf)
147 {
148 uint64_t n = num;
149 int index = 0;
150 char *u;
151
152 while (n >= 1024) {
153 n = (n + (1024 / 2)) / 1024; /* Round up or down */
154 index++;
155 }
156
157 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2];
158
159 if (index == 0) {
160 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu",
161 (u_longlong_t)n);
162 } else if (n < 10 && (num & (num - 1)) != 0) {
163 (void) snprintfrac(buf, NICENUM_BUFLEN,
164 num, 1ULL << 10 * index, 2);
165 strcat(buf, u);
166 } else if (n < 100 && (num & (num - 1)) != 0) {
167 (void) snprintfrac(buf, NICENUM_BUFLEN,
168 num, 1ULL << 10 * index, 1);
169 strcat(buf, u);
170 } else {
171 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s",
172 (u_longlong_t)n, u);
173 }
174 }
175
176 static int verbose;
177
178 static int
179 freelist_walk_init(mdb_walk_state_t *wsp)
180 {
181 if (wsp->walk_addr == NULL) {
182 mdb_warn("must supply starting address\n");
183 return (WALK_ERR);
184 }
185
186 wsp->walk_data = 0; /* Index into the freelist */
187 return (WALK_NEXT);
188 }
189
190 static int
191 freelist_walk_step(mdb_walk_state_t *wsp)
192 {
193 uint64_t entry;
194 uintptr_t number = (uintptr_t)wsp->walk_data;
195 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
196 "INVALID", "INVALID", "INVALID", "INVALID" };
197 int mapshift = SPA_MINBLOCKSHIFT;
198
199 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) {
200 mdb_warn("failed to read freelist entry %p", wsp->walk_addr);
201 return (WALK_DONE);
202 }
203 wsp->walk_addr += sizeof (entry);
204 wsp->walk_data = (void *)(number + 1);
205
206 if (SM_DEBUG_DECODE(entry)) {
207 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n",
208 number,
209 ddata[SM_DEBUG_ACTION_DECODE(entry)],
210 SM_DEBUG_TXG_DECODE(entry),
211 SM_DEBUG_SYNCPASS_DECODE(entry));
212 } else {
213 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c "
214 "size=%06llx", number,
215 SM_OFFSET_DECODE(entry) << mapshift,
216 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) <<
217 mapshift,
218 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
219 SM_RUN_DECODE(entry) << mapshift);
220 if (verbose)
221 mdb_printf(" (raw=%012llx)\n", entry);
222 mdb_printf("\n");
223 }
224 return (WALK_NEXT);
225 }
226
227 static int
228 mdb_dsl_dir_name(uintptr_t addr, char *buf)
229 {
230 static int gotid;
231 static mdb_ctf_id_t dd_id;
232 uintptr_t dd_parent;
233 char dd_myname[MAXNAMELEN];
234
235 if (!gotid) {
236 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dir",
237 &dd_id) == -1) {
238 mdb_warn("couldn't find struct dsl_dir");
239 return (DCMD_ERR);
240 }
241 gotid = TRUE;
242 }
243 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) ||
244 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) {
245 return (DCMD_ERR);
246 }
247
248 if (dd_parent) {
249 if (mdb_dsl_dir_name(dd_parent, buf))
250 return (DCMD_ERR);
251 strcat(buf, "/");
252 }
253
254 if (dd_myname[0])
255 strcat(buf, dd_myname);
256 else
257 strcat(buf, "???");
258
259 return (0);
260 }
261
262 static int
263 objset_name(uintptr_t addr, char *buf)
264 {
265 static int gotid;
266 static mdb_ctf_id_t os_id, ds_id;
267 uintptr_t os_dsl_dataset;
268 char ds_snapname[MAXNAMELEN];
269 uintptr_t ds_dir;
270
271 buf[0] = '\0';
272
273 if (!gotid) {
274 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "objset",
275 &os_id) == -1) {
276 mdb_warn("couldn't find struct objset");
277 return (DCMD_ERR);
278 }
279 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dataset",
280 &ds_id) == -1) {
281 mdb_warn("couldn't find struct dsl_dataset");
282 return (DCMD_ERR);
283 }
284
285 gotid = TRUE;
286 }
287
288 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset))
289 return (DCMD_ERR);
290
291 if (os_dsl_dataset == 0) {
292 strcat(buf, "mos");
293 return (0);
294 }
295
296 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) ||
297 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) {
298 return (DCMD_ERR);
299 }
300
301 if (ds_dir && mdb_dsl_dir_name(ds_dir, buf))
302 return (DCMD_ERR);
303
304 if (ds_snapname[0]) {
305 strcat(buf, "@");
306 strcat(buf, ds_snapname);
307 }
308 return (0);
309 }
310
311 static void
312 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val,
313 const char *prefix)
314 {
315 const char *cp;
316 size_t len = strlen(prefix);
317
318 if ((cp = mdb_ctf_enum_name(id, val)) != NULL) {
319 if (strncmp(cp, prefix, len) == 0)
320 cp += len;
321 (void) strncpy(out, cp, size);
322 } else {
323 mdb_snprintf(out, size, "? (%d)", val);
324 }
325 }
326
327 /* ARGSUSED */
328 static int
329 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
330 {
331 /*
332 * This table can be approximately generated by running:
333 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2
334 */
335 static const char *params[] = {
336 "arc_reduce_dnlc_percent",
337 "arc_lotsfree_percent",
338 "zfs_dirty_data_max",
339 "zfs_dirty_data_sync",
340 "zfs_delay_max_ns",
341 "zfs_delay_min_dirty_percent",
342 "zfs_delay_scale",
343 "zfs_vdev_max_active",
344 "zfs_vdev_sync_read_min_active",
345 "zfs_vdev_sync_read_max_active",
346 "zfs_vdev_sync_write_min_active",
347 "zfs_vdev_sync_write_max_active",
348 "zfs_vdev_async_read_min_active",
349 "zfs_vdev_async_read_max_active",
350 "zfs_vdev_async_write_min_active",
351 "zfs_vdev_async_write_max_active",
352 "zfs_vdev_scrub_min_active",
353 "zfs_vdev_scrub_max_active",
354 "zfs_vdev_async_write_active_min_dirty_percent",
355 "zfs_vdev_async_write_active_max_dirty_percent",
356 "spa_asize_inflation",
357 "zfs_arc_max",
358 "zfs_arc_min",
359 "arc_shrink_shift",
360 "zfs_mdcomp_disable",
361 "zfs_prefetch_disable",
362 "zfetch_max_streams",
363 "zfetch_min_sec_reap",
364 "zfetch_block_cap",
365 "zfetch_array_rd_sz",
366 "zfs_default_bs",
367 "zfs_default_ibs",
368 "metaslab_aliquot",
369 "reference_tracking_enable",
370 "reference_history",
371 "spa_max_replication_override",
372 "spa_mode_global",
373 "zfs_flags",
374 "zfs_txg_timeout",
375 "zfs_vdev_cache_max",
376 "zfs_vdev_cache_size",
377 "zfs_vdev_cache_bshift",
378 "vdev_mirror_shift",
379 "zfs_scrub_limit",
380 "zfs_no_scrub_io",
381 "zfs_no_scrub_prefetch",
382 "zfs_vdev_aggregation_limit",
383 "fzap_default_block_shift",
384 "zfs_immediate_write_sz",
385 "zfs_read_chunk_size",
386 "zfs_nocacheflush",
387 "zil_replay_disable",
388 "metaslab_gang_bang",
389 "metaslab_df_alloc_threshold",
390 "metaslab_df_free_pct",
391 "zio_injection_enabled",
392 "zvol_immediate_write_sz",
393 };
394
395 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) {
396 int sz;
397 uint64_t val64;
398 uint32_t *val32p = (uint32_t *)&val64;
399
400 sz = mdb_readvar(&val64, params[i]);
401 if (sz == 4) {
402 mdb_printf("%s = 0x%x\n", params[i], *val32p);
403 } else if (sz == 8) {
404 mdb_printf("%s = 0x%llx\n", params[i], val64);
405 } else {
406 mdb_warn("variable %s not found", params[i]);
407 }
408 }
409
410 return (DCMD_OK);
411 }
412
413 /* ARGSUSED */
414 static int
415 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
416 {
417 mdb_ctf_id_t type_enum, checksum_enum, compress_enum;
418 char type[80], checksum[80], compress[80];
419 blkptr_t blk, *bp = &blk;
420 char buf[BP_SPRINTF_LEN];
421
422 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) {
423 mdb_warn("failed to read blkptr_t");
424 return (DCMD_ERR);
425 }
426
427 if (mdb_ctf_lookup_by_name("enum dmu_object_type", &type_enum) == -1 ||
428 mdb_ctf_lookup_by_name("enum zio_checksum", &checksum_enum) == -1 ||
429 mdb_ctf_lookup_by_name("enum zio_compress", &compress_enum) == -1) {
430 mdb_warn("Could not find blkptr enumerated types");
431 return (DCMD_ERR);
432 }
433
434 enum_lookup(type, sizeof (type), type_enum,
435 BP_GET_TYPE(bp), "DMU_OT_");
436 enum_lookup(checksum, sizeof (checksum), checksum_enum,
437 BP_GET_CHECKSUM(bp), "ZIO_CHECKSUM_");
438 enum_lookup(compress, sizeof (compress), compress_enum,
439 BP_GET_COMPRESS(bp), "ZIO_COMPRESS_");
440
441 SNPRINTF_BLKPTR(mdb_snprintf, '\n', buf, sizeof (buf), bp, type,
442 checksum, compress);
443
444 mdb_printf("%s\n", buf);
445
446 return (DCMD_OK);
447 }
448
449 typedef struct mdb_dmu_buf_impl {
450 struct {
451 uint64_t db_object;
452 } db;
453 uintptr_t db_objset;
454 uint64_t db_level;
455 uint64_t db_blkid;
456 struct {
457 uint64_t rc_count;
458 } db_holds;
459 } mdb_dmu_buf_impl_t;
460
461 /* ARGSUSED */
462 static int
463 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
464 {
465 mdb_dmu_buf_impl_t db;
466 char objectname[32];
467 char blkidname[32];
468 char path[MAXNAMELEN];
469
470 if (DCMD_HDRSPEC(flags))
471 mdb_printf(" addr object lvl blkid holds os\n");
472
473 if (mdb_ctf_vread(&db, ZFS_STRUCT "dmu_buf_impl", "mdb_dmu_buf_impl_t",
474 addr, 0) == -1)
475 return (DCMD_ERR);
476
477 if (db.db.db_object == DMU_META_DNODE_OBJECT)
478 (void) strcpy(objectname, "mdn");
479 else
480 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx",
481 (u_longlong_t)db.db.db_object);
482
483 if (db.db_blkid == DMU_BONUS_BLKID)
484 (void) strcpy(blkidname, "bonus");
485 else
486 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx",
487 (u_longlong_t)db.db_blkid);
488
489 if (objset_name(db.db_objset, path)) {
490 return (DCMD_ERR);
491 }
492
493 mdb_printf("%p %8s %1u %9s %2llu %s\n", addr,
494 objectname, (int)db.db_level, blkidname,
495 db.db_holds.rc_count, path);
496
497 return (DCMD_OK);
498 }
499
500 /* ARGSUSED */
501 static int
502 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
503 {
504 #define HISTOSZ 32
505 uintptr_t dbp;
506 dmu_buf_impl_t db;
507 dbuf_hash_table_t ht;
508 uint64_t bucket, ndbufs;
509 uint64_t histo[HISTOSZ];
510 uint64_t histo2[HISTOSZ];
511 int i, maxidx;
512
513 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) {
514 mdb_warn("failed to read 'dbuf_hash_table'");
515 return (DCMD_ERR);
516 }
517
518 for (i = 0; i < HISTOSZ; i++) {
519 histo[i] = 0;
520 histo2[i] = 0;
521 }
522
523 ndbufs = 0;
524 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) {
525 int len;
526
527 if (mdb_vread(&dbp, sizeof (void *),
528 (uintptr_t)(ht.hash_table+bucket)) == -1) {
529 mdb_warn("failed to read hash bucket %u at %p",
530 bucket, ht.hash_table+bucket);
531 return (DCMD_ERR);
532 }
533
534 len = 0;
535 while (dbp != 0) {
536 if (mdb_vread(&db, sizeof (dmu_buf_impl_t),
537 dbp) == -1) {
538 mdb_warn("failed to read dbuf at %p", dbp);
539 return (DCMD_ERR);
540 }
541 dbp = (uintptr_t)db.db_hash_next;
542 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--)
543 histo2[i]++;
544 len++;
545 ndbufs++;
546 }
547
548 if (len >= HISTOSZ)
549 len = HISTOSZ-1;
550 histo[len]++;
551 }
552
553 mdb_printf("hash table has %llu buckets, %llu dbufs "
554 "(avg %llu buckets/dbuf)\n",
555 ht.hash_table_mask+1, ndbufs,
556 (ht.hash_table_mask+1)/ndbufs);
557
558 mdb_printf("\n");
559 maxidx = 0;
560 for (i = 0; i < HISTOSZ; i++)
561 if (histo[i] > 0)
562 maxidx = i;
563 mdb_printf("hash chain length number of buckets\n");
564 for (i = 0; i <= maxidx; i++)
565 mdb_printf("%u %llu\n", i, histo[i]);
566
567 mdb_printf("\n");
568 maxidx = 0;
569 for (i = 0; i < HISTOSZ; i++)
570 if (histo2[i] > 0)
571 maxidx = i;
572 mdb_printf("hash chain depth number of dbufs\n");
573 for (i = 0; i <= maxidx; i++)
574 mdb_printf("%u or more %llu %llu%%\n",
575 i, histo2[i], histo2[i]*100/ndbufs);
576
577
578 return (DCMD_OK);
579 }
580
581 #define CHAIN_END 0xffff
582 /*
583 * ::zap_leaf [-v]
584 *
585 * Print a zap_leaf_phys_t, assumed to be 16k
586 */
587 /* ARGSUSED */
588 static int
589 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
590 {
591 char buf[16*1024];
592 int verbose = B_FALSE;
593 int four = B_FALSE;
594 dmu_buf_t l_dbuf;
595 zap_leaf_t l;
596 zap_leaf_phys_t *zlp = (void *)buf;
597 int i;
598
599 if (mdb_getopts(argc, argv,
600 'v', MDB_OPT_SETBITS, TRUE, &verbose,
601 '4', MDB_OPT_SETBITS, TRUE, &four,
602 NULL) != argc)
603 return (DCMD_USAGE);
604
605 l_dbuf.db_data = zlp;
606 l.l_dbuf = &l_dbuf;
607 l.l_bs = 14; /* assume 16k blocks */
608 if (four)
609 l.l_bs = 12;
610
611 if (!(flags & DCMD_ADDRSPEC)) {
612 return (DCMD_USAGE);
613 }
614
615 if (mdb_vread(buf, sizeof (buf), addr) == -1) {
616 mdb_warn("failed to read zap_leaf_phys_t at %p", addr);
617 return (DCMD_ERR);
618 }
619
620 if (zlp->l_hdr.lh_block_type != ZBT_LEAF ||
621 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) {
622 mdb_warn("This does not appear to be a zap_leaf_phys_t");
623 return (DCMD_ERR);
624 }
625
626 mdb_printf("zap_leaf_phys_t at %p:\n", addr);
627 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len);
628 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix);
629 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries);
630 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree,
631 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l)));
632 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist);
633 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags,
634 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ?
