?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
5368 ARC should cache more metadata
[unleashed.git] / usr / src / cmd / mdb / common / modules / zfs / zfs.c
blobac00609767e927737d86f7957880a89e5c94f326
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 zap_leaf_t l;
595 zap_leaf_phys_t *zlp = (void *)buf;
596 int i;
597
598 if (mdb_getopts(argc, argv,
599 'v', MDB_OPT_SETBITS, TRUE, &verbose,
600 '4', MDB_OPT_SETBITS, TRUE, &four,
601 NULL) != argc)
602 return (DCMD_USAGE);
603
604 l.l_phys = zlp;
605 l.l_bs = 14; /* assume 16k blocks */
606 if (four)
607 l.l_bs = 12;
608
609 if (!(flags & DCMD_ADDRSPEC)) {
610 return (DCMD_USAGE);
611 }
612
613 if (mdb_vread(buf, sizeof (buf), addr) == -1) {
614 mdb_warn("failed to read zap_leaf_phys_t at %p", addr);
615 return (DCMD_ERR);
616 }
617
618 if (zlp->l_hdr.lh_block_type != ZBT_LEAF ||
619 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) {
620 mdb_warn("This does not appear to be a zap_leaf_phys_t");
621 return (DCMD_ERR);
622 }
623
624 mdb_printf("zap_leaf_phys_t at %p:\n", addr);
625 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len);
626 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix);
627 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries);
628 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree,
629 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l)));
630 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist);
631 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags,
632 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ?
633 "ENTRIES_CDSORTED" : "");
634
635 if (verbose) {
636 mdb_printf(" hash table:\n");
637 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) {
638 if (zlp->l_hash[i] != CHAIN_END)
639 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]);
640 }
641 }
642
643 mdb_printf(" chunks:\n");
644 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) {
645 /* LINTED: alignment */
646 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i);
647 switch (zlc->l_entry.le_type) {
648 case ZAP_CHUNK_FREE:
649 if (verbose) {
650 mdb_printf(" %u: free; lf_next = %u\n",
651 i, zlc->l_free.lf_next);
652 }
653 break;
654 case ZAP_CHUNK_ENTRY:
655 mdb_printf(" %u: entry\n", i);
656 if (verbose) {
657 mdb_printf(" le_next = %u\n",
658 zlc->l_entry.le_next);
659 }
660 mdb_printf(" le_name_chunk = %u\n",
661 zlc->l_entry.le_name_chunk);
662 mdb_printf(" le_name_numints = %u\n",
663 zlc->l_entry.le_name_numints);
664 mdb_printf(" le_value_chunk = %u\n",
665 zlc->l_entry.le_value_chunk);
666 mdb_printf(" le_value_intlen = %u\n",
667 zlc->l_entry.le_value_intlen);
668 mdb_printf(" le_value_numints = %u\n",
669 zlc->l_entry.le_value_numints);
670 mdb_printf(" le_cd = %u\n",
671 zlc->l_entry.le_cd);
672 mdb_printf(" le_hash = %llx\n",
673 zlc->l_entry.le_hash);
674 break;
675 case ZAP_CHUNK_ARRAY:
676 mdb_printf(" %u: array", i);
677 if (strisprint((char *)zlc->l_array.la_array))
678 mdb_printf(" \"%s\"", zlc->l_array.la_array);
679 mdb_printf("\n");
680 if (verbose) {
681 int j;
682 mdb_printf(" ");
683 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) {
684 mdb_printf("%02x ",
685 zlc->l_array.la_array[j]);
686 }
687 mdb_printf("\n");
688 }
689 if (zlc->l_array.la_next != CHAIN_END) {
690 mdb_printf(" lf_next = %u\n",
691 zlc->l_array.la_next);
692 }
693 break;
694 default:
695 mdb_printf(" %u: undefined type %u\n",
696 zlc->l_entry.le_type);
697 }
698 }
699
700 return (DCMD_OK);
701 }
702
703 typedef struct dbufs_data {
704 mdb_ctf_id_t id;
705 uint64_t objset;
706 uint64_t object;
707 uint64_t level;
708 uint64_t blkid;
709 char *osname;
710 } dbufs_data_t;
711
712 #define DBUFS_UNSET (0xbaddcafedeadbeefULL)
713
714 /* ARGSUSED */
715 static int
716 dbufs_cb(uintptr_t addr, const void *unknown, void *arg)
717 {
718 dbufs_data_t *data = arg;
719 uintptr_t objset;
720 dmu_buf_t db;
721 uint8_t level;
722 uint64_t blkid;
723 char osname[MAXNAMELEN];
724
725 if (GETMEMBID(addr, &data->id, db_objset, objset) ||
726 GETMEMBID(addr, &data->id, db, db) ||
727 GETMEMBID(addr, &data->id, db_level, level) ||
728 GETMEMBID(addr, &data->id, db_blkid, blkid)) {
729 return (WALK_ERR);
730 }
731
732 if ((data->objset == DBUFS_UNSET || data->objset == objset) &&
733 (data->osname == NULL || (objset_name(objset, osname) == 0 &&
734 strcmp(data->osname, osname) == 0)) &&
735 (data->object == DBUFS_UNSET || data->object == db.db_object) &&
736 (data->level == DBUFS_UNSET || data->level == level) &&
737 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) {
738 mdb_printf("%#lr\n", addr);
739 }
740 return (WALK_NEXT);
741 }
742
743 /* ARGSUSED */
744 static int
745 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
746 {
747 dbufs_data_t data;
748 char *object = NULL;
749 char *blkid = NULL;
750
751 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET;
752 data.osname = NULL;
753
754 if (mdb_getopts(argc, argv,
755 'O', MDB_OPT_UINT64, &data.objset,
756 'n', MDB_OPT_STR, &data.osname,
757 'o', MDB_OPT_STR, &object,
758 'l', MDB_OPT_UINT64, &data.level,
759 'b', MDB_OPT_STR, &blkid) != argc) {
760 return (DCMD_USAGE);
761 }
762
763 if (object) {
764 if (strcmp(object, "mdn") == 0) {
765 data.object = DMU_META_DNODE_OBJECT;
766 } else {
767 data.object = mdb_strtoull(object);
768 }
769 }
770
771 if (blkid) {
772 if (strcmp(blkid, "bonus") == 0) {
773 data.blkid = DMU_BONUS_BLKID;
774 } else {
775 data.blkid = mdb_strtoull(blkid);
776 }
777 }
778
779 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dmu_buf_impl", &data.id) == -1) {
780 mdb_warn("couldn't find struct dmu_buf_impl_t");
781 return (DCMD_ERR);
782 }
783
784 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) {
785 mdb_warn("can't walk dbufs");
786 return (DCMD_ERR);
787 }
788
789 return (DCMD_OK);
790 }
791
792 typedef struct abuf_find_data {
793 dva_t dva;
794 mdb_ctf_id_t id;
795 } abuf_find_data_t;
796
797 /* ARGSUSED */
798 static int
799 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg)
800 {
801 abuf_find_data_t *data = arg;
802 dva_t dva;
803
804 if (GETMEMBID(addr, &data->id, b_dva, dva)) {
805 return (WALK_ERR);
806 }
807
808 if (dva.dva_word[0] == data->dva.dva_word[0] &&
809 dva.dva_word[1] == data->dva.dva_word[1]) {
810 mdb_printf("%#lr\n", addr);
811 }
812 return (WALK_NEXT);
813 }
814
815 /* ARGSUSED */
816 static int
817 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
818 {
819 abuf_find_data_t data;
820 GElf_Sym sym;
821 int i;
822 const char *syms[] = {