635 "ENTRIES_CDSORTED" : "");
636
637 if (verbose) {
638 mdb_printf(" hash table:\n");
639 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) {
640 if (zlp->l_hash[i] != CHAIN_END)
641 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]);
642 }
643 }
644
645 mdb_printf(" chunks:\n");
646 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) {
647 /* LINTED: alignment */
648 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i);
649 switch (zlc->l_entry.le_type) {
650 case ZAP_CHUNK_FREE:
651 if (verbose) {
652 mdb_printf(" %u: free; lf_next = %u\n",
653 i, zlc->l_free.lf_next);
654 }
655 break;
656 case ZAP_CHUNK_ENTRY:
657 mdb_printf(" %u: entry\n", i);
658 if (verbose) {
659 mdb_printf(" le_next = %u\n",
660 zlc->l_entry.le_next);
661 }
662 mdb_printf(" le_name_chunk = %u\n",
663 zlc->l_entry.le_name_chunk);
664 mdb_printf(" le_name_numints = %u\n",
665 zlc->l_entry.le_name_numints);
666 mdb_printf(" le_value_chunk = %u\n",
667 zlc->l_entry.le_value_chunk);
668 mdb_printf(" le_value_intlen = %u\n",
669 zlc->l_entry.le_value_intlen);
670 mdb_printf(" le_value_numints = %u\n",
671 zlc->l_entry.le_value_numints);
672 mdb_printf(" le_cd = %u\n",
673 zlc->l_entry.le_cd);
674 mdb_printf(" le_hash = %llx\n",
675 zlc->l_entry.le_hash);
676 break;
677 case ZAP_CHUNK_ARRAY:
678 mdb_printf(" %u: array", i);
679 if (strisprint((char *)zlc->l_array.la_array))
680 mdb_printf(" \"%s\"", zlc->l_array.la_array);
681 mdb_printf("\n");
682 if (verbose) {
683 int j;
684 mdb_printf(" ");
685 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) {
686 mdb_printf("%02x ",
687 zlc->l_array.la_array[j]);
688 }
689 mdb_printf("\n");
690 }
691 if (zlc->l_array.la_next != CHAIN_END) {
692 mdb_printf(" lf_next = %u\n",
693 zlc->l_array.la_next);
694 }
695 break;
696 default:
697 mdb_printf(" %u: undefined type %u\n",
698 zlc->l_entry.le_type);
699 }
700 }
701
702 return (DCMD_OK);
703 }
704
705 typedef struct dbufs_data {
706 mdb_ctf_id_t id;
707 uint64_t objset;
708 uint64_t object;
709 uint64_t level;
710 uint64_t blkid;
711 char *osname;
712 } dbufs_data_t;
713
714 #define DBUFS_UNSET (0xbaddcafedeadbeefULL)
715
716 /* ARGSUSED */
717 static int
718 dbufs_cb(uintptr_t addr, const void *unknown, void *arg)
719 {
720 dbufs_data_t *data = arg;
721 uintptr_t objset;
722 dmu_buf_t db;
723 uint8_t level;
724 uint64_t blkid;
725 char osname[MAXNAMELEN];
726
727 if (GETMEMBID(addr, &data->id, db_objset, objset) ||
728 GETMEMBID(addr, &data->id, db, db) ||
729 GETMEMBID(addr, &data->id, db_level, level) ||
730 GETMEMBID(addr, &data->id, db_blkid, blkid)) {
731 return (WALK_ERR);
732 }
733
734 if ((data->objset == DBUFS_UNSET || data->objset == objset) &&
735 (data->osname == NULL || (objset_name(objset, osname) == 0 &&
736 strcmp(data->osname, osname) == 0)) &&
737 (data->object == DBUFS_UNSET || data->object == db.db_object) &&
738 (data->level == DBUFS_UNSET || data->level == level) &&
739 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) {
740 mdb_printf("%#lr\n", addr);
741 }
742 return (WALK_NEXT);
743 }
744
745 /* ARGSUSED */
746 static int
747 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
748 {
749 dbufs_data_t data;
750 char *object = NULL;
751 char *blkid = NULL;
752
753 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET;
754 data.osname = NULL;
755
756 if (mdb_getopts(argc, argv,
757 'O', MDB_OPT_UINT64, &data.objset,
758 'n', MDB_OPT_STR, &data.osname,
759 'o', MDB_OPT_STR, &object,
760 'l', MDB_OPT_UINT64, &data.level,
761 'b', MDB_OPT_STR, &blkid) != argc) {
762 return (DCMD_USAGE);
763 }
764
765 if (object) {
766 if (strcmp(object, "mdn") == 0) {
767 data.object = DMU_META_DNODE_OBJECT;
768 } else {
769 data.object = mdb_strtoull(object);
770 }
771 }
772
773 if (blkid) {
774 if (strcmp(blkid, "bonus") == 0) {
775 data.blkid = DMU_BONUS_BLKID;
776 } else {
777 data.blkid = mdb_strtoull(blkid);
778 }
779 }
780
781 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dmu_buf_impl", &data.id) == -1) {
782 mdb_warn("couldn't find struct dmu_buf_impl_t");
783 return (DCMD_ERR);
784 }
785
786 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) {
787 mdb_warn("can't walk dbufs");
788 return (DCMD_ERR);
789 }
790
791 return (DCMD_OK);
792 }
793
794 typedef struct abuf_find_data {
795 dva_t dva;
796 mdb_ctf_id_t id;
797 } abuf_find_data_t;
798
799 /* ARGSUSED */
800 static int
801 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg)
802 {
803 abuf_find_data_t *data = arg;
804 dva_t dva;
805
806 if (GETMEMBID(addr, &data->id, b_dva, dva)) {
807 return (WALK_ERR);
808 }
809
810 if (dva.dva_word[0] == data->dva.dva_word[0] &&
811 dva.dva_word[1] == data->dva.dva_word[1]) {
812 mdb_printf("%#lr\n", addr);
813 }
814 return (WALK_NEXT);
815 }
816
817 /* ARGSUSED */
818 static int
819 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
820 {
821 abuf_find_data_t data;
822 GElf_Sym sym;
823 int i;
824 const char *syms[] = {
825 "ARC_mru",
826 "ARC_mru_ghost",
827 "ARC_mfu",
828 "ARC_mfu_ghost",
829 };
830
831 if (argc != 2)
832 return (DCMD_USAGE);
833
834 for (i = 0; i < 2; i ++) {
835 switch (argv[i].a_type) {
836 case MDB_TYPE_STRING:
837 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str);
838 break;
839 case MDB_TYPE_IMMEDIATE:
840 data.dva.dva_word[i] = argv[i].a_un.a_val;
841 break;
842 default:
843 return (DCMD_USAGE);
844 }
845 }
846
847 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "arc_buf_hdr", &data.id) == -1) {
848 mdb_warn("couldn't find struct arc_buf_hdr");
849 return (DCMD_ERR);
850 }
851
852 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) {
853 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, syms[i], &sym)) {
854 mdb_warn("can't find symbol %s", syms[i]);
855 return (DCMD_ERR);
856 }
857
858 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) {
859 mdb_warn("can't walk %s", syms[i]);
860 return (DCMD_ERR);
861 }
862 }
863
864 return (DCMD_OK);
865 }
866
867
868 typedef struct dbgmsg_arg {
869 boolean_t da_verbose;
870 boolean_t da_address;
871 } dbgmsg_arg_t;
872
873 /* ARGSUSED */
874 static int
875 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg)
876 {
877 static mdb_ctf_id_t id;
878 static boolean_t gotid;
879 static ulong_t off;
880
881 dbgmsg_arg_t *da = arg;
882 time_t timestamp;
883 char buf[1024];
884
885 if (!gotid) {
886 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "zfs_dbgmsg", &id) ==
887 -1) {
888 mdb_warn("couldn't find struct zfs_dbgmsg");
889 return (WALK_ERR);
890 }
891 gotid = TRUE;
892 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) {
893 mdb_warn("couldn't find zdm_msg");
894 return (WALK_ERR);
895 }
896 off /= 8;
897 }
898
899
900 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) {
901 return (WALK_ERR);
902 }
903
904 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) {
905 mdb_warn("failed to read zdm_msg at %p\n", addr + off);
906 return (DCMD_ERR);
907 }
908
909 if (da->da_address)
910 mdb_printf("%p ", addr);
911 if (da->da_verbose)
912 mdb_printf("%Y ", timestamp);
913
914 mdb_printf("%s\n", buf);
915
916 if (da->da_verbose)
917 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
918
919 return (WALK_NEXT);
920 }
921
922 /* ARGSUSED */
923 static int
924 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
925 {
926 GElf_Sym sym;
927 dbgmsg_arg_t da = { 0 };
928
929 if (mdb_getopts(argc, argv,
930 'v', MDB_OPT_SETBITS, B_TRUE, &da.da_verbose,
931 'a', MDB_OPT_SETBITS, B_TRUE, &da.da_address,
932 NULL) != argc)
933 return (DCMD_USAGE);
934
935 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "zfs_dbgmsgs", &sym)) {
936 mdb_warn("can't find zfs_dbgmsgs");
937 return (DCMD_ERR);
938 }
939
940 if (mdb_pwalk("list", dbgmsg_cb, &da, sym.st_value) != 0) {
941 mdb_warn("can't walk zfs_dbgmsgs");
942 return (DCMD_ERR);
943 }
944
945 return (DCMD_OK);
946 }
947
948 /*ARGSUSED*/
949 static int
950 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
951 {
952 kstat_named_t *stats;
953 GElf_Sym sym;
954 int nstats, i;
955 uint_t opt_a = FALSE;
956 uint_t opt_b = FALSE;
957 uint_t shift = 0;
958 const char *suffix;
959
960 static const char *bytestats[] = {
961 "p", "c", "c_min", "c_max", "size", "duplicate_buffers_size",
962 "arc_meta_used", "arc_meta_limit", "arc_meta_max",
963 "arc_meta_min", "hdr_size", "data_size", "metadata_size",
964 "other_size", "anon_size", "anon_evictable_data",
965 "anon_evictable_metadata", "mru_size", "mru_evictable_data",
966 "mru_evictable_metadata", "mru_ghost_size",
967 "mru_ghost_evictable_data", "mru_ghost_evictable_metadata",
968 "mfu_size", "mfu_evictable_data", "mfu_evictable_metadata",
969 "mfu_ghost_size", "mfu_ghost_evictable_data",
970 "mfu_ghost_evictable_metadata", "evict_l2_cached",
971 "evict_l2_eligible", "evict_l2_ineligible", "l2_read_bytes",
972 "l2_write_bytes", "l2_size", "l2_asize", "l2_hdr_size",
973 NULL
974 };
975
976 static const char *extras[] = {
977 "arc_no_grow", "arc_tempreserve",
978 NULL
979 };
980
981 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "arc_stats", &sym) == -1) {
982 mdb_warn("failed to find 'arc_stats'");
983 return (DCMD_ERR);
984 }
985
986 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC);
987
988 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) {
989 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value);
990 return (DCMD_ERR);
991 }
992
993 nstats = sym.st_size / sizeof (kstat_named_t);
994
995 /* NB: -a / opt_a are ignored for backwards compatability */
996 if (mdb_getopts(argc, argv,
997 'a', MDB_OPT_SETBITS, TRUE, &opt_a,
998 'b', MDB_OPT_SETBITS, TRUE, &opt_b,
999 'k', MDB_OPT_SETBITS, 10, &shift,
1000 'm', MDB_OPT_SETBITS, 20, &shift,
1001 'g', MDB_OPT_SETBITS, 30, &shift,
1002 NULL) != argc)
1003 return (DCMD_USAGE);
1004
1005 if (!opt_b && !shift)
1006 shift = 20;
1007
1008 switch (shift) {
1009 case 0:
1010 suffix = "B";
1011 break;
1012 case 10:
1013 suffix = "KB";
1014 break;
1015 case 20:
1016 suffix = "MB";
1017 break;
1018 case 30:
1019 suffix = "GB";
1020 break;
1021 default:
1022 suffix = "XX";
1023 }
1024
1025 for (i = 0; i < nstats; i++) {
1026 int j;
1027 boolean_t bytes = B_FALSE;
1028
1029 for (j = 0; bytestats[j]; j++) {
1030 if (strcmp(stats[i].name, bytestats[j]) == 0) {
1031 bytes = B_TRUE;
1032 break;
1033 }
1034 }
1035
1036 if (bytes) {
1037 mdb_printf("%-25s = %9llu %s\n", stats[i].name,
1038 stats[i].value.ui64 >> shift, suffix);
1039 } else {
1040 mdb_printf("%-25s = %9llu\n", stats[i].name,
1041 stats[i].value.ui64);
1042 }
1043 }
1044
1045 for (i = 0; extras[i]; i++) {
1046 uint64_t buf;
1047
1048 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, extras[i], &sym) == -1) {
1049 mdb_warn("failed to find '%s'", extras[i]);
1050 return (DCMD_ERR);
1051 }
1052
1053 if (sym.st_size != sizeof (uint64_t) &&
1054 sym.st_size != sizeof (uint32_t)) {
1055 mdb_warn("expected scalar for variable '%s'\n",
1056 extras[i]);
1057 return (DCMD_ERR);
1058 }
1059
1060 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) {
1061 mdb_warn("couldn't read '%s'", extras[i]);
1062 return (DCMD_ERR);
1063 }
1064
1065 mdb_printf("%-25s = ", extras[i]);
1066
1067 /* NB: all the 64-bit extras happen to be byte counts */
1068 if (sym.st_size == sizeof (uint64_t))
1069 mdb_printf("%9llu %s\n", buf >> shift, suffix);
1070
1071 if (sym.st_size == sizeof (uint32_t))
1072 mdb_printf("%9d\n", *((uint32_t *)&buf));
1073 }
1074 return (DCMD_OK);
1075 }
1076
1077 typedef struct mdb_spa_print {
1078 pool_state_t spa_state;
1079 char spa_name[MAXNAMELEN];
1080 } mdb_spa_print_t;
1081
1082 /*
1083 * ::spa
1084 *
1085 * -c Print configuration information as well
1086 * -v Print vdev state
1087 * -e Print vdev error stats
1088 * -m Print vdev metaslab info
1089 * -M print vdev metaslab group info
1090 * -h Print histogram info (must be combined with -m or -M)
1091 *
1092 * Print a summarized spa_t. When given no arguments, prints out a table of all
1093 * active pools on the system.