823 "ARC_mru",
824 "ARC_mru_ghost",
825 "ARC_mfu",
826 "ARC_mfu_ghost",
827 };
828
829 if (argc != 2)
830 return (DCMD_USAGE);
831
832 for (i = 0; i < 2; i ++) {
833 switch (argv[i].a_type) {
834 case MDB_TYPE_STRING:
835 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str);
836 break;
837 case MDB_TYPE_IMMEDIATE:
838 data.dva.dva_word[i] = argv[i].a_un.a_val;
839 break;
840 default:
841 return (DCMD_USAGE);
842 }
843 }
844
845 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "arc_buf_hdr", &data.id) == -1) {
846 mdb_warn("couldn't find struct arc_buf_hdr");
847 return (DCMD_ERR);
848 }
849
850 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) {
851 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, syms[i], &sym)) {
852 mdb_warn("can't find symbol %s", syms[i]);
853 return (DCMD_ERR);
854 }
855
856 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) {
857 mdb_warn("can't walk %s", syms[i]);
858 return (DCMD_ERR);
859 }
860 }
861
862 return (DCMD_OK);
863 }
864
865
866 typedef struct dbgmsg_arg {
867 boolean_t da_verbose;
868 boolean_t da_address;
869 } dbgmsg_arg_t;
870
871 /* ARGSUSED */
872 static int
873 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg)
874 {
875 static mdb_ctf_id_t id;
876 static boolean_t gotid;
877 static ulong_t off;
878
879 dbgmsg_arg_t *da = arg;
880 time_t timestamp;
881 char buf[1024];
882
883 if (!gotid) {
884 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "zfs_dbgmsg", &id) ==
885 -1) {
886 mdb_warn("couldn't find struct zfs_dbgmsg");
887 return (WALK_ERR);
888 }
889 gotid = TRUE;
890 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) {
891 mdb_warn("couldn't find zdm_msg");
892 return (WALK_ERR);
893 }
894 off /= 8;
895 }
896
897
898 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) {
899 return (WALK_ERR);
900 }
901
902 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) {
903 mdb_warn("failed to read zdm_msg at %p\n", addr + off);
904 return (DCMD_ERR);
905 }
906
907 if (da->da_address)
908 mdb_printf("%p ", addr);
909 if (da->da_verbose)
910 mdb_printf("%Y ", timestamp);
911
912 mdb_printf("%s\n", buf);
913
914 if (da->da_verbose)
915 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
916
917 return (WALK_NEXT);
918 }
919
920 /* ARGSUSED */
921 static int
922 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
923 {
924 GElf_Sym sym;
925 dbgmsg_arg_t da = { 0 };
926
927 if (mdb_getopts(argc, argv,
928 'v', MDB_OPT_SETBITS, B_TRUE, &da.da_verbose,
929 'a', MDB_OPT_SETBITS, B_TRUE, &da.da_address,
930 NULL) != argc)
931 return (DCMD_USAGE);
932
933 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "zfs_dbgmsgs", &sym)) {
934 mdb_warn("can't find zfs_dbgmsgs");
935 return (DCMD_ERR);
936 }
937
938 if (mdb_pwalk("list", dbgmsg_cb, &da, sym.st_value) != 0) {
939 mdb_warn("can't walk zfs_dbgmsgs");
940 return (DCMD_ERR);
941 }
942
943 return (DCMD_OK);
944 }
945
946 /*ARGSUSED*/
947 static int
948 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
949 {
950 kstat_named_t *stats;
951 GElf_Sym sym;
952 int nstats, i;
953 uint_t opt_a = FALSE;
954 uint_t opt_b = FALSE;
955 uint_t shift = 0;
956 const char *suffix;
957
958 static const char *bytestats[] = {
959 "p", "c", "c_min", "c_max", "size", "duplicate_buffers_size",
960 "arc_meta_used", "arc_meta_limit", "arc_meta_max",
961 "arc_meta_min", "hdr_size", "data_size", "other_size",
962 NULL
963 };
964
965 static const char *extras[] = {
966 "arc_no_grow", "arc_tempreserve",
967 NULL
968 };
969
970 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "arc_stats", &sym) == -1) {
971 mdb_warn("failed to find 'arc_stats'");
972 return (DCMD_ERR);
973 }
974
975 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC);
976
977 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) {
978 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value);
979 return (DCMD_ERR);
980 }
981
982 nstats = sym.st_size / sizeof (kstat_named_t);
983
984 /* NB: -a / opt_a are ignored for backwards compatability */
985 if (mdb_getopts(argc, argv,
986 'a', MDB_OPT_SETBITS, TRUE, &opt_a,
987 'b', MDB_OPT_SETBITS, TRUE, &opt_b,
988 'k', MDB_OPT_SETBITS, 10, &shift,
989 'm', MDB_OPT_SETBITS, 20, &shift,
990 'g', MDB_OPT_SETBITS, 30, &shift,
991 NULL) != argc)
992 return (DCMD_USAGE);
993
994 if (!opt_b && !shift)
995 shift = 20;
996
997 switch (shift) {
998 case 0:
999 suffix = "B";
1000 break;
1001 case 10:
1002 suffix = "KB";
1003 break;
1004 case 20:
1005 suffix = "MB";
1006 break;
1007 case 30:
1008 suffix = "GB";
1009 break;
1010 default:
1011 suffix = "XX";
1012 }
1013
1014 for (i = 0; i < nstats; i++) {
1015 int j;
1016 boolean_t bytes = B_FALSE;
1017
1018 for (j = 0; bytestats[j]; j++) {
1019 if (strcmp(stats[i].name, bytestats[j]) == 0) {
1020 bytes = B_TRUE;
1021 break;
1022 }
1023 }
1024
1025 if (bytes) {
1026 mdb_printf("%-25s = %9llu %s\n", stats[i].name,
1027 stats[i].value.ui64 >> shift, suffix);
1028 } else {
1029 mdb_printf("%-25s = %9llu\n", stats[i].name,
1030 stats[i].value.ui64);
1031 }
1032 }
1033
1034 for (i = 0; extras[i]; i++) {
1035 uint64_t buf;
1036
1037 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, extras[i], &sym) == -1) {
1038 mdb_warn("failed to find '%s'", extras[i]);
1039 return (DCMD_ERR);
1040 }
1041
1042 if (sym.st_size != sizeof (uint64_t) &&
1043 sym.st_size != sizeof (uint32_t)) {
1044 mdb_warn("expected scalar for variable '%s'\n",
1045 extras[i]);
1046 return (DCMD_ERR);
1047 }
1048
1049 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) {
1050 mdb_warn("couldn't read '%s'", extras[i]);
1051 return (DCMD_ERR);
1052 }
1053
1054 mdb_printf("%-25s = ", extras[i]);
1055
1056 /* NB: all the 64-bit extras happen to be byte counts */
1057 if (sym.st_size == sizeof (uint64_t))
1058 mdb_printf("%9llu %s\n", buf >> shift, suffix);
1059
1060 if (sym.st_size == sizeof (uint32_t))
1061 mdb_printf("%9d\n", *((uint32_t *)&buf));
1062 }
1063 return (DCMD_OK);
1064 }
1065
1066 typedef struct mdb_spa_print {
1067 pool_state_t spa_state;
1068 char spa_name[MAXNAMELEN];
1069 } mdb_spa_print_t;
1070
1071 /*
1072 * ::spa
1073 *
1074 * -c Print configuration information as well
1075 * -v Print vdev state
1076 * -e Print vdev error stats
1077 * -m Print vdev metaslab info
1078 * -M print vdev metaslab group info
1079 * -h Print histogram info (must be combined with -m or -M)
1080 *
1081 * Print a summarized spa_t. When given no arguments, prints out a table of all
1082 * active pools on the system.
1083 */
1084 /* ARGSUSED */
1085 static int
1086 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1087 {
1088 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED",
1089 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" };
1090 const char *state;
1091 int spa_flags = 0;
1092
1093 if (mdb_getopts(argc, argv,
1094 'c', MDB_OPT_SETBITS, SPA_FLAG_CONFIG, &spa_flags,
1095 'v', MDB_OPT_SETBITS, SPA_FLAG_VDEVS, &spa_flags,
1096 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags,
1097 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags,
1098 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags,
1099 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags,
1100 NULL) != argc)
1101 return (DCMD_USAGE);
1102
1103 if (!(flags & DCMD_ADDRSPEC)) {
1104 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) {
1105 mdb_warn("can't walk spa");
1106 return (DCMD_ERR);
1107 }
1108
1109 return (DCMD_OK);
1110 }
1111
1112 if (flags & DCMD_PIPE_OUT) {
1113 mdb_printf("%#lr\n", addr);
1114 return (DCMD_OK);
1115 }
1116
1117 if (DCMD_HDRSPEC(flags))
1118 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE",
1119 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME");
1120
1121 mdb_spa_print_t spa;
1122 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_print_t", addr, 0) == -1)
1123 return (DCMD_ERR);
1124
1125 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL)
1126 state = "UNKNOWN";
1127 else
1128 state = statetab[spa.spa_state];
1129
1130 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name);
1131
1132 if (spa_flags & SPA_FLAG_CONFIG) {
1133 mdb_printf("\n");
1134 mdb_inc_indent(4);
1135 if (mdb_call_dcmd("spa_config", addr, flags, 0,
1136 NULL) != DCMD_OK)
1137 return (DCMD_ERR);
1138 mdb_dec_indent(4);
1139 }
1140
1141 if (spa_flags & SPA_FLAG_ALL_VDEV) {
1142 mdb_arg_t v;
1143 char opts[100] = "-";
1144 int args =
1145 (spa_flags | SPA_FLAG_VDEVS) == SPA_FLAG_VDEVS ? 0 : 1;
1146
1147 if (spa_flags & SPA_FLAG_ERRORS)
1148 strcat(opts, "e");
1149 if (spa_flags & SPA_FLAG_METASLABS)
1150 strcat(opts, "m");
1151 if (spa_flags & SPA_FLAG_METASLAB_GROUPS)
1152 strcat(opts, "M");
1153 if (spa_flags & SPA_FLAG_HISTOGRAMS)
1154 strcat(opts, "h");
1155
1156 v.a_type = MDB_TYPE_STRING;
1157 v.a_un.a_str = opts;
1158
1159 mdb_printf("\n");
1160 mdb_inc_indent(4);
1161 if (mdb_call_dcmd("spa_vdevs", addr, flags, args,
1162 &v) != DCMD_OK)
1163 return (DCMD_ERR);
1164 mdb_dec_indent(4);
1165 }
1166
1167 return (DCMD_OK);
1168 }
1169
1170 typedef struct mdb_spa_config_spa {
1171 uintptr_t spa_config;
1172 } mdb_spa_config_spa_t;
1173
1174 /*
1175 * ::spa_config
1176 *
1177 * Given a spa_t, print the configuration information stored in spa_config.
1178 * Since it's just an nvlist, format it as an indented list of name=value pairs.
1179 * We simply read the value of spa_config and pass off to ::nvlist.