1094 */
1095 /* ARGSUSED */
1096 static int
1097 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1098 {
1099 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED",
1100 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" };
1101 const char *state;
1102 int spa_flags = 0;
1103
1104 if (mdb_getopts(argc, argv,
1105 'c', MDB_OPT_SETBITS, SPA_FLAG_CONFIG, &spa_flags,
1106 'v', MDB_OPT_SETBITS, SPA_FLAG_VDEVS, &spa_flags,
1107 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags,
1108 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags,
1109 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags,
1110 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags,
1111 NULL) != argc)
1112 return (DCMD_USAGE);
1113
1114 if (!(flags & DCMD_ADDRSPEC)) {
1115 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) {
1116 mdb_warn("can't walk spa");
1117 return (DCMD_ERR);
1118 }
1119
1120 return (DCMD_OK);
1121 }
1122
1123 if (flags & DCMD_PIPE_OUT) {
1124 mdb_printf("%#lr\n", addr);
1125 return (DCMD_OK);
1126 }
1127
1128 if (DCMD_HDRSPEC(flags))
1129 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE",
1130 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME");
1131
1132 mdb_spa_print_t spa;
1133 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_print_t", addr, 0) == -1)
1134 return (DCMD_ERR);
1135
1136 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL)
1137 state = "UNKNOWN";
1138 else
1139 state = statetab[spa.spa_state];
1140
1141 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name);
1142
1143 if (spa_flags & SPA_FLAG_CONFIG) {
1144 mdb_printf("\n");
1145 mdb_inc_indent(4);
1146 if (mdb_call_dcmd("spa_config", addr, flags, 0,
1147 NULL) != DCMD_OK)
1148 return (DCMD_ERR);
1149 mdb_dec_indent(4);
1150 }
1151
1152 if (spa_flags & SPA_FLAG_ALL_VDEV) {
1153 mdb_arg_t v;
1154 char opts[100] = "-";
1155 int args =
1156 (spa_flags | SPA_FLAG_VDEVS) == SPA_FLAG_VDEVS ? 0 : 1;
1157
1158 if (spa_flags & SPA_FLAG_ERRORS)
1159 strcat(opts, "e");
1160 if (spa_flags & SPA_FLAG_METASLABS)
1161 strcat(opts, "m");
1162 if (spa_flags & SPA_FLAG_METASLAB_GROUPS)
1163 strcat(opts, "M");
1164 if (spa_flags & SPA_FLAG_HISTOGRAMS)
1165 strcat(opts, "h");
1166
1167 v.a_type = MDB_TYPE_STRING;
1168 v.a_un.a_str = opts;
1169
1170 mdb_printf("\n");
1171 mdb_inc_indent(4);
1172 if (mdb_call_dcmd("spa_vdevs", addr, flags, args,
1173 &v) != DCMD_OK)
1174 return (DCMD_ERR);
1175 mdb_dec_indent(4);
1176 }
1177
1178 return (DCMD_OK);
1179 }
1180
1181 typedef struct mdb_spa_config_spa {
1182 uintptr_t spa_config;
1183 } mdb_spa_config_spa_t;
1184
1185 /*
1186 * ::spa_config
1187 *
1188 * Given a spa_t, print the configuration information stored in spa_config.
1189 * Since it's just an nvlist, format it as an indented list of name=value pairs.
1190 * We simply read the value of spa_config and pass off to ::nvlist.
1191 */
1192 /* ARGSUSED */
1193 static int
1194 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1195 {
1196 mdb_spa_config_spa_t spa;
1197
1198 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1199 return (DCMD_USAGE);
1200
1201 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_config_spa_t",
1202 addr, 0) == -1)
1203 return (DCMD_ERR);
1204
1205 if (spa.spa_config == 0) {
1206 mdb_printf("(none)\n");
1207 return (DCMD_OK);
1208 }
1209
1210 return (mdb_call_dcmd("nvlist", spa.spa_config, flags,
1211 0, NULL));
1212 }
1213
1214 const char histo_stars[] = "****************************************";
1215 const int histo_width = sizeof (histo_stars) - 1;
1216
1217 static void
1218 dump_histogram(const uint64_t *histo, int size, int offset)
1219 {
1220 int i;
1221 int minidx = size - 1;
1222 int maxidx = 0;
1223 uint64_t max = 0;
1224
1225 for (i = 0; i < size; i++) {
1226 if (histo[i] > max)
1227 max = histo[i];
1228 if (histo[i] > 0 && i > maxidx)
1229 maxidx = i;
1230 if (histo[i] > 0 && i < minidx)
1231 minidx = i;
1232 }
1233
1234 if (max < histo_width)
1235 max = histo_width;
1236
1237 for (i = minidx; i <= maxidx; i++) {
1238 mdb_printf("%3u: %6llu %s\n",
1239 i + offset, (u_longlong_t)histo[i],
1240 &histo_stars[(max - histo[i]) * histo_width / max]);
1241 }
1242 }
1243
1244 typedef struct mdb_range_tree {
1245 uint64_t rt_space;
1246 } mdb_range_tree_t;
1247
1248 typedef struct mdb_metaslab_group {
1249 uint64_t mg_fragmentation;
1250 uint64_t mg_histogram[RANGE_TREE_HISTOGRAM_SIZE];
1251 } mdb_metaslab_group_t;
1252
1253 typedef struct mdb_metaslab {
1254 uint64_t ms_id;
1255 uint64_t ms_start;
1256 uint64_t ms_size;
1257 uint64_t ms_fragmentation;
1258 uintptr_t ms_alloctree[TXG_SIZE];
1259 uintptr_t ms_freetree[TXG_SIZE];
1260 uintptr_t ms_tree;
1261 uintptr_t ms_sm;
1262 } mdb_metaslab_t;
1263
1264 typedef struct mdb_space_map_phys_t {
1265 uint64_t smp_alloc;
1266 uint64_t smp_histogram[SPACE_MAP_HISTOGRAM_SIZE];
1267 } mdb_space_map_phys_t;
1268
1269 typedef struct mdb_space_map {
1270 uint64_t sm_size;
1271 uint8_t sm_shift;
1272 uint64_t sm_alloc;
1273 uintptr_t sm_phys;
1274 } mdb_space_map_t;
1275
1276 typedef struct mdb_vdev {
1277 uintptr_t vdev_ms;
1278 uint64_t vdev_ms_count;
1279 vdev_stat_t vdev_stat;
1280 } mdb_vdev_t;
1281
1282 static int
1283 metaslab_stats(uintptr_t addr, int spa_flags)
1284 {
1285 mdb_vdev_t vdev;
1286 uintptr_t *vdev_ms;
1287
1288 if (mdb_ctf_vread(&vdev, "vdev_t", "mdb_vdev_t",
1289 (uintptr_t)addr, 0) == -1) {
1290 mdb_warn("failed to read vdev at %p\n", addr);
1291 return (DCMD_ERR);
1292 }
1293
1294 mdb_inc_indent(4);
1295 mdb_printf("%<u>%-?s %6s %20s %10s %9s%</u>\n", "ADDR", "ID",
1296 "OFFSET", "FREE", "FRAGMENTATION");
1297
1298 vdev_ms = mdb_alloc(vdev.vdev_ms_count * sizeof (void *),
1299 UM_SLEEP | UM_GC);
1300 if (mdb_vread(vdev_ms, vdev.vdev_ms_count * sizeof (void *),
1301 (uintptr_t)vdev.vdev_ms) == -1) {
1302 mdb_warn("failed to read vdev_ms at %p\n", vdev.vdev_ms);
1303 return (DCMD_ERR);
1304 }
1305
1306 for (int m = 0; m < vdev.vdev_ms_count; m++) {
1307 mdb_metaslab_t ms;
1308 mdb_space_map_t sm = { 0 };
1309 char free[NICENUM_BUFLEN];
1310
1311 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t",
1312 (uintptr_t)vdev_ms[m], 0) == -1)
1313 return (DCMD_ERR);
1314
1315 if (ms.ms_sm != NULL &&
1316 mdb_ctf_vread(&sm, "space_map_t", "mdb_space_map_t",
1317 ms.ms_sm, 0) == -1)
1318 return (DCMD_ERR);
1319
1320 mdb_nicenum(ms.ms_size - sm.sm_alloc, free);
1321
1322 mdb_printf("%0?p %6llu %20llx %10s ", vdev_ms[m], ms.ms_id,
1323 ms.ms_start, free);
1324 if (ms.ms_fragmentation == ZFS_FRAG_INVALID)
1325 mdb_printf("%9s\n", "-");
1326 else
1327 mdb_printf("%9llu%%\n", ms.ms_fragmentation);
1328
1329 if ((spa_flags & SPA_FLAG_HISTOGRAMS) && ms.ms_sm != NULL) {
1330 mdb_space_map_phys_t smp;
1331
1332 if (sm.sm_phys == NULL)
1333 continue;
1334
1335 (void) mdb_ctf_vread(&smp, "space_map_phys_t",
1336 "mdb_space_map_phys_t", sm.sm_phys, 0);
1337
1338 dump_histogram(smp.smp_histogram,
1339 SPACE_MAP_HISTOGRAM_SIZE, sm.sm_shift);
1340 }
1341 }
1342 mdb_dec_indent(4);
1343 return (DCMD_OK);
1344 }
1345
1346 static int
1347 metaslab_group_stats(uintptr_t addr, int spa_flags)
1348 {
1349 mdb_metaslab_group_t mg;
1350 if (mdb_ctf_vread(&mg, "metaslab_group_t", "mdb_metaslab_group_t",
1351 (uintptr_t)addr, 0) == -1) {
1352 mdb_warn("failed to read vdev_mg at %p\n", addr);
1353 return (DCMD_ERR);
1354 }
1355
1356 mdb_inc_indent(4);
1357 mdb_printf("%<u>%-?s %15s%</u>\n", "ADDR", "FRAGMENTATION");
1358 if (mg.mg_fragmentation == ZFS_FRAG_INVALID)
1359 mdb_printf("%0?p %15s\n", addr, "-");
1360 else
1361 mdb_printf("%0?p %15llu%%\n", addr, mg.mg_fragmentation);
1362
1363 if (spa_flags & SPA_FLAG_HISTOGRAMS)
1364 dump_histogram(mg.mg_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0);
1365 mdb_dec_indent(4);
1366 return (DCMD_OK);
1367 }
1368
1369 /*
1370 * ::vdev
1371 *
1372 * Print out a summarized vdev_t, in the following form:
1373 *
1374 * ADDR STATE AUX DESC
1375 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0
1376 *
1377 * If '-r' is specified, recursively visit all children.