1180 */
1181 /* ARGSUSED */
1182 static int
1183 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1184 {
1185 mdb_spa_config_spa_t spa;
1186
1187 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1188 return (DCMD_USAGE);
1189
1190 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_config_spa_t",
1191 addr, 0) == -1)
1192 return (DCMD_ERR);
1193
1194 if (spa.spa_config == 0) {
1195 mdb_printf("(none)\n");
1196 return (DCMD_OK);
1197 }
1198
1199 return (mdb_call_dcmd("nvlist", spa.spa_config, flags,
1200 0, NULL));
1201 }
1202
1203 const char histo_stars[] = "****************************************";
1204 const int histo_width = sizeof (histo_stars) - 1;
1205
1206 static void
1207 dump_histogram(const uint64_t *histo, int size, int offset)
1208 {
1209 int i;
1210 int minidx = size - 1;
1211 int maxidx = 0;
1212 uint64_t max = 0;
1213
1214 for (i = 0; i < size; i++) {
1215 if (histo[i] > max)
1216 max = histo[i];
1217 if (histo[i] > 0 && i > maxidx)
1218 maxidx = i;
1219 if (histo[i] > 0 && i < minidx)
1220 minidx = i;
1221 }
1222
1223 if (max < histo_width)
1224 max = histo_width;
1225
1226 for (i = minidx; i <= maxidx; i++) {
1227 mdb_printf("%3u: %6llu %s\n",
1228 i + offset, (u_longlong_t)histo[i],
1229 &histo_stars[(max - histo[i]) * histo_width / max]);
1230 }
1231 }
1232
1233 typedef struct mdb_range_tree {
1234 uint64_t rt_space;
1235 } mdb_range_tree_t;
1236
1237 typedef struct mdb_metaslab_group {
1238 uint64_t mg_fragmentation;
1239 uint64_t mg_histogram[RANGE_TREE_HISTOGRAM_SIZE];
1240 } mdb_metaslab_group_t;
1241
1242 typedef struct mdb_metaslab {
1243 uint64_t ms_id;
1244 uint64_t ms_start;
1245 uint64_t ms_size;
1246 uint64_t ms_fragmentation;
1247 uintptr_t ms_alloctree[TXG_SIZE];
1248 uintptr_t ms_freetree[TXG_SIZE];
1249 uintptr_t ms_tree;
1250 uintptr_t ms_sm;
1251 } mdb_metaslab_t;
1252
1253 typedef struct mdb_space_map_phys_t {
1254 uint64_t smp_alloc;
1255 uint64_t smp_histogram[SPACE_MAP_HISTOGRAM_SIZE];
1256 } mdb_space_map_phys_t;
1257
1258 typedef struct mdb_space_map {
1259 uint64_t sm_size;
1260 uint8_t sm_shift;
1261 uint64_t sm_alloc;
1262 uintptr_t sm_phys;
1263 } mdb_space_map_t;
1264
1265 typedef struct mdb_vdev {
1266 uintptr_t vdev_ms;
1267 uint64_t vdev_ms_count;
1268 vdev_stat_t vdev_stat;
1269 } mdb_vdev_t;
1270
1271 static int
1272 metaslab_stats(uintptr_t addr, int spa_flags)
1273 {
1274 mdb_vdev_t vdev;
1275 uintptr_t *vdev_ms;
1276
1277 if (mdb_ctf_vread(&vdev, "vdev_t", "mdb_vdev_t",
1278 (uintptr_t)addr, 0) == -1) {
1279 mdb_warn("failed to read vdev at %p\n", addr);
1280 return (DCMD_ERR);
1281 }
1282
1283 mdb_inc_indent(4);
1284 mdb_printf("%<u>%-?s %6s %20s %10s %9s%</u>\n", "ADDR", "ID",
1285 "OFFSET", "FREE", "FRAGMENTATION");
1286
1287 vdev_ms = mdb_alloc(vdev.vdev_ms_count * sizeof (void *),
1288 UM_SLEEP | UM_GC);
1289 if (mdb_vread(vdev_ms, vdev.vdev_ms_count * sizeof (void *),
1290 (uintptr_t)vdev.vdev_ms) == -1) {
1291 mdb_warn("failed to read vdev_ms at %p\n", vdev.vdev_ms);
1292 return (DCMD_ERR);
1293 }
1294
1295 for (int m = 0; m < vdev.vdev_ms_count; m++) {
1296 mdb_metaslab_t ms;
1297 mdb_space_map_t sm = { 0 };
1298 char free[NICENUM_BUFLEN];
1299
1300 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t",
1301 (uintptr_t)vdev_ms[m], 0) == -1)
1302 return (DCMD_ERR);
1303
1304 if (ms.ms_sm != NULL &&
1305 mdb_ctf_vread(&sm, "space_map_t", "mdb_space_map_t",
1306 ms.ms_sm, 0) == -1)
1307 return (DCMD_ERR);
1308
1309 mdb_nicenum(ms.ms_size - sm.sm_alloc, free);
1310
1311 mdb_printf("%0?p %6llu %20llx %10s ", vdev_ms[m], ms.ms_id,
1312 ms.ms_start, free);
1313 if (ms.ms_fragmentation == ZFS_FRAG_INVALID)
1314 mdb_printf("%9s\n", "-");
1315 else
1316 mdb_printf("%9llu%%\n", ms.ms_fragmentation);
1317
1318 if ((spa_flags & SPA_FLAG_HISTOGRAMS) && ms.ms_sm != NULL) {
1319 mdb_space_map_phys_t smp;
1320
1321 if (sm.sm_phys == NULL)
1322 continue;
1323
1324 (void) mdb_ctf_vread(&smp, "space_map_phys_t",
1325 "mdb_space_map_phys_t", sm.sm_phys, 0);
1326
1327 dump_histogram(smp.smp_histogram,
1328 SPACE_MAP_HISTOGRAM_SIZE, sm.sm_shift);
1329 }
1330 }
1331 mdb_dec_indent(4);
1332 return (DCMD_OK);
1333 }
1334
1335 static int
1336 metaslab_group_stats(uintptr_t addr, int spa_flags)
1337 {
1338 mdb_metaslab_group_t mg;
1339 if (mdb_ctf_vread(&mg, "metaslab_group_t", "mdb_metaslab_group_t",
1340 (uintptr_t)addr, 0) == -1) {
1341 mdb_warn("failed to read vdev_mg at %p\n", addr);
1342 return (DCMD_ERR);
1343 }
1344
1345 mdb_inc_indent(4);
1346 mdb_printf("%<u>%-?s %15s%</u>\n", "ADDR", "FRAGMENTATION");
1347 if (mg.mg_fragmentation == ZFS_FRAG_INVALID)
1348 mdb_printf("%0?p %15s\n", addr, "-");
1349 else
1350 mdb_printf("%0?p %15llu%%\n", addr, mg.mg_fragmentation);
1351
1352 if (spa_flags & SPA_FLAG_HISTOGRAMS)
1353 dump_histogram(mg.mg_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0);
1354 mdb_dec_indent(4);
1355 return (DCMD_OK);
1356 }
1357
1358 /*
1359 * ::vdev
1360 *
1361 * Print out a summarized vdev_t, in the following form:
1362 *
1363 * ADDR STATE AUX DESC
1364 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0
1365 *
1366 * If '-r' is specified, recursively visit all children.
1367 *
1368 * With '-e', the statistics associated with the vdev are printed as well.