1378 *
1379 * With '-e', the statistics associated with the vdev are printed as well.
1380 */
1381 static int
1382 do_print_vdev(uintptr_t addr, int flags, int depth, boolean_t recursive,
1383 int spa_flags)
1384 {
1385 vdev_t vdev;
1386 char desc[MAXNAMELEN];
1387 int c, children;
1388 uintptr_t *child;
1389 const char *state, *aux;
1390
1391 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) {
1392 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr);
1393 return (DCMD_ERR);
1394 }
1395
1396 if (flags & DCMD_PIPE_OUT) {
1397 mdb_printf("%#lr\n", addr);
1398 } else {
1399 if (vdev.vdev_path != NULL) {
1400 if (mdb_readstr(desc, sizeof (desc),
1401 (uintptr_t)vdev.vdev_path) == -1) {
1402 mdb_warn("failed to read vdev_path at %p\n",
1403 vdev.vdev_path);
1404 return (DCMD_ERR);
1405 }
1406 } else if (vdev.vdev_ops != NULL) {
1407 vdev_ops_t ops;
1408 if (mdb_vread(&ops, sizeof (ops),
1409 (uintptr_t)vdev.vdev_ops) == -1) {
1410 mdb_warn("failed to read vdev_ops at %p\n",
1411 vdev.vdev_ops);
1412 return (DCMD_ERR);
1413 }
1414 (void) strcpy(desc, ops.vdev_op_type);
1415 } else {
1416 (void) strcpy(desc, "<unknown>");
1417 }
1418
1419 if (depth == 0 && DCMD_HDRSPEC(flags))
1420 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n",
1421 "ADDR", "STATE", "AUX",
1422 sizeof (uintptr_t) == 4 ? 43 : 35,
1423 "DESCRIPTION");
1424
1425 mdb_printf("%0?p ", addr);
1426
1427 switch (vdev.vdev_state) {
1428 case VDEV_STATE_CLOSED:
1429 state = "CLOSED";
1430 break;
1431 case VDEV_STATE_OFFLINE:
1432 state = "OFFLINE";
1433 break;
1434 case VDEV_STATE_CANT_OPEN:
1435 state = "CANT_OPEN";
1436 break;
1437 case VDEV_STATE_DEGRADED:
1438 state = "DEGRADED";
1439 break;
1440 case VDEV_STATE_HEALTHY:
1441 state = "HEALTHY";
1442 break;
1443 case VDEV_STATE_REMOVED:
1444 state = "REMOVED";
1445 break;
1446 case VDEV_STATE_FAULTED:
1447 state = "FAULTED";
1448 break;
1449 default:
1450 state = "UNKNOWN";
1451 break;
1452 }
1453
1454 switch (vdev.vdev_stat.vs_aux) {
1455 case VDEV_AUX_NONE:
1456 aux = "-";
1457 break;
1458 case VDEV_AUX_OPEN_FAILED:
1459 aux = "OPEN_FAILED";
1460 break;
1461 case VDEV_AUX_CORRUPT_DATA:
1462 aux = "CORRUPT_DATA";
1463 break;
1464 case VDEV_AUX_NO_REPLICAS:
1465 aux = "NO_REPLICAS";
1466 break;
1467 case VDEV_AUX_BAD_GUID_SUM:
1468 aux = "BAD_GUID_SUM";
1469 break;
1470 case VDEV_AUX_TOO_SMALL:
1471 aux = "TOO_SMALL";
1472 break;
1473 case VDEV_AUX_BAD_LABEL:
1474 aux = "BAD_LABEL";
1475 break;
1476 case VDEV_AUX_VERSION_NEWER:
1477 aux = "VERS_NEWER";
1478 break;
1479 case VDEV_AUX_VERSION_OLDER:
1480 aux = "VERS_OLDER";
1481 break;
1482 case VDEV_AUX_UNSUP_FEAT:
1483 aux = "UNSUP_FEAT";
1484 break;
1485 case VDEV_AUX_SPARED:
1486 aux = "SPARED";
1487 break;
1488 case VDEV_AUX_ERR_EXCEEDED:
1489 aux = "ERR_EXCEEDED";
1490 break;
1491 case VDEV_AUX_IO_FAILURE:
1492 aux = "IO_FAILURE";
1493 break;
1494 case VDEV_AUX_BAD_LOG:
1495 aux = "BAD_LOG";
1496 break;
1497 case VDEV_AUX_EXTERNAL:
1498 aux = "EXTERNAL";
1499 break;
1500 case VDEV_AUX_SPLIT_POOL:
1501 aux = "SPLIT_POOL";
1502 break;
1503 default:
1504 aux = "UNKNOWN";
1505 break;
1506 }
1507
1508 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc);
1509
1510 if (spa_flags & SPA_FLAG_ERRORS) {
1511 vdev_stat_t *vs = &vdev.vdev_stat;
1512 int i;
1513
1514 mdb_inc_indent(4);
1515 mdb_printf("\n");
1516 mdb_printf("%<u> %12s %12s %12s %12s "
1517 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM",
1518 "IOCTL");
1519 mdb_printf("OPS ");
1520 for (i = 1; i < ZIO_TYPES; i++)
1521 mdb_printf("%11#llx%s", vs->vs_ops[i],
1522 i == ZIO_TYPES - 1 ? "" : " ");
1523 mdb_printf("\n");
1524 mdb_printf("BYTES ");
1525 for (i = 1; i < ZIO_TYPES; i++)
1526 mdb_printf("%11#llx%s", vs->vs_bytes[i],
1527 i == ZIO_TYPES - 1 ? "" : " ");
1528
1529
1530 mdb_printf("\n");
1531 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors);
1532 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors);
1533 mdb_printf("ECKSUM %10#llx\n",
1534 vs->vs_checksum_errors);
1535 mdb_dec_indent(4);
1536 mdb_printf("\n");
1537 }
1538
1539 if (spa_flags & SPA_FLAG_METASLAB_GROUPS &&
1540 vdev.vdev_mg != NULL) {
1541 metaslab_group_stats((uintptr_t)vdev.vdev_mg,
1542 spa_flags);
1543 }
1544 if (spa_flags & SPA_FLAG_METASLABS && vdev.vdev_ms != NULL) {
1545 metaslab_stats((uintptr_t)addr, spa_flags);
1546 }
1547 }
1548
1549 children = vdev.vdev_children;
1550
1551 if (children == 0 || !recursive)
1552 return (DCMD_OK);
1553
1554 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC);
1555 if (mdb_vread(child, children * sizeof (void *),
1556 (uintptr_t)vdev.vdev_child) == -1) {
1557 mdb_warn("failed to read vdev children at %p", vdev.vdev_child);
1558 return (DCMD_ERR);
1559 }
1560
1561 for (c = 0; c < children; c++) {
1562 if (do_print_vdev(child[c], flags, depth + 2, recursive,
1563 spa_flags)) {
1564 return (DCMD_ERR);
1565 }
1566 }
1567
1568 return (DCMD_OK);
1569 }
1570
1571 static int
1572 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1573 {
1574 uint64_t depth = 0;
1575 boolean_t recursive = B_FALSE;
1576 int spa_flags = 0;
1577
1578 if (mdb_getopts(argc, argv,
1579 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags,
1580 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags,
1581 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags,
1582 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags,
1583 'r', MDB_OPT_SETBITS, TRUE, &recursive,
1584 'd', MDB_OPT_UINT64, &depth, NULL) != argc)
1585 return (DCMD_USAGE);
1586
1587 if (!(flags & DCMD_ADDRSPEC)) {
1588 mdb_warn("no vdev_t address given\n");
1589 return (DCMD_ERR);
1590 }
1591
1592 return (do_print_vdev(addr, flags, (int)depth, recursive, spa_flags));
1593 }
1594
1595 typedef struct metaslab_walk_data {
1596 uint64_t mw_numvdevs;
1597 uintptr_t *mw_vdevs;
1598 int mw_curvdev;
1599 uint64_t mw_nummss;
1600 uintptr_t *mw_mss;
1601 int mw_curms;
1602 } metaslab_walk_data_t;
1603
1604 static int
1605 metaslab_walk_step(mdb_walk_state_t *wsp)
1606 {
1607 metaslab_walk_data_t *mw = wsp->walk_data;
1608 metaslab_t ms;
1609 uintptr_t msp;
1610
1611 if (mw->mw_curvdev >= mw->mw_numvdevs)
1612 return (WALK_DONE);
1613
1614 if (mw->mw_mss == NULL) {
1615 uintptr_t mssp;
1616 uintptr_t vdevp;
1617
1618 ASSERT(mw->mw_curms == 0);
1619 ASSERT(mw->mw_nummss == 0);
1620
1621 vdevp = mw->mw_vdevs[mw->mw_curvdev];
1622 if (GETMEMB(vdevp, "vdev", vdev_ms, mssp) ||
1623 GETMEMB(vdevp, "vdev", vdev_ms_count, mw->mw_nummss)) {
1624 return (WALK_ERR);
1625 }
1626
1627 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*),
1628 UM_SLEEP | UM_GC);
1629 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*),
1630 mssp) == -1) {
1631 mdb_warn("failed to read vdev_ms at %p", mssp);
1632 return (WALK_ERR);
1633 }
1634 }
1635
1636 if (mw->mw_curms >= mw->mw_nummss) {
1637 mw->mw_mss = NULL;
1638 mw->mw_curms = 0;
1639 mw->mw_nummss = 0;
1640 mw->mw_curvdev++;
1641 return (WALK_NEXT);
1642 }
1643
1644 msp = mw->mw_mss[mw->mw_curms];
1645 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) {
1646 mdb_warn("failed to read metaslab_t at %p", msp);
1647 return (WALK_ERR);
1648 }
1649
1650 mw->mw_curms++;
1651
1652 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata));
1653 }
1654
1655 /* ARGSUSED */
1656 static int
1657 metaslab_walk_init(mdb_walk_state_t *wsp)
1658 {
1659 metaslab_walk_data_t *mw;
1660 uintptr_t root_vdevp;
1661 uintptr_t childp;
1662
1663 if (wsp->walk_addr == NULL) {
1664 mdb_warn("must supply address of spa_t\n");
1665 return (WALK_ERR);
1666 }
1667
1668 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC);
1669
1670 if (GETMEMB(wsp->walk_addr, "spa", spa_root_vdev, root_vdevp) ||
1671 GETMEMB(root_vdevp, "vdev", vdev_children, mw->mw_numvdevs) ||
1672 GETMEMB(root_vdevp, "vdev", vdev_child, childp)) {
1673 return (DCMD_ERR);
1674 }
1675
1676 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *),
1677 UM_SLEEP | UM_GC);
1678 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *),
1679 childp) == -1) {
1680 mdb_warn("failed to read root vdev children at %p", childp);
1681 return (DCMD_ERR);
1682 }
1683
1684 wsp->walk_data = mw;
1685
1686 return (WALK_NEXT);
1687 }
1688
1689 typedef struct mdb_spa {
1690 uintptr_t spa_dsl_pool;
1691 uintptr_t spa_root_vdev;
1692 } mdb_spa_t;
1693
1694 typedef struct mdb_dsl_dir {
1695 uintptr_t dd_phys;
1696 int64_t dd_space_towrite[TXG_SIZE];
1697 } mdb_dsl_dir_t;
1698
1699 typedef struct mdb_dsl_dir_phys {
1700 uint64_t dd_used_bytes;
1701 uint64_t dd_compressed_bytes;
1702 uint64_t dd_uncompressed_bytes;
1703 } mdb_dsl_dir_phys_t;
1704
1705 typedef struct space_data {
1706 uint64_t ms_alloctree[TXG_SIZE];
1707 uint64_t ms_freetree[TXG_SIZE];
1708 uint64_t ms_tree;
1709 uint64_t avail;
1710 uint64_t nowavail;
1711 } space_data_t;
1712
1713 /* ARGSUSED */
1714 static int
1715 space_cb(uintptr_t addr, const void *unknown, void *arg)
1716 {
1717 space_data_t *sd = arg;
1718 mdb_metaslab_t ms;
1719 mdb_range_tree_t rt;
1720 mdb_space_map_t sm = { 0 };
1721 mdb_space_map_phys_t smp = { 0 };
1722 int i;
1723
1724 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t",
1725 addr, 0) == -1)
1726 return (WALK_ERR);
1727
1728 for (i = 0; i < TXG_SIZE; i++) {
1729 if (mdb_ctf_vread(&rt, "range_tree_t",
1730 "mdb_range_tree_t", ms.ms_alloctree[i], 0) == -1)
1731 return (WALK_ERR);
1732
1733 sd->ms_alloctree[i] += rt.rt_space;
1734
1735 if (mdb_ctf_vread(&rt, "range_tree_t",
1736 "mdb_range_tree_t", ms.ms_freetree[i], 0) == -1)
1737 return (WALK_ERR);
1738
1739 sd->ms_freetree[i] += rt.rt_space;
1740 }
1741
1742 if (mdb_ctf_vread(&rt, "range_tree_t",
1743 "mdb_range_tree_t", ms.ms_tree, 0) == -1)
1744 return (WALK_ERR);
1745
1746 if (ms.ms_sm != NULL &&
1747 mdb_ctf_vread(&sm, "space_map_t",
1748 "mdb_space_map_t", ms.ms_sm, 0) == -1)
1749 return (WALK_ERR);
1750
1751 if (sm.sm_phys != NULL) {
1752 (void) mdb_ctf_vread(&smp, "space_map_phys_t",
1753 "mdb_space_map_phys_t", sm.sm_phys, 0);
1754 }
1755
1756 sd->ms_tree += rt.rt_space;
1757 sd->avail += sm.sm_size - sm.sm_alloc;
1758 sd->nowavail += sm.sm_size - smp.smp_alloc;
1759
1760 return (WALK_NEXT);
1761 }
1762
1763 /*
1764 * ::spa_space [-b]
1765 *
1766 * Given a spa_t, print out it's on-disk space usage and in-core
1767 * estimates of future usage. If -b is given, print space in bytes.