1369 */
1370 static int
1371 do_print_vdev(uintptr_t addr, int flags, int depth, boolean_t recursive,
1372 int spa_flags)
1373 {
1374 vdev_t vdev;
1375 char desc[MAXNAMELEN];
1376 int c, children;
1377 uintptr_t *child;
1378 const char *state, *aux;
1379
1380 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) {
1381 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr);
1382 return (DCMD_ERR);
1383 }
1384
1385 if (flags & DCMD_PIPE_OUT) {
1386 mdb_printf("%#lr\n", addr);
1387 } else {
1388 if (vdev.vdev_path != NULL) {
1389 if (mdb_readstr(desc, sizeof (desc),
1390 (uintptr_t)vdev.vdev_path) == -1) {
1391 mdb_warn("failed to read vdev_path at %p\n",
1392 vdev.vdev_path);
1393 return (DCMD_ERR);
1394 }
1395 } else if (vdev.vdev_ops != NULL) {
1396 vdev_ops_t ops;
1397 if (mdb_vread(&ops, sizeof (ops),
1398 (uintptr_t)vdev.vdev_ops) == -1) {
1399 mdb_warn("failed to read vdev_ops at %p\n",
1400 vdev.vdev_ops);
1401 return (DCMD_ERR);
1402 }
1403 (void) strcpy(desc, ops.vdev_op_type);
1404 } else {
1405 (void) strcpy(desc, "<unknown>");
1406 }
1407
1408 if (depth == 0 && DCMD_HDRSPEC(flags))
1409 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n",
1410 "ADDR", "STATE", "AUX",
1411 sizeof (uintptr_t) == 4 ? 43 : 35,
1412 "DESCRIPTION");
1413
1414 mdb_printf("%0?p ", addr);
1415
1416 switch (vdev.vdev_state) {
1417 case VDEV_STATE_CLOSED:
1418 state = "CLOSED";
1419 break;
1420 case VDEV_STATE_OFFLINE:
1421 state = "OFFLINE";
1422 break;
1423 case VDEV_STATE_CANT_OPEN:
1424 state = "CANT_OPEN";
1425 break;
1426 case VDEV_STATE_DEGRADED:
1427 state = "DEGRADED";
1428 break;
1429 case VDEV_STATE_HEALTHY:
1430 state = "HEALTHY";
1431 break;
1432 case VDEV_STATE_REMOVED:
1433 state = "REMOVED";
1434 break;
1435 case VDEV_STATE_FAULTED:
1436 state = "FAULTED";
1437 break;
1438 default:
1439 state = "UNKNOWN";
1440 break;
1441 }
1442
1443 switch (vdev.vdev_stat.vs_aux) {
1444 case VDEV_AUX_NONE:
1445 aux = "-";
1446 break;
1447 case VDEV_AUX_OPEN_FAILED:
1448 aux = "OPEN_FAILED";
1449 break;
1450 case VDEV_AUX_CORRUPT_DATA:
1451 aux = "CORRUPT_DATA";
1452 break;
1453 case VDEV_AUX_NO_REPLICAS:
1454 aux = "NO_REPLICAS";
1455 break;
1456 case VDEV_AUX_BAD_GUID_SUM:
1457 aux = "BAD_GUID_SUM";
1458 break;
1459 case VDEV_AUX_TOO_SMALL:
1460 aux = "TOO_SMALL";
1461 break;
1462 case VDEV_AUX_BAD_LABEL:
1463 aux = "BAD_LABEL";
1464 break;
1465 case VDEV_AUX_VERSION_NEWER:
1466 aux = "VERS_NEWER";
1467 break;
1468 case VDEV_AUX_VERSION_OLDER:
1469 aux = "VERS_OLDER";
1470 break;
1471 case VDEV_AUX_UNSUP_FEAT:
1472 aux = "UNSUP_FEAT";
1473 break;
1474 case VDEV_AUX_SPARED:
1475 aux = "SPARED";
1476 break;
1477 case VDEV_AUX_ERR_EXCEEDED:
1478 aux = "ERR_EXCEEDED";
1479 break;
1480 case VDEV_AUX_IO_FAILURE:
1481 aux = "IO_FAILURE";
1482 break;
1483 case VDEV_AUX_BAD_LOG:
1484 aux = "BAD_LOG";
1485 break;
1486 case VDEV_AUX_EXTERNAL:
1487 aux = "EXTERNAL";
1488 break;
1489 case VDEV_AUX_SPLIT_POOL:
1490 aux = "SPLIT_POOL";
1491 break;
1492 default:
1493 aux = "UNKNOWN";
1494 break;
1495 }
1496
1497 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc);
1498
1499 if (spa_flags & SPA_FLAG_ERRORS) {
1500 vdev_stat_t *vs = &vdev.vdev_stat;
1501 int i;
1502
1503 mdb_inc_indent(4);
1504 mdb_printf("\n");
1505 mdb_printf("%<u> %12s %12s %12s %12s "
1506 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM",
1507 "IOCTL");
1508 mdb_printf("OPS ");
1509 for (i = 1; i < ZIO_TYPES; i++)
1510 mdb_printf("%11#llx%s", vs->vs_ops[i],
1511 i == ZIO_TYPES - 1 ? "" : " ");
1512 mdb_printf("\n");
1513 mdb_printf("BYTES ");
1514 for (i = 1; i < ZIO_TYPES; i++)
1515 mdb_printf("%11#llx%s", vs->vs_bytes[i],
1516 i == ZIO_TYPES - 1 ? "" : " ");
1517
1518
1519 mdb_printf("\n");
1520 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors);
1521 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors);
1522 mdb_printf("ECKSUM %10#llx\n",
1523 vs->vs_checksum_errors);
1524 mdb_dec_indent(4);
1525 mdb_printf("\n");
1526 }
1527
1528 if (spa_flags & SPA_FLAG_METASLAB_GROUPS &&
1529 vdev.vdev_mg != NULL) {
1530 metaslab_group_stats((uintptr_t)vdev.vdev_mg,
1531 spa_flags);
1532 }
1533 if (spa_flags & SPA_FLAG_METASLABS && vdev.vdev_ms != NULL) {
1534 metaslab_stats((uintptr_t)addr, spa_flags);
1535 }
1536 }
1537
1538 children = vdev.vdev_children;
1539
1540 if (children == 0 || !recursive)
1541 return (DCMD_OK);
1542
1543 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC);
1544 if (mdb_vread(child, children * sizeof (void *),
1545 (uintptr_t)vdev.vdev_child) == -1) {
1546 mdb_warn("failed to read vdev children at %p", vdev.vdev_child);
1547 return (DCMD_ERR);
1548 }
1549
1550 for (c = 0; c < children; c++) {
1551 if (do_print_vdev(child[c], flags, depth + 2, recursive,
1552 spa_flags)) {
1553 return (DCMD_ERR);
1554 }
1555 }
1556
1557 return (DCMD_OK);
1558 }
1559
1560 static int
1561 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1562 {
1563 uint64_t depth = 0;
1564 boolean_t recursive = B_FALSE;
1565 int spa_flags = 0;
1566
1567 if (mdb_getopts(argc, argv,
1568 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags,
1569 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags,
1570 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags,
1571 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags,
1572 'r', MDB_OPT_SETBITS, TRUE, &recursive,
1573 'd', MDB_OPT_UINT64, &depth, NULL) != argc)
1574 return (DCMD_USAGE);
1575
1576 if (!(flags & DCMD_ADDRSPEC)) {
1577 mdb_warn("no vdev_t address given\n");
1578 return (DCMD_ERR);
1579 }
1580
1581 return (do_print_vdev(addr, flags, (int)depth, recursive, spa_flags));
1582 }
1583
1584 typedef struct metaslab_walk_data {
1585 uint64_t mw_numvdevs;
1586 uintptr_t *mw_vdevs;
1587 int mw_curvdev;
1588 uint64_t mw_nummss;
1589 uintptr_t *mw_mss;
1590 int mw_curms;
1591 } metaslab_walk_data_t;
1592
1593 static int
1594 metaslab_walk_step(mdb_walk_state_t *wsp)
1595 {
1596 metaslab_walk_data_t *mw = wsp->walk_data;
1597 metaslab_t ms;
1598 uintptr_t msp;
1599
1600 if (mw->mw_curvdev >= mw->mw_numvdevs)
1601 return (WALK_DONE);
1602
1603 if (mw->mw_mss == NULL) {
1604 uintptr_t mssp;
1605 uintptr_t vdevp;
1606
1607 ASSERT(mw->mw_curms == 0);
1608 ASSERT(mw->mw_nummss == 0);
1609
1610 vdevp = mw->mw_vdevs[mw->mw_curvdev];
1611 if (GETMEMB(vdevp, "vdev", vdev_ms, mssp) ||
1612 GETMEMB(vdevp, "vdev", vdev_ms_count, mw->mw_nummss)) {
1613 return (WALK_ERR);
1614 }
1615
1616 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*),
1617 UM_SLEEP | UM_GC);
1618 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*),
1619 mssp) == -1) {
1620 mdb_warn("failed to read vdev_ms at %p", mssp);
1621 return (WALK_ERR);
1622 }
1623 }
1624
1625 if (mw->mw_curms >= mw->mw_nummss) {
1626 mw->mw_mss = NULL;
1627 mw->mw_curms = 0;
1628 mw->mw_nummss = 0;
1629 mw->mw_curvdev++;
1630 return (WALK_NEXT);
1631 }
1632
1633 msp = mw->mw_mss[mw->mw_curms];
1634 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) {
1635 mdb_warn("failed to read metaslab_t at %p", msp);
1636 return (WALK_ERR);
1637 }
1638
1639 mw->mw_curms++;
1640
1641 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata));
1642 }
1643
1644 /* ARGSUSED */
1645 static int
1646 metaslab_walk_init(mdb_walk_state_t *wsp)
1647 {
1648 metaslab_walk_data_t *mw;
1649 uintptr_t root_vdevp;
1650 uintptr_t childp;
1651
1652 if (wsp->walk_addr == NULL) {
1653 mdb_warn("must supply address of spa_t\n");
1654 return (WALK_ERR);
1655 }
1656
1657 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC);
1658
1659 if (GETMEMB(wsp->walk_addr, "spa", spa_root_vdev, root_vdevp) ||
1660 GETMEMB(root_vdevp, "vdev", vdev_children, mw->mw_numvdevs) ||
1661 GETMEMB(root_vdevp, "vdev", vdev_child, childp)) {
1662 return (DCMD_ERR);
1663 }
1664
1665 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *),
1666 UM_SLEEP | UM_GC);
1667 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *),
1668 childp) == -1) {
1669 mdb_warn("failed to read root vdev children at %p", childp);
1670 return (DCMD_ERR);
1671 }
1672
1673 wsp->walk_data = mw;
1674
1675 return (WALK_NEXT);
1676 }
1677
1678 typedef struct mdb_spa {
1679 uintptr_t spa_dsl_pool;
1680 uintptr_t spa_root_vdev;
1681 } mdb_spa_t;
1682
1683 typedef struct mdb_dsl_dir {
1684 uintptr_t dd_phys;
1685 int64_t dd_space_towrite[TXG_SIZE];
1686 } mdb_dsl_dir_t;
1687
1688 typedef struct mdb_dsl_dir_phys {
1689 uint64_t dd_used_bytes;
1690 uint64_t dd_compressed_bytes;
1691 uint64_t dd_uncompressed_bytes;
1692 } mdb_dsl_dir_phys_t;
1693
1694 typedef struct space_data {
1695 uint64_t ms_alloctree[TXG_SIZE];
1696 uint64_t ms_freetree[TXG_SIZE];
1697 uint64_t ms_tree;
1698 uint64_t avail;
1699 uint64_t nowavail;
1700 } space_data_t;
1701
1702 /* ARGSUSED */
1703 static int
1704 space_cb(uintptr_t addr, const void *unknown, void *arg)
1705 {
1706 space_data_t *sd = arg;
1707 mdb_metaslab_t ms;
1708 mdb_range_tree_t rt;
1709 mdb_space_map_t sm = { 0 };
1710 mdb_space_map_phys_t smp = { 0 };
1711 int i;
1712
1713 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t",
1714 addr, 0) == -1)
1715 return (WALK_ERR);
1716
1717 for (i = 0; i < TXG_SIZE; i++) {
1718 if (mdb_ctf_vread(&rt, "range_tree_t",
1719 "mdb_range_tree_t", ms.ms_alloctree[i], 0) == -1)
1720 return (WALK_ERR);
1721
1722 sd->ms_alloctree[i] += rt.rt_space;
1723
1724 if (mdb_ctf_vread(&rt, "range_tree_t",
1725 "mdb_range_tree_t", ms.ms_freetree[i], 0) == -1)
1726 return (WALK_ERR);
1727
1728 sd->ms_freetree[i] += rt.rt_space;
1729 }
1730
1731 if (mdb_ctf_vread(&rt, "range_tree_t",
1732 "mdb_range_tree_t", ms.ms_tree, 0) == -1)
1733 return (WALK_ERR);
1734
1735 if (ms.ms_sm != NULL &&
1736 mdb_ctf_vread(&sm, "space_map_t",
1737 "mdb_space_map_t", ms.ms_sm, 0) == -1)
1738 return (WALK_ERR);
1739
1740 if (sm.sm_phys != NULL) {
1741 (void) mdb_ctf_vread(&smp, "space_map_phys_t",
1742 "mdb_space_map_phys_t", sm.sm_phys, 0);
1743 }
1744
1745 sd->ms_tree += rt.rt_space;
1746 sd->avail += sm.sm_size - sm.sm_alloc;
1747 sd->nowavail += sm.sm_size - smp.smp_alloc;
1748
1749 return (WALK_NEXT);
1750 }
1751
1752 /*
1753 * ::spa_space [-b]
1754 *
1755 * Given a spa_t, print out it's on-disk space usage and in-core
1756 * estimates of future usage. If -b is given, print space in bytes.
1757 * Otherwise print in megabytes.