1768 * Otherwise print in megabytes.
1769 */
1770 /* ARGSUSED */
1771 static int
1772 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1773 {
1774 mdb_spa_t spa;
1775 uintptr_t dp_root_dir;
1776 mdb_dsl_dir_t dd;
1777 mdb_dsl_dir_phys_t dsp;
1778 uint64_t children;
1779 uintptr_t childaddr;
1780 space_data_t sd;
1781 int shift = 20;
1782 char *suffix = "M";
1783 int bytes = B_FALSE;
1784
1785 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) !=
1786 argc)
1787 return (DCMD_USAGE);
1788 if (!(flags & DCMD_ADDRSPEC))
1789 return (DCMD_USAGE);
1790
1791 if (bytes) {
1792 shift = 0;
1793 suffix = "";
1794 }
1795
1796 if (GETMEMB(addr, "spa", spa_dsl_pool, spa.spa_dsl_pool) ||
1797 GETMEMB(addr, "spa", spa_root_vdev, spa.spa_root_vdev) ||
1798 GETMEMB(spa.spa_root_vdev, "vdev", vdev_children, children) ||
1799 GETMEMB(spa.spa_root_vdev, "vdev", vdev_child, childaddr) ||
1800 GETMEMB(spa.spa_dsl_pool, "dsl_pool",
1801 dp_root_dir, dp_root_dir) ||
1802 GETMEMB(dp_root_dir, "dsl_dir", dd_phys, dd.dd_phys) ||
1803 GETMEMB(dp_root_dir, "dsl_dir",
1804 dd_space_towrite, dd.dd_space_towrite) ||
1805 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1806 dd_used_bytes, dsp.dd_used_bytes) ||
1807 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1808 dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1809 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1810 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1811 return (DCMD_ERR);
1812 }
1813
1814 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1815 dd.dd_space_towrite[0] >> shift, suffix,
1816 dd.dd_space_towrite[1] >> shift, suffix,
1817 dd.dd_space_towrite[2] >> shift, suffix,
1818 dd.dd_space_towrite[3] >> shift, suffix);
1819
1820 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1821 dsp.dd_used_bytes >> shift, suffix);
1822 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1823 dsp.dd_compressed_bytes >> shift, suffix);
1824 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1825 dsp.dd_uncompressed_bytes >> shift, suffix);
1826
1827 bzero(&sd, sizeof (sd));
1828 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1829 mdb_warn("can't walk metaslabs");
1830 return (DCMD_ERR);
1831 }
1832
1833 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1834 sd.ms_alloctree[0] >> shift, suffix,
1835 sd.ms_alloctree[1] >> shift, suffix,
1836 sd.ms_alloctree[2] >> shift, suffix,
1837 sd.ms_alloctree[3] >> shift, suffix);
1838 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1839 sd.ms_freetree[0] >> shift, suffix,
1840 sd.ms_freetree[1] >> shift, suffix,
1841 sd.ms_freetree[2] >> shift, suffix,
1842 sd.ms_freetree[3] >> shift, suffix);
1843 mdb_printf("ms_tree = %llu%s\n", sd.ms_tree >> shift, suffix);
1844 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1845 mdb_printf("current syncing avail = %llu%s\n",
1846 sd.nowavail >> shift, suffix);
1847
1848 return (DCMD_OK);
1849 }
1850
1851 typedef struct mdb_spa_aux_vdev {
1852 int sav_count;
1853 uintptr_t sav_vdevs;
1854 } mdb_spa_aux_vdev_t;
1855
1856 typedef struct mdb_spa_vdevs {
1857 uintptr_t spa_root_vdev;
1858 mdb_spa_aux_vdev_t spa_l2cache;
1859 mdb_spa_aux_vdev_t spa_spares;
1860 } mdb_spa_vdevs_t;
1861
1862 static int
1863 spa_print_aux(mdb_spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v,
1864 const char *name)
1865 {
1866 uintptr_t *aux;
1867 size_t len;
1868 int ret, i;
1869
1870 /*
1871 * Iterate over aux vdevs and print those out as well. This is a
1872 * little annoying because we don't have a root vdev to pass to ::vdev.
1873 * Instead, we print a single line and then call it for each child
1874 * vdev.
1875 */
1876 if (sav->sav_count != 0) {
1877 v[1].a_type = MDB_TYPE_STRING;
1878 v[1].a_un.a_str = "-d";
1879 v[2].a_type = MDB_TYPE_IMMEDIATE;
1880 v[2].a_un.a_val = 2;
1881
1882 len = sav->sav_count * sizeof (uintptr_t);
1883 aux = mdb_alloc(len, UM_SLEEP);
1884 if (mdb_vread(aux, len, sav->sav_vdevs) == -1) {
1885 mdb_free(aux, len);
1886 mdb_warn("failed to read l2cache vdevs at %p",
1887 sav->sav_vdevs);
1888 return (DCMD_ERR);
1889 }
1890
1891 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name);
1892
1893 for (i = 0; i < sav->sav_count; i++) {
1894 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v);
1895 if (ret != DCMD_OK) {
1896 mdb_free(aux, len);
1897 return (ret);
1898 }
1899 }
1900
1901 mdb_free(aux, len);
1902 }
1903
1904 return (0);
1905 }
1906
1907 /*
1908 * ::spa_vdevs
1909 *
1910 * -e Include error stats
1911 * -m Include metaslab information
1912 * -M Include metaslab group information
1913 * -h Include histogram information (requires -m or -M)
1914 *
1915 * Print out a summarized list of vdevs for the given spa_t.
1916 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as
1917 * iterating over the cache devices.
1918 */
1919 /* ARGSUSED */
1920 static int
1921 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1922 {
1923 mdb_arg_t v[3];
1924 int ret;
1925 char opts[100] = "-r";
1926 int spa_flags = 0;
1927
1928 if (mdb_getopts(argc, argv,
1929 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags,
1930 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags,
1931 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags,
1932 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags,
1933 NULL) != argc)
1934 return (DCMD_USAGE);
1935
1936 if (!(flags & DCMD_ADDRSPEC))
1937 return (DCMD_USAGE);
1938
1939 mdb_spa_vdevs_t spa;
1940 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_vdevs_t", addr, 0) == -1)
1941 return (DCMD_ERR);
1942
1943 /*
1944 * Unitialized spa_t structures can have a NULL root vdev.
1945 */
1946 if (spa.spa_root_vdev == NULL) {
1947 mdb_printf("no associated vdevs\n");
1948 return (DCMD_OK);
1949 }
1950
1951 if (spa_flags & SPA_FLAG_ERRORS)
1952 strcat(opts, "e");
1953 if (spa_flags & SPA_FLAG_METASLABS)
1954 strcat(opts, "m");
1955 if (spa_flags & SPA_FLAG_METASLAB_GROUPS)
1956 strcat(opts, "M");
1957 if (spa_flags & SPA_FLAG_HISTOGRAMS)
1958 strcat(opts, "h");
1959
1960 v[0].a_type = MDB_TYPE_STRING;
1961 v[0].a_un.a_str = opts;
1962
1963 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1964 flags, 1, v);
1965 if (ret != DCMD_OK)
1966 return (ret);
1967
1968 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 ||
1969 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0)
1970 return (DCMD_ERR);
1971
1972 return (DCMD_OK);
1973 }
1974
1975 /*
1976 * ::zio
1977 *
1978 * Print a summary of zio_t and all its children. This is intended to display a
1979 * zio tree, and hence we only pick the most important pieces of information for
1980 * the main summary. More detailed information can always be found by doing a
1981 * '::print zio' on the underlying zio_t. The columns we display are:
1982 *
1983 * ADDRESS TYPE STAGE WAITER TIME_ELAPSED
1984 *
1985 * The 'address' column is indented by one space for each depth level as we
1986 * descend down the tree.
1987 */
1988
1989 #define ZIO_MAXINDENT 7
1990 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT)
1991 #define ZIO_WALK_SELF 0
1992 #define ZIO_WALK_CHILD 1
1993 #define ZIO_WALK_PARENT 2
1994
1995 typedef struct zio_print_args {
1996 int zpa_current_depth;
1997 int zpa_min_depth;
1998 int zpa_max_depth;
1999 int zpa_type;
2000 uint_t zpa_flags;
2001 } zio_print_args_t;
2002
2003 typedef struct mdb_zio {
2004 enum zio_type io_type;
2005 enum zio_stage io_stage;
2006 uintptr_t io_waiter;
2007 uintptr_t io_spa;
2008 struct {
2009 struct {
2010 uintptr_t list_next;
2011 } list_head;
2012 } io_parent_list;
2013 int io_error;
2014 } mdb_zio_t;
2015
2016 typedef struct mdb_zio_timestamp {
2017 hrtime_t io_timestamp;
2018 } mdb_zio_timestamp_t;
2019
2020 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg);
2021
2022 static int
2023 zio_print_cb(uintptr_t addr, zio_print_args_t *zpa)
2024 {
2025 mdb_ctf_id_t type_enum, stage_enum;
2026 int indent = zpa->zpa_current_depth;
2027 const char *type, *stage;
2028 uintptr_t laddr;
2029 mdb_zio_t zio;
2030 mdb_zio_timestamp_t zio_timestamp = { 0 };
2031
2032 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1)
2033 return (WALK_ERR);
2034 (void) mdb_ctf_vread(&zio_timestamp, ZFS_STRUCT "zio",
2035 "mdb_zio_timestamp_t", addr, MDB_CTF_VREAD_QUIET);
2036
2037 if (indent > ZIO_MAXINDENT)
2038 indent = ZIO_MAXINDENT;
2039
2040 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 ||
2041 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) {
2042 mdb_warn("failed to lookup zio enums");
2043 return (WALK_ERR);
2044 }
2045
2046 if ((type = mdb_ctf_enum_name(type_enum, zio.io_type)) != NULL)
2047 type += sizeof ("ZIO_TYPE_") - 1;
2048 else
2049 type = "?";
2050
2051 if (zio.io_error == 0) {
2052 stage = mdb_ctf_enum_name(stage_enum, zio.io_stage);
2053 if (stage != NULL)
2054 stage += sizeof ("ZIO_STAGE_") - 1;
2055 else
2056 stage = "?";
2057 } else {
2058 stage = "FAILED";
2059 }
2060
2061 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) {
2062 if (zpa->zpa_flags & DCMD_PIPE_OUT) {
2063 mdb_printf("%?p\n", addr);
2064 } else {
2065 mdb_printf("%*s%-*p %-5s %-16s ", indent, "",
2066 ZIO_MAXWIDTH - indent, addr, type, stage);
2067 if (zio.io_waiter != 0)
2068 mdb_printf("%-16lx ", zio.io_waiter);
2069 else
2070 mdb_printf("%-16s ", "-");
2071 #ifdef _KERNEL
2072 if (zio_timestamp.io_timestamp != 0) {
2073 mdb_printf("%llums", (mdb_gethrtime() -
2074 zio_timestamp.io_timestamp) /
2075 1000000);
2076 } else {
2077 mdb_printf("%-12s ", "-");
2078 }
2079 #else
2080 mdb_printf("%-12s ", "-");
2081 #endif
2082 mdb_printf("\n");
2083 }
2084 }
2085
2086 if (zpa->zpa_current_depth >= zpa->zpa_max_depth)
2087 return (WALK_NEXT);
2088
2089 if (zpa->zpa_type == ZIO_WALK_PARENT)
2090 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
2091 "io_parent_list");
2092 else
2093 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
2094 "io_child_list");
2095
2096 zpa->zpa_current_depth++;
2097 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) {
2098 mdb_warn("failed to walk zio_t children at %p\n", laddr);
2099 return (WALK_ERR);
2100 }
2101 zpa->zpa_current_depth--;
2102
2103 return (WALK_NEXT);
2104 }
2105
2106 /* ARGSUSED */
2107 static int
2108 zio_child_cb(uintptr_t addr, const void *unknown, void *arg)
2109 {
2110 zio_link_t zl;
2111 uintptr_t ziop;
2112 zio_print_args_t *zpa = arg;
2113
2114 if (mdb_vread(&zl, sizeof (zl), addr) == -1) {
2115 mdb_warn("failed to read zio_link_t at %p", addr);
2116 return (WALK_ERR);
2117 }
2118
2119 if (zpa->zpa_type == ZIO_WALK_PARENT)
2120 ziop = (uintptr_t)zl.zl_parent;
2121 else
2122 ziop = (uintptr_t)zl.zl_child;
2123
2124 return (zio_print_cb(ziop, zpa));
2125 }
2126
2127 /* ARGSUSED */
2128 static int
2129 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2130 {
2131 zio_print_args_t zpa = { 0 };
2132
2133 if (!(flags & DCMD_ADDRSPEC))
2134 return (DCMD_USAGE);
2135
2136 if (mdb_getopts(argc, argv,
2137 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth,
2138 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type,
2139 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type,
2140 NULL) != argc)
2141 return (DCMD_USAGE);
2142
2143 zpa.zpa_flags = flags;
2144 if (zpa.zpa_max_depth != 0) {
2145 if (zpa.zpa_type == ZIO_WALK_SELF)
2146 zpa.zpa_type = ZIO_WALK_CHILD;
2147 } else if (zpa.zpa_type != ZIO_WALK_SELF) {
2148 zpa.zpa_min_depth = 1;
2149 zpa.zpa_max_depth = 1;
2150 }
2151
2152 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) {
2153 mdb_printf("%<u>%-*s %-5s %-16s %-16s %-12s%</u>\n",
2154 ZIO_MAXWIDTH, "ADDRESS", "TYPE", "STAGE", "WAITER",
2155 "TIME_ELAPSED");
2156 }
2157
2158 if (zio_print_cb(addr, &zpa) != WALK_NEXT)
2159 return (DCMD_ERR);
2160
2161 return (DCMD_OK);
2162 }
2163
2164 /*
2165 * [addr]::zio_state
2166 *
2167 * Print a summary of all zio_t structures on the system, or for a particular
2168 * pool. This is equivalent to '::walk zio_root | ::zio'.
2169 */
2170 /*ARGSUSED*/
2171 static int
2172 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2173 {
2174 /*
2175 * MDB will remember the last address of the pipeline, so if we don't
2176 * zero this we'll end up trying to walk zio structures for a
2177 * non-existent spa_t.
2178 */
2179 if (!(flags & DCMD_ADDRSPEC))
2180 addr = 0;
2181
2182 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr));
2183 }
2184
2185 typedef struct txg_list_walk_data {
2186 uintptr_t lw_head[TXG_SIZE];
2187 int lw_txgoff;
2188 int lw_maxoff;
2189 size_t lw_offset;
2190 void *lw_obj;
2191 } txg_list_walk_data_t;
2192
2193 static int
2194 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
2195 {
2196 txg_list_walk_data_t *lwd;
2197 txg_list_t list;
2198 int i;
2199
2200 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
2201 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
2202 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
2203 return (WALK_ERR);
2204 }
2205
2206 for (i = 0; i < TXG_SIZE; i++)
2207 lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
2208 lwd->lw_offset = list.tl_offset;
2209 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
2210 UM_SLEEP | UM_GC);
2211 lwd->lw_txgoff = txg;
2212 lwd->lw_maxoff = maxoff;
2213
2214 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
2215 wsp->walk_data = lwd;
2216
2217 return (WALK_NEXT);
2218 }
2219
2220 static int
2221 txg_list_walk_init(mdb_walk_state_t *wsp)
2222 {
2223 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
2224 }
2225
2226 static int
2227 txg_list0_walk_init(mdb_walk_state_t *wsp)
2228 {
2229 return (txg_list_walk_init_common(wsp, 0, 0));
2230 }
2231
2232 static int
2233 txg_list1_walk_init(mdb_walk_state_t *wsp)
2234 {
2235 return (txg_list_walk_init_common(wsp, 1, 1));
2236 }
2237
2238 static int
2239 txg_list2_walk_init(mdb_walk_state_t *wsp)
2240 {
2241 return (txg_list_walk_init_common(wsp, 2, 2));
2242 }
2243
2244 static int
2245 txg_list3_walk_init(mdb_walk_state_t *wsp)
2246 {
2247 return (txg_list_walk_init_common(wsp, 3, 3));
2248 }
2249
2250 static int
2251 txg_list_walk_step(mdb_walk_state_t *wsp)
2252 {
2253 txg_list_walk_data_t *lwd = wsp->walk_data;
2254 uintptr_t addr;
2255 txg_node_t *node;
2256 int status;
2257
2258 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
2259 lwd->lw_txgoff++;
2260 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
2261 }
2262
2263 if (wsp->walk_addr == NULL)
2264 return (WALK_DONE);
2265
2266 addr = wsp->walk_addr - lwd->lw_offset;
2267
2268 if (mdb_vread(lwd->lw_obj,
2269 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
2270 mdb_warn("failed to read list element at %#lx", addr);
2271 return (WALK_ERR);
2272 }
2273
2274 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
2275 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
2276 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
2277
2278 return (status);
2279 }
2280
2281 /*
2282 * ::walk spa
2283 *
2284 * Walk all named spa_t structures in the namespace. This is nothing more than
2285 * a layered avl walk.
2286 */
2287 static int
2288 spa_walk_init(mdb_walk_state_t *wsp)
2289 {
2290 GElf_Sym sym;
2291
2292 if (wsp->walk_addr != NULL) {
2293 mdb_warn("spa walk only supports global walks\n");
2294 return (WALK_ERR);
2295 }
2296
2297 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
2298 mdb_warn("failed to find symbol 'spa_namespace_avl'");
2299 return (WALK_ERR);
2300 }
2301
2302 wsp->walk_addr = (uintptr_t)sym.st_value;
2303
2304 if (mdb_layered_walk("avl", wsp) == -1) {
2305 mdb_warn("failed to walk 'avl'\n");
2306 return (WALK_ERR);
2307 }
2308
2309 return (WALK_NEXT);
2310 }
2311
2312 static int
2313 spa_walk_step(mdb_walk_state_t *wsp)
2314 {
2315 return (wsp->walk_callback(wsp->walk_addr, NULL, wsp->walk_cbdata));
2316 }
2317
2318 /*
2319 * [addr]::walk zio
2320 *
2321 * Walk all active zio_t structures on the system. This is simply a layered
2322 * walk on top of ::walk zio_cache, with the optional ability to limit the
2323 * structures to a particular pool.
2324 */
2325 static int
2326 zio_walk_init(mdb_walk_state_t *wsp)
2327 {
2328 wsp->walk_data = (void *)wsp->walk_addr;
2329
2330 if (mdb_layered_walk("zio_cache", wsp) == -1) {
2331 mdb_warn("failed to walk 'zio_cache'\n");
2332 return (WALK_ERR);
2333 }
2334
2335 return (WALK_NEXT);
2336 }
2337
2338 static int
2339 zio_walk_step(mdb_walk_state_t *wsp)
2340 {
2341 mdb_zio_t zio;
2342 uintptr_t spa = (uintptr_t)wsp->walk_data;
2343
2344 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2345 wsp->walk_addr, 0) == -1)
2346 return (WALK_ERR);
2347
2348 if (spa != 0 && spa != zio.io_spa)
2349 return (WALK_NEXT);
2350
2351 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2352 }
2353
2354 /*
2355 * [addr]::walk zio_root
2356 *
2357 * Walk only root zio_t structures, optionally for a particular spa_t.
2358 */
2359 static int
2360 zio_walk_root_step(mdb_walk_state_t *wsp)
2361 {
2362 mdb_zio_t zio;
2363 uintptr_t spa = (uintptr_t)wsp->walk_data;
2364
2365 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2366 wsp->walk_addr, 0) == -1)
2367 return (WALK_ERR);
2368
2369 if (spa != 0 && spa != zio.io_spa)
2370 return (WALK_NEXT);
2371
2372 /* If the parent list is not empty, ignore */
2373 if (zio.io_parent_list.list_head.list_next !=
2374 wsp->walk_addr +
2375 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") +
2376 mdb_ctf_offsetof_by_name("struct list", "list_head"))
2377 return (WALK_NEXT);
2378
2379 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2380 }
2381
2382 /*
2383 * ::zfs_blkstats
2384 *
2385 * -v print verbose per-level information
2386 *
2387 */
2388 static int
2389 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2390 {
2391 boolean_t verbose = B_FALSE;
2392 zfs_all_blkstats_t stats;
2393 dmu_object_type_t t;
2394 zfs_blkstat_t *tzb;
2395 uint64_t ditto;
2396 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10];
2397 /* +10 in case it grew */
2398
2399 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) {
2400 mdb_warn("failed to read 'dmu_ot'");
2401 return (DCMD_ERR);
2402 }
2403
2404 if (mdb_getopts(argc, argv,
2405 'v', MDB_OPT_SETBITS, TRUE, &verbose,
2406 NULL) != argc)
2407 return (DCMD_USAGE);
2408
2409 if (!(flags & DCMD_ADDRSPEC))
2410 return (DCMD_USAGE);
2411
2412 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) ||
2413 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) ||
2414 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) {
2415 mdb_warn("failed to read data at %p;", addr);
2416 mdb_printf("maybe no stats? run \"zpool scrub\" first.");
2417 return (DCMD_ERR);
2418 }
2419
2420 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL];
2421 if (tzb->zb_gangs != 0) {
2422 mdb_printf("Ganged blocks: %llu\n",
2423 (longlong_t)tzb->zb_gangs);
2424 }
2425
2426 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev +
2427 tzb->zb_ditto_3_of_3_samevdev;
2428 if (ditto != 0) {
2429 mdb_printf("Dittoed blocks on same vdev: %llu\n",
2430 (longlong_t)ditto);
2431 }
2432
2433 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
2434 "\t avg\t comp\t%%Total\tType\n");
2435
2436 for (t = 0; t <= DMU_OT_TOTAL; t++) {
2437 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN];
2438 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN];
2439 char avg[NICENUM_BUFLEN];
2440 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN];
2441 char typename[64];
2442 int l;
2443
2444
2445 if (t == DMU_OT_DEFERRED)
2446 strcpy(typename, "deferred free");
2447 else if (t == DMU_OT_OTHER)
2448 strcpy(typename, "other");
2449 else if (t == DMU_OT_TOTAL)
2450 strcpy(typename, "Total");
2451 else if (mdb_readstr(typename, sizeof (typename),
2452 (uintptr_t)dmu_ot[t].ot_name) == -1) {
2453 mdb_warn("failed to read type name");
2454 return (DCMD_ERR);
2455 }
2456
2457 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0)
2458 continue;
2459
2460 for (l = -1; l < DN_MAX_LEVELS; l++) {
2461 int level = (l == -1 ? DN_MAX_LEVELS : l);
2462 zfs_blkstat_t *zb = &stats.zab_type[level][t];
2463
2464 if (zb->zb_asize == 0)
2465 continue;
2466
2467 /*
2468 * Don't print each level unless requested.
2469 */
2470 if (!verbose && level != DN_MAX_LEVELS)
2471 continue;
2472
2473 /*
2474 * If all the space is level 0, don't print the
2475 * level 0 separately.