1758 */
1759 /* ARGSUSED */
1760 static int
1761 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1762 {
1763 mdb_spa_t spa;
1764 uintptr_t dp_root_dir;
1765 mdb_dsl_dir_t dd;
1766 mdb_dsl_dir_phys_t dsp;
1767 uint64_t children;
1768 uintptr_t childaddr;
1769 space_data_t sd;
1770 int shift = 20;
1771 char *suffix = "M";
1772 int bytes = B_FALSE;
1773
1774 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) !=
1775 argc)
1776 return (DCMD_USAGE);
1777 if (!(flags & DCMD_ADDRSPEC))
1778 return (DCMD_USAGE);
1779
1780 if (bytes) {
1781 shift = 0;
1782 suffix = "";
1783 }
1784
1785 if (GETMEMB(addr, "spa", spa_dsl_pool, spa.spa_dsl_pool) ||
1786 GETMEMB(addr, "spa", spa_root_vdev, spa.spa_root_vdev) ||
1787 GETMEMB(spa.spa_root_vdev, "vdev", vdev_children, children) ||
1788 GETMEMB(spa.spa_root_vdev, "vdev", vdev_child, childaddr) ||
1789 GETMEMB(spa.spa_dsl_pool, "dsl_pool",
1790 dp_root_dir, dp_root_dir) ||
1791 GETMEMB(dp_root_dir, "dsl_dir", dd_phys, dd.dd_phys) ||
1792 GETMEMB(dp_root_dir, "dsl_dir",
1793 dd_space_towrite, dd.dd_space_towrite) ||
1794 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1795 dd_used_bytes, dsp.dd_used_bytes) ||
1796 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1797 dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1798 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1799 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1800 return (DCMD_ERR);
1801 }
1802
1803 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1804 dd.dd_space_towrite[0] >> shift, suffix,
1805 dd.dd_space_towrite[1] >> shift, suffix,
1806 dd.dd_space_towrite[2] >> shift, suffix,
1807 dd.dd_space_towrite[3] >> shift, suffix);
1808
1809 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1810 dsp.dd_used_bytes >> shift, suffix);
1811 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1812 dsp.dd_compressed_bytes >> shift, suffix);
1813 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1814 dsp.dd_uncompressed_bytes >> shift, suffix);
1815
1816 bzero(&sd, sizeof (sd));
1817 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1818 mdb_warn("can't walk metaslabs");
1819 return (DCMD_ERR);
1820 }
1821
1822 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1823 sd.ms_alloctree[0] >> shift, suffix,
1824 sd.ms_alloctree[1] >> shift, suffix,
1825 sd.ms_alloctree[2] >> shift, suffix,
1826 sd.ms_alloctree[3] >> shift, suffix);
1827 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1828 sd.ms_freetree[0] >> shift, suffix,
1829 sd.ms_freetree[1] >> shift, suffix,
1830 sd.ms_freetree[2] >> shift, suffix,
1831 sd.ms_freetree[3] >> shift, suffix);
1832 mdb_printf("ms_tree = %llu%s\n", sd.ms_tree >> shift, suffix);
1833 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1834 mdb_printf("current syncing avail = %llu%s\n",
1835 sd.nowavail >> shift, suffix);
1836
1837 return (DCMD_OK);
1838 }
1839
1840 typedef struct mdb_spa_aux_vdev {
1841 int sav_count;
1842 uintptr_t sav_vdevs;
1843 } mdb_spa_aux_vdev_t;
1844
1845 typedef struct mdb_spa_vdevs {
1846 uintptr_t spa_root_vdev;
1847 mdb_spa_aux_vdev_t spa_l2cache;
1848 mdb_spa_aux_vdev_t spa_spares;
1849 } mdb_spa_vdevs_t;
1850
1851 static int
1852 spa_print_aux(mdb_spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v,
1853 const char *name)
1854 {
1855 uintptr_t *aux;
1856 size_t len;
1857 int ret, i;
1858
1859 /*
1860 * Iterate over aux vdevs and print those out as well. This is a
1861 * little annoying because we don't have a root vdev to pass to ::vdev.
1862 * Instead, we print a single line and then call it for each child
1863 * vdev.
1864 */
1865 if (sav->sav_count != 0) {
1866 v[1].a_type = MDB_TYPE_STRING;
1867 v[1].a_un.a_str = "-d";
1868 v[2].a_type = MDB_TYPE_IMMEDIATE;
1869 v[2].a_un.a_val = 2;
1870
1871 len = sav->sav_count * sizeof (uintptr_t);
1872 aux = mdb_alloc(len, UM_SLEEP);
1873 if (mdb_vread(aux, len, sav->sav_vdevs) == -1) {
1874 mdb_free(aux, len);
1875 mdb_warn("failed to read l2cache vdevs at %p",
1876 sav->sav_vdevs);
1877 return (DCMD_ERR);
1878 }
1879
1880 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name);
1881
1882 for (i = 0; i < sav->sav_count; i++) {
1883 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v);
1884 if (ret != DCMD_OK) {
1885 mdb_free(aux, len);
1886 return (ret);
1887 }
1888 }
1889
1890 mdb_free(aux, len);
1891 }
1892
1893 return (0);
1894 }
1895
1896 /*
1897 * ::spa_vdevs
1898 *
1899 * -e Include error stats
1900 * -m Include metaslab information
1901 * -M Include metaslab group information
1902 * -h Include histogram information (requires -m or -M)
1903 *
1904 * Print out a summarized list of vdevs for the given spa_t.
1905 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as
1906 * iterating over the cache devices.
1907 */
1908 /* ARGSUSED */
1909 static int
1910 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1911 {
1912 mdb_arg_t v[3];
1913 int ret;
1914 char opts[100] = "-r";
1915 int spa_flags = 0;
1916
1917 if (mdb_getopts(argc, argv,
1918 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags,
1919 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags,
1920 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags,
1921 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags,
1922 NULL) != argc)
1923 return (DCMD_USAGE);
1924
1925 if (!(flags & DCMD_ADDRSPEC))
1926 return (DCMD_USAGE);
1927
1928 mdb_spa_vdevs_t spa;
1929 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_vdevs_t", addr, 0) == -1)
1930 return (DCMD_ERR);
1931
1932 /*
1933 * Unitialized spa_t structures can have a NULL root vdev.
1934 */
1935 if (spa.spa_root_vdev == NULL) {
1936 mdb_printf("no associated vdevs\n");
1937 return (DCMD_OK);
1938 }
1939
1940 if (spa_flags & SPA_FLAG_ERRORS)
1941 strcat(opts, "e");
1942 if (spa_flags & SPA_FLAG_METASLABS)
1943 strcat(opts, "m");
1944 if (spa_flags & SPA_FLAG_METASLAB_GROUPS)
1945 strcat(opts, "M");
1946 if (spa_flags & SPA_FLAG_HISTOGRAMS)
1947 strcat(opts, "h");
1948
1949 v[0].a_type = MDB_TYPE_STRING;
1950 v[0].a_un.a_str = opts;
1951
1952 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1953 flags, 1, v);
1954 if (ret != DCMD_OK)
1955 return (ret);
1956
1957 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 ||
1958 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0)
1959 return (DCMD_ERR);
1960
1961 return (DCMD_OK);
1962 }
1963
1964 /*
1965 * ::zio
1966 *
1967 * Print a summary of zio_t and all its children. This is intended to display a
1968 * zio tree, and hence we only pick the most important pieces of information for
1969 * the main summary. More detailed information can always be found by doing a
1970 * '::print zio' on the underlying zio_t. The columns we display are:
1971 *
1972 * ADDRESS TYPE STAGE WAITER TIME_ELAPSED
1973 *
1974 * The 'address' column is indented by one space for each depth level as we
1975 * descend down the tree.