2476 */
2477 if (level == 0 && zb->zb_asize ==
2478 stats.zab_type[DN_MAX_LEVELS][t].zb_asize)
2479 continue;
2480
2481 mdb_nicenum(zb->zb_count, csize);
2482 mdb_nicenum(zb->zb_lsize, lsize);
2483 mdb_nicenum(zb->zb_psize, psize);
2484 mdb_nicenum(zb->zb_asize, asize);
2485 mdb_nicenum(zb->zb_asize / zb->zb_count, avg);
2486 (void) snprintfrac(comp, NICENUM_BUFLEN,
2487 zb->zb_lsize, zb->zb_psize, 2);
2488 (void) snprintfrac(pct, NICENUM_BUFLEN,
2489 100 * zb->zb_asize, tzb->zb_asize, 2);
2490
2491 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s"
2492 "\t%5s\t%6s\t",
2493 csize, lsize, psize, asize, avg, comp, pct);
2494
2495 if (level == DN_MAX_LEVELS)
2496 mdb_printf("%s\n", typename);
2497 else
2498 mdb_printf(" L%d %s\n",
2499 level, typename);
2500 }
2501 }
2502
2503 return (DCMD_OK);
2504 }
2505
2506 typedef struct mdb_reference {
2507 uintptr_t ref_holder;
2508 uintptr_t ref_removed;
2509 uint64_t ref_number;
2510 } mdb_reference_t;
2511
2512 /* ARGSUSED */
2513 static int
2514 reference_cb(uintptr_t addr, const void *ignored, void *arg)
2515 {
2516 mdb_reference_t ref;
2517 boolean_t holder_is_str = B_FALSE;
2518 char holder_str[128];
2519 boolean_t removed = (boolean_t)arg;
2520
2521 if (mdb_ctf_vread(&ref, "reference_t", "mdb_reference_t", addr,
2522 0) == -1)
2523 return (DCMD_ERR);
2524
2525 if (mdb_readstr(holder_str, sizeof (holder_str),
2526 ref.ref_holder) != -1)
2527 holder_is_str = strisprint(holder_str);
2528
2529 if (removed)
2530 mdb_printf("removed ");
2531 mdb_printf("reference ");
2532 if (ref.ref_number != 1)
2533 mdb_printf("with count=%llu ", ref.ref_number);
2534 mdb_printf("with tag %lx", ref.ref_holder);
2535 if (holder_is_str)
2536 mdb_printf(" \"%s\"", holder_str);
2537 mdb_printf(", held at:\n");
2538
2539 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
2540
2541 if (removed) {
2542 mdb_printf("removed at:\n");
2543 (void) mdb_call_dcmd("whatis", ref.ref_removed,
2544 DCMD_ADDRSPEC, 0, NULL);
2545 }
2546
2547 mdb_printf("\n");
2548
2549 return (WALK_NEXT);
2550 }
2551
2552 typedef struct mdb_refcount {
2553 uint64_t rc_count;
2554 } mdb_refcount_t;
2555
2556 typedef struct mdb_refcount_removed {
2557 uint64_t rc_removed_count;
2558 } mdb_refcount_removed_t;
2559
2560 typedef struct mdb_refcount_tracked {
2561 boolean_t rc_tracked;
2562 } mdb_refcount_tracked_t;
2563
2564 /* ARGSUSED */
2565 static int
2566 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2567 {
2568 mdb_refcount_t rc;
2569 mdb_refcount_removed_t rcr;
2570 mdb_refcount_tracked_t rct;
2571 int off;
2572 boolean_t released = B_FALSE;
2573
2574 if (!(flags & DCMD_ADDRSPEC))
2575 return (DCMD_USAGE);
2576
2577 if (mdb_getopts(argc, argv,
2578 'r', MDB_OPT_SETBITS, B_TRUE, &released,
2579 NULL) != argc)
2580 return (DCMD_USAGE);
2581
2582 if (mdb_ctf_vread(&rc, "refcount_t", "mdb_refcount_t", addr,
2583 0) == -1)
2584 return (DCMD_ERR);
2585
2586 if (mdb_ctf_vread(&rcr, "refcount_t", "mdb_refcount_removed_t", addr,
2587 MDB_CTF_VREAD_QUIET) == -1) {
2588 mdb_printf("refcount_t at %p has %llu holds (untracked)\n",
2589 addr, (longlong_t)rc.rc_count);
2590 return (DCMD_OK);
2591 }
2592
2593 if (mdb_ctf_vread(&rct, "refcount_t", "mdb_refcount_tracked_t", addr,
2594 MDB_CTF_VREAD_QUIET) == -1) {
2595 /* If this is an old target, it might be tracked. */
2596 rct.rc_tracked = B_TRUE;
2597 }
2598
2599 mdb_printf("refcount_t at %p has %llu current holds, "
2600 "%llu recently released holds\n",
2601 addr, (longlong_t)rc.rc_count, (longlong_t)rcr.rc_removed_count);
2602
2603 if (rct.rc_tracked && rc.rc_count > 0)
2604 mdb_printf("current holds:\n");
2605 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_list");
2606 if (off == -1)
2607 return (DCMD_ERR);
2608 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off);
2609
2610 if (released && rcr.rc_removed_count > 0) {
2611 mdb_printf("released holds:\n");
2612
2613 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_removed");
2614 if (off == -1)
2615 return (DCMD_ERR);
2616 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off);
2617 }
2618
2619 return (DCMD_OK);
2620 }
2621
2622 /* ARGSUSED */
2623 static int
2624 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2625 {
2626 sa_attr_table_t *table;
2627 sa_os_t sa_os;
2628 char *name;
2629 int i;
2630
2631 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) {
2632 mdb_warn("failed to read sa_os at %p", addr);
2633 return (DCMD_ERR);
2634 }
2635
2636 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2637 UM_SLEEP | UM_GC);
2638 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC);
2639
2640 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2641 (uintptr_t)sa_os.sa_attr_table) == -1) {
2642 mdb_warn("failed to read sa_os at %p", addr);
2643 return (DCMD_ERR);
2644 }
2645
2646 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n",
2647 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME");
2648 for (i = 0; i != sa_os.sa_num_attrs; i++) {
2649 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name);
2650 mdb_printf("%5x %8x %8x %8x %-s\n",
2651 (int)table[i].sa_attr, (int)table[i].sa_registered,
2652 (int)table[i].sa_length, table[i].sa_byteswap, name);
2653 }
2654
2655 return (DCMD_OK);
2656 }
2657
2658 static int
2659 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count)
2660 {
2661 uintptr_t idx_table;
2662
2663 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) {
2664 mdb_printf("can't find offset table in sa_idx_tab\n");
2665 return (-1);
2666 }
2667
2668 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t),
2669 UM_SLEEP | UM_GC);
2670
2671 if (mdb_vread(*off_tab,
2672 attr_count * sizeof (uint32_t), idx_table) == -1) {
2673 mdb_warn("failed to attribute offset table %p", idx_table);
2674 return (-1);
2675 }
2676
2677 return (DCMD_OK);
2678 }
2679
2680 /*ARGSUSED*/
2681 static int
2682 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2683 {
2684 uint32_t *offset_tab;
2685 int attr_count;
2686 uint64_t attr_id;
2687 uintptr_t attr_addr;
2688 uintptr_t bonus_tab, spill_tab;
2689 uintptr_t db_bonus, db_spill;
2690 uintptr_t os, os_sa;
2691 uintptr_t db_data;
2692
2693 if (argc != 1)
2694 return (DCMD_USAGE);
2695
2696 if (argv[0].a_type == MDB_TYPE_STRING)
2697 attr_id = mdb_strtoull(argv[0].a_un.a_str);
2698 else
2699 return (DCMD_USAGE);
2700
2701 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) ||
2702 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) ||
2703 GETMEMB(addr, "sa_handle", sa_os, os) ||
2704 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) ||
2705 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) {
2706 mdb_printf("Can't find necessary information in sa_handle "
2707 "in sa_handle\n");
2708 return (DCMD_ERR);
2709 }
2710
2711 if (GETMEMB(os, "objset", os_sa, os_sa)) {
2712 mdb_printf("Can't find os_sa in objset\n");
2713 return (DCMD_ERR);
2714 }
2715
2716 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) {
2717 mdb_printf("Can't find sa_num_attrs\n");
2718 return (DCMD_ERR);
2719 }
2720
2721 if (attr_id > attr_count) {
2722 mdb_printf("attribute id number is out of range\n");
2723 return (DCMD_ERR);
2724 }
2725
2726 if (bonus_tab) {
2727 if (sa_get_off_table(bonus_tab, &offset_tab,
2728 attr_count) == -1) {
2729 return (DCMD_ERR);
2730 }
2731
2732 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) {
2733 mdb_printf("can't find db_data in bonus dbuf\n");
2734 return (DCMD_ERR);
2735 }
2736 }
2737
2738 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) &&
2739 spill_tab == NULL) {
2740 mdb_printf("Attribute does not exist\n");
2741 return (DCMD_ERR);
2742 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) {
2743 if (sa_get_off_table(spill_tab, &offset_tab,
2744 attr_count) == -1) {
2745 return (DCMD_ERR);
2746 }
2747 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) {
2748 mdb_printf("can't find db_data in spill dbuf\n");
2749 return (DCMD_ERR);
2750 }
2751 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) {
2752 mdb_printf("Attribute does not exist\n");
2753 return (DCMD_ERR);
2754 }
2755 }
2756 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]);
2757 mdb_printf("%p\n", attr_addr);
2758 return (DCMD_OK);
2759 }
2760
2761 /* ARGSUSED */
2762 static int
2763 zfs_ace_print_common(uintptr_t addr, uint_t flags,
2764 uint64_t id, uint32_t access_mask, uint16_t ace_flags,
2765 uint16_t ace_type, int verbose)
2766 {
2767 if (DCMD_HDRSPEC(flags) && !verbose)
2768 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n",
2769 "ADDR", "FLAGS", "MASK", "TYPE", "ID");
2770
2771 if (!verbose) {
2772 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr,
2773 ace_flags, access_mask, ace_type, id);
2774 return (DCMD_OK);
2775 }
2776
2777 switch (ace_flags & ACE_TYPE_FLAGS) {
2778 case ACE_OWNER:
2779 mdb_printf("owner@:");
2780 break;
2781 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2782 mdb_printf("group@:");
2783 break;
2784 case ACE_EVERYONE:
2785 mdb_printf("everyone@:");
2786 break;
2787 case ACE_IDENTIFIER_GROUP:
2788 mdb_printf("group:%llx:", (u_longlong_t)id);
2789 break;
2790 case 0: /* User entry */
2791 mdb_printf("user:%llx:", (u_longlong_t)id);
2792 break;
2793 }
2794
2795 /* print out permission mask */
2796 if (access_mask & ACE_READ_DATA)
2797 mdb_printf("r");
2798 else
2799 mdb_printf("-");
2800 if (access_mask & ACE_WRITE_DATA)
2801 mdb_printf("w");
2802 else
2803 mdb_printf("-");
2804 if (access_mask & ACE_EXECUTE)
2805 mdb_printf("x");
2806 else
2807 mdb_printf("-");
2808 if (access_mask & ACE_APPEND_DATA)
2809 mdb_printf("p");
2810 else
2811 mdb_printf("-");
2812 if (access_mask & ACE_DELETE)
2813 mdb_printf("d");
2814 else
2815 mdb_printf("-");
2816 if (access_mask & ACE_DELETE_CHILD)
2817 mdb_printf("D");
2818 else
2819 mdb_printf("-");
2820 if (access_mask & ACE_READ_ATTRIBUTES)
2821 mdb_printf("a");
2822 else
2823 mdb_printf("-");
2824 if (access_mask & ACE_WRITE_ATTRIBUTES)
2825 mdb_printf("A");
2826 else
2827 mdb_printf("-");
2828 if (access_mask & ACE_READ_NAMED_ATTRS)
2829 mdb_printf("R");
2830 else
2831 mdb_printf("-");
2832 if (access_mask & ACE_WRITE_NAMED_ATTRS)
2833 mdb_printf("W");
2834 else
2835 mdb_printf("-");
2836 if (access_mask & ACE_READ_ACL)
2837 mdb_printf("c");
2838 else
2839 mdb_printf("-");
2840 if (access_mask & ACE_WRITE_ACL)
2841 mdb_printf("C");
2842 else
2843 mdb_printf("-");
2844 if (access_mask & ACE_WRITE_OWNER)
2845 mdb_printf("o");
2846 else
2847 mdb_printf("-");
2848 if (access_mask & ACE_SYNCHRONIZE)
2849 mdb_printf("s");
2850 else
2851 mdb_printf("-");
2852
2853 mdb_printf(":");
2854
2855 /* Print out inheritance flags */
2856 if (ace_flags & ACE_FILE_INHERIT_ACE)
2857 mdb_printf("f");
2858 else
2859 mdb_printf("-");
2860 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE)
2861 mdb_printf("d");
2862 else
2863 mdb_printf("-");
2864 if (ace_flags & ACE_INHERIT_ONLY_ACE)
2865 mdb_printf("i");
2866 else
2867 mdb_printf("-");
2868 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE)
2869 mdb_printf("n");
2870 else
2871 mdb_printf("-");
2872 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG)
2873 mdb_printf("S");
2874 else
2875 mdb_printf("-");
2876 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG)
2877 mdb_printf("F");
2878 else
2879 mdb_printf("-");
2880 if (ace_flags & ACE_INHERITED_ACE)
2881 mdb_printf("I");
2882 else
2883 mdb_printf("-");
2884
2885 switch (ace_type) {
2886 case ACE_ACCESS_ALLOWED_ACE_TYPE:
2887 mdb_printf(":allow\n");
2888 break;
2889 case ACE_ACCESS_DENIED_ACE_TYPE:
2890 mdb_printf(":deny\n");
2891 break;
2892 case ACE_SYSTEM_AUDIT_ACE_TYPE:
2893 mdb_printf(":audit\n");
2894 break;
2895 case ACE_SYSTEM_ALARM_ACE_TYPE:
2896 mdb_printf(":alarm\n");
2897 break;
2898 default:
2899 mdb_printf(":?\n");
2900 }
2901 return (DCMD_OK);
2902 }
2903
2904 /* ARGSUSED */
2905 static int
2906 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2907 {
2908 zfs_ace_t zace;
2909 int verbose = FALSE;
2910 uint64_t id;
2911
2912 if (!(flags & DCMD_ADDRSPEC))
2913 return (DCMD_USAGE);
2914
2915 if (mdb_getopts(argc, argv,
2916 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2917 return (DCMD_USAGE);
2918
2919 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) {
2920 mdb_warn("failed to read zfs_ace_t");
2921 return (DCMD_ERR);
2922 }
2923
2924 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 ||
2925 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2926 id = zace.z_fuid;
2927 else
2928 id = -1;
2929
2930 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask,
2931 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose));
2932 }
2933
2934 /* ARGSUSED */
2935 static int
2936 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2937 {
2938 ace_t ace;
2939 uint64_t id;
2940 int verbose = FALSE;
2941
2942 if (!(flags & DCMD_ADDRSPEC))
2943 return (DCMD_USAGE);
2944
2945 if (mdb_getopts(argc, argv,
2946 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2947 return (DCMD_USAGE);
2948
2949 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) {
2950 mdb_warn("failed to read ace_t");
2951 return (DCMD_ERR);
2952 }
2953
2954 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 ||
2955 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2956 id = ace.a_who;
2957 else
2958 id = -1;
2959
2960 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask,
2961 ace.a_flags, ace.a_type, verbose));
2962 }
2963
2964 typedef struct acl_dump_args {
2965 int a_argc;
2966 const mdb_arg_t *a_argv;
2967 uint16_t a_version;
2968 int a_flags;
2969 } acl_dump_args_t;
2970
2971 /* ARGSUSED */
2972 static int
2973 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg)
2974 {
2975 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2976
2977 if (acl_args->a_version == 1) {
2978 if (mdb_call_dcmd("zfs_ace", addr,
2979 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2980 acl_args->a_argv) != DCMD_OK) {
2981 return (WALK_ERR);
2982 }
2983 } else {
2984 if (mdb_call_dcmd("zfs_ace0", addr,
2985 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2986 acl_args->a_argv) != DCMD_OK) {
2987 return (WALK_ERR);
2988 }
2989 }
2990 acl_args->a_flags = DCMD_LOOP;
2991 return (WALK_NEXT);
2992 }
2993
2994 /* ARGSUSED */
2995 static int
2996 acl_cb(uintptr_t addr, const void *unknown, void *arg)
2997 {
2998 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2999
3000 if (acl_args->a_version == 1) {
3001 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb,
3002 arg, addr) != 0) {
3003 mdb_warn("can't walk ACEs");
3004 return (DCMD_ERR);
3005 }
3006 } else {
3007 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb,
3008 arg, addr) != 0) {
3009 mdb_warn("can't walk ACEs");
3010 return (DCMD_ERR);
3011 }
3012 }
3013 return (WALK_NEXT);
3014 }
3015
3016 /* ARGSUSED */
3017 static int
3018 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
3019 {
3020 zfs_acl_t zacl;
3021 int verbose = FALSE;
3022 acl_dump_args_t acl_args;
3023
3024 if (!(flags & DCMD_ADDRSPEC))
3025 return (DCMD_USAGE);
3026
3027 if (mdb_getopts(argc, argv,
3028 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
3029 return (DCMD_USAGE);
3030
3031 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) {
3032 mdb_warn("failed to read zfs_acl_t");
3033 return (DCMD_ERR);
3034 }
3035
3036 acl_args.a_argc = argc;
3037 acl_args.a_argv = argv;
3038 acl_args.a_version = zacl.z_version;
3039 acl_args.a_flags = DCMD_LOOPFIRST;
3040
3041 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) {
3042 mdb_warn("can't walk ACL");
3043 return (DCMD_ERR);
3044 }
3045
3046 return (DCMD_OK);
3047 }
3048
3049 /* ARGSUSED */
3050 static int
3051 zfs_acl_node_walk_init(mdb_walk_state_t *wsp)
3052 {
3053 if (wsp->walk_addr == NULL) {
3054 mdb_warn("must supply address of zfs_acl_node_t\n");
3055 return (WALK_ERR);
3056 }
3057
3058 wsp->walk_addr +=
3059 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl");
3060
3061 if (mdb_layered_walk("list", wsp) == -1) {
3062 mdb_warn("failed to walk 'list'\n");
3063 return (WALK_ERR);
3064 }
3065
3066 return (WALK_NEXT);
3067 }
3068
3069 static int
3070 zfs_acl_node_walk_step(mdb_walk_state_t *wsp)
3071 {
3072 zfs_acl_node_t aclnode;
3073
3074 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t),
3075 wsp->walk_addr) == -1) {
3076 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr);
3077 return (WALK_ERR);
3078 }
3079
3080 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata));
3081 }
3082
3083 typedef struct ace_walk_data {
3084 int ace_count;
3085 int ace_version;
3086 } ace_walk_data_t;
3087
3088 static int
3089 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version,
3090 int ace_count, uintptr_t ace_data)
3091 {
3092 ace_walk_data_t *ace_walk_data;
3093
3094 if (wsp->walk_addr == NULL) {
3095 mdb_warn("must supply address of zfs_acl_node_t\n");
3096 return (WALK_ERR);
3097 }
3098
3099 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC);
3100
3101 ace_walk_data->ace_count = ace_count;
3102 ace_walk_data->ace_version = version;
3103
3104 wsp->walk_addr = ace_data;
3105 wsp->walk_data = ace_walk_data;
3106
3107 return (WALK_NEXT);
3108 }
3109
3110 static int
3111 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version)
3112 {
3113 static int gotid;
3114 static mdb_ctf_id_t acl_id;
3115 int z_ace_count;
3116 uintptr_t z_acldata;
3117
3118 if (!gotid) {
3119 if (mdb_ctf_lookup_by_name("struct zfs_acl_node",
3120 &acl_id) == -1) {
3121 mdb_warn("couldn't find struct zfs_acl_node");
3122 return (DCMD_ERR);
3123 }
3124 gotid = TRUE;
3125 }
3126
3127 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) {
3128 return (DCMD_ERR);
3129 }
3130 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) {
3131 return (DCMD_ERR);
3132 }
3133
3134 return (zfs_aces_walk_init_common(wsp, version,
3135 z_ace_count, z_acldata));
3136 }
3137
3138 /* ARGSUSED */
3139 static int
3140 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp)
3141 {
3142 return (zfs_acl_node_aces_walk_init_common(wsp, 1));
3143 }
3144
3145 /* ARGSUSED */
3146 static int
3147 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp)
3148 {
3149 return (zfs_acl_node_aces_walk_init_common(wsp, 0));
3150 }
3151
3152 static int
3153 zfs_aces_walk_step(mdb_walk_state_t *wsp)
3154 {
3155 ace_walk_data_t *ace_data = wsp->walk_data;
3156 zfs_ace_t zace;
3157 ace_t *acep;
3158 int status;
3159 int entry_type;
3160 int allow_type;
3161 uintptr_t ptr;
3162
3163 if (ace_data->ace_count == 0)
3164 return (WALK_DONE);
3165
3166 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) {
3167 mdb_warn("failed to read zfs_ace_t at %#lx",
3168 wsp->walk_addr);
3169 return (WALK_ERR);
3170 }
3171
3172 switch (ace_data->ace_version) {
3173 case 0:
3174 acep = (ace_t *)&zace;
3175 entry_type = acep->a_flags & ACE_TYPE_FLAGS;
3176 allow_type = acep->a_type;
3177 break;
3178 case 1:
3179 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS;
3180 allow_type = zace.z_hdr.z_type;
3181 break;
3182 default:
3183 return (WALK_ERR);
3184 }
3185
3186 ptr = (uintptr_t)wsp->walk_addr;
3187 switch (entry_type) {
3188 case ACE_OWNER:
3189 case ACE_EVERYONE:
3190 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
3191 ptr += ace_data->ace_version == 0 ?
3192 sizeof (ace_t) : sizeof (zfs_ace_hdr_t);
3193 break;
3194 case ACE_IDENTIFIER_GROUP:
3195 default:
3196 switch (allow_type) {
3197 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
3198 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
3199 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
3200 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
3201 ptr += ace_data->ace_version == 0 ?
3202 sizeof (ace_t) : sizeof (zfs_object_ace_t);
3203 break;
3204 default:
3205 ptr += ace_data->ace_version == 0 ?
3206 sizeof (ace_t) : sizeof (zfs_ace_t);
3207 break;
3208 }
3209 }
3210
3211 ace_data->ace_count--;
3212 status = wsp->walk_callback(wsp->walk_addr,
3213 (void *)(uintptr_t)&zace, wsp->walk_cbdata);
3214
3215 wsp->walk_addr = ptr;
3216 return (status);
3217 }
3218
3219 typedef struct mdb_zfs_rrwlock {
3220 uintptr_t rr_writer;
3221 boolean_t rr_writer_wanted;
3222 } mdb_zfs_rrwlock_t;
3223
3224 static uint_t rrw_key;
3225
3226 /* ARGSUSED */
3227 static int
3228 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
3229 {
3230 mdb_zfs_rrwlock_t rrw;
3231
3232 if (rrw_key == 0) {
3233 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1)
3234 return (DCMD_ERR);
3235 }
3236
3237 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr,
3238 0) == -1)
3239 return (DCMD_ERR);
3240
3241 if (rrw.rr_writer != 0) {
3242 mdb_printf("write lock held by thread %lx\n", rrw.rr_writer);
3243 return (DCMD_OK);
3244 }
3245
3246 if (rrw.rr_writer_wanted) {
3247 mdb_printf("writer wanted\n");
3248 }
3249
3250 mdb_printf("anonymous references:\n");
3251 (void) mdb_call_dcmd("refcount", addr +
3252 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"),
3253 DCMD_ADDRSPEC, 0, NULL);
3254
3255 mdb_printf("linked references:\n");
3256 (void) mdb_call_dcmd("refcount", addr +
3257 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"),
3258 DCMD_ADDRSPEC, 0, NULL);
3259
3260 /*
3261 * XXX This should find references from
3262 * "::walk thread | ::tsd -v <rrw_key>", but there is no support
3263 * for programmatic consumption of dcmds, so this would be
3264 * difficult, potentially requiring reimplementing ::tsd (both
3265 * user and kernel versions) in this MDB module.
3266 */
3267
3268 return (DCMD_OK);
3269 }
3270
3271 /*
3272 * MDB module linkage information:
3273 *
3274 * We declare a list of structures describing our dcmds, and a function
3275 * named _mdb_init to return a pointer to our module information.
3276 */
3277
3278 static const mdb_dcmd_t dcmds[] = {
3279 { "arc", "[-bkmg]", "print ARC variables", arc_print },
3280 { "blkptr", ":", "print blkptr_t", blkptr },
3281 { "dbuf", ":", "print dmu_buf_impl_t", dbuf },
3282 { "dbuf_stats", ":", "dbuf stats", dbuf_stats },
3283 { "dbufs",
3284 "\t[-O objset_t*] [-n objset_name | \"mos\"] "
3285 "[-o object | \"mdn\"] \n"
3286 "\t[-l level] [-b blkid | \"bonus\"]",
3287 "find dmu_buf_impl_t's that match specified criteria", dbufs },
3288 { "abuf_find", "dva_word[0] dva_word[1]",
3289 "find arc_buf_hdr_t of a specified DVA",
3290 abuf_find },
3291 { "spa", "?[-cevmMh]\n"
3292 "\t-c display spa config\n"
3293 "\t-e display vdev statistics\n"
3294 "\t-v display vdev information\n"
3295 "\t-m display metaslab statistics\n"
3296 "\t-M display metaslab group statistics\n"
3297 "\t-h display histogram (requires -m or -M)\n",
3298 "spa_t summary", spa_print },
3299 { "spa_config", ":", "print spa_t configuration", spa_print_config },
3300 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
3301 { "spa_vdevs", ":[-emMh]\n"
3302 "\t-e display vdev statistics\n"
3303 "\t-m dispaly metaslab statistics\n"
3304 "\t-M display metaslab group statistic\n"
3305 "\t-h display histogram (requires -m or -M)\n",
3306 "given a spa_t, print vdev summary", spa_vdevs },
3307 { "vdev", ":[-re]\n"
3308 "\t-r display recursively\n"
3309 "\t-e display statistics\n"
3310 "\t-m display metaslab statistics\n"
3311 "\t-M display metaslab group statistics\n"
3312 "\t-h display histogram (requires -m or -M)\n",
3313 "vdev_t summary", vdev_print },
3314 { "zio", ":[-cpr]\n"
3315 "\t-c display children\n"
3316 "\t-p display parents\n"
3317 "\t-r display recursively",
3318 "zio_t summary", zio_print },
3319 { "zio_state", "?", "print out all zio_t structures on system or "
3320 "for a particular pool", zio_state },
3321 { "zfs_blkstats", ":[-v]",
3322 "given a spa_t, print block type stats from last scrub",
3323 zfs_blkstats },
3324 { "zfs_params", "", "print zfs tunable parameters", zfs_params },
3325 { "refcount", ":[-r]\n"
3326 "\t-r display recently removed references",
3327 "print refcount_t holders", refcount },
3328 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf },
3329 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t",
3330 zfs_acl_dump },
3331 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print },
3332 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print },
3333 { "sa_attr_table", ":", "print SA attribute table from sa_os_t",
3334 sa_attr_table},
3335 { "sa_attr", ": attr_id",
3336 "print SA attribute address when given sa_handle_t", sa_attr_print},
3337 { "zfs_dbgmsg", ":[-va]",
3338 "print zfs debug log", dbgmsg},
3339 { "rrwlock", ":",
3340 "print rrwlock_t, including readers", rrwlock},
3341 { NULL }
3342 };
3343
3344 static const mdb_walker_t walkers[] = {
3345 { "zms_freelist", "walk ZFS metaslab freelist",
3346 freelist_walk_init, freelist_walk_step, NULL },
3347 { "txg_list", "given any txg_list_t *, walk all entries in all txgs",
3348 txg_list_walk_init, txg_list_walk_step, NULL },
3349 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
3350 txg_list0_walk_init, txg_list_walk_step, NULL },
3351 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
3352 txg_list1_walk_init, txg_list_walk_step, NULL },
3353 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
3354 txg_list2_walk_init, txg_list_walk_step, NULL },
3355 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
3356 txg_list3_walk_init, txg_list_walk_step, NULL },
3357 { "zio", "walk all zio structures, optionally for a particular spa_t",
3358 zio_walk_init, zio_walk_step, NULL },
3359 { "zio_root",
3360 "walk all root zio_t structures, optionally for a particular spa_t",
3361 zio_walk_init, zio_walk_root_step, NULL },
3362 { "spa", "walk all spa_t entries in the namespace",
3363 spa_walk_init, spa_walk_step, NULL },
3364 { "metaslab", "given a spa_t *, walk all metaslab_t structures",
3365 metaslab_walk_init, metaslab_walk_step, NULL },
3366 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes",
3367 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL },
3368 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs",
3369 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL },
3370 { "zfs_acl_node_aces0",
3371 "given a zfs_acl_node_t, walk all ACEs as ace_t",
3372 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL },
3373 { NULL }
3374 };
3375
3376 static const mdb_modinfo_t modinfo = {
3377 MDB_API_VERSION, dcmds, walkers
3378 };
3379
3380 const mdb_modinfo_t *
3381 _mdb_init(void)
3382 {
3383 return (&modinfo);
3384 }
Unleashed OS