1976 */
1977
1978 #define ZIO_MAXINDENT 7
1979 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT)
1980 #define ZIO_WALK_SELF 0
1981 #define ZIO_WALK_CHILD 1
1982 #define ZIO_WALK_PARENT 2
1983
1984 typedef struct zio_print_args {
1985 int zpa_current_depth;
1986 int zpa_min_depth;
1987 int zpa_max_depth;
1988 int zpa_type;
1989 uint_t zpa_flags;
1990 } zio_print_args_t;
1991
1992 typedef struct mdb_zio {
1993 enum zio_type io_type;
1994 enum zio_stage io_stage;
1995 uintptr_t io_waiter;
1996 uintptr_t io_spa;
1997 struct {
1998 struct {
1999 uintptr_t list_next;
2000 } list_head;
2001 } io_parent_list;
2002 int io_error;
2003 } mdb_zio_t;
2004
2005 typedef struct mdb_zio_timestamp {
2006 hrtime_t io_timestamp;
2007 } mdb_zio_timestamp_t;
2008
2009 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg);
2010
2011 static int
2012 zio_print_cb(uintptr_t addr, zio_print_args_t *zpa)
2013 {
2014 mdb_ctf_id_t type_enum, stage_enum;
2015 int indent = zpa->zpa_current_depth;
2016 const char *type, *stage;
2017 uintptr_t laddr;
2018 mdb_zio_t zio;
2019 mdb_zio_timestamp_t zio_timestamp = { 0 };
2020
2021 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1)
2022 return (WALK_ERR);
2023 (void) mdb_ctf_vread(&zio_timestamp, ZFS_STRUCT "zio",
2024 "mdb_zio_timestamp_t", addr, MDB_CTF_VREAD_QUIET);
2025
2026 if (indent > ZIO_MAXINDENT)
2027 indent = ZIO_MAXINDENT;
2028
2029 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 ||
2030 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) {
2031 mdb_warn("failed to lookup zio enums");
2032 return (WALK_ERR);
2033 }
2034
2035 if ((type = mdb_ctf_enum_name(type_enum, zio.io_type)) != NULL)
2036 type += sizeof ("ZIO_TYPE_") - 1;
2037 else
2038 type = "?";
2039
2040 if (zio.io_error == 0) {
2041 stage = mdb_ctf_enum_name(stage_enum, zio.io_stage);
2042 if (stage != NULL)
2043 stage += sizeof ("ZIO_STAGE_") - 1;
2044 else
2045 stage = "?";
2046 } else {
2047 stage = "FAILED";
2048 }
2049
2050 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) {
2051 if (zpa->zpa_flags & DCMD_PIPE_OUT) {
2052 mdb_printf("%?p\n", addr);
2053 } else {
2054 mdb_printf("%*s%-*p %-5s %-16s ", indent, "",
2055 ZIO_MAXWIDTH - indent, addr, type, stage);
2056 if (zio.io_waiter != 0)
2057 mdb_printf("%-16lx ", zio.io_waiter);
2058 else
2059 mdb_printf("%-16s ", "-");
2060 #ifdef _KERNEL
2061 if (zio_timestamp.io_timestamp != 0) {
2062 mdb_printf("%llums", (mdb_gethrtime() -
2063 zio_timestamp.io_timestamp) /
2064 1000000);
2065 } else {
2066 mdb_printf("%-12s ", "-");
2067 }
2068 #else
2069 mdb_printf("%-12s ", "-");
2070 #endif
2071 mdb_printf("\n");
2072 }
2073 }
2074
2075 if (zpa->zpa_current_depth >= zpa->zpa_max_depth)
2076 return (WALK_NEXT);
2077
2078 if (zpa->zpa_type == ZIO_WALK_PARENT)
2079 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
2080 "io_parent_list");
2081 else
2082 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
2083 "io_child_list");
2084
2085 zpa->zpa_current_depth++;
2086 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) {
2087 mdb_warn("failed to walk zio_t children at %p\n", laddr);
2088 return (WALK_ERR);
2089 }
2090 zpa->zpa_current_depth--;
2091
2092 return (WALK_NEXT);
2093 }
2094
2095 /* ARGSUSED */
2096 static int
2097 zio_child_cb(uintptr_t addr, const void *unknown, void *arg)
2098 {
2099 zio_link_t zl;
2100 uintptr_t ziop;
2101 zio_print_args_t *zpa = arg;
2102
2103 if (mdb_vread(&zl, sizeof (zl), addr) == -1) {
2104 mdb_warn("failed to read zio_link_t at %p", addr);
2105 return (WALK_ERR);
2106 }
2107
2108 if (zpa->zpa_type == ZIO_WALK_PARENT)
2109 ziop = (uintptr_t)zl.zl_parent;
2110 else
2111 ziop = (uintptr_t)zl.zl_child;
2112
2113 return (zio_print_cb(ziop, zpa));
2114 }
2115
2116 /* ARGSUSED */
2117 static int
2118 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2119 {
2120 zio_print_args_t zpa = { 0 };
2121
2122 if (!(flags & DCMD_ADDRSPEC))
2123 return (DCMD_USAGE);
2124
2125 if (mdb_getopts(argc, argv,
2126 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth,
2127 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type,
2128 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type,
2129 NULL) != argc)
2130 return (DCMD_USAGE);
2131
2132 zpa.zpa_flags = flags;
2133 if (zpa.zpa_max_depth != 0) {
2134 if (zpa.zpa_type == ZIO_WALK_SELF)
2135 zpa.zpa_type = ZIO_WALK_CHILD;
2136 } else if (zpa.zpa_type != ZIO_WALK_SELF) {
2137 zpa.zpa_min_depth = 1;
2138 zpa.zpa_max_depth = 1;
2139 }
2140
2141 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) {
2142 mdb_printf("%<u>%-*s %-5s %-16s %-16s %-12s%</u>\n",
2143 ZIO_MAXWIDTH, "ADDRESS", "TYPE", "STAGE", "WAITER",
2144 "TIME_ELAPSED");
2145 }
2146
2147 if (zio_print_cb(addr, &zpa) != WALK_NEXT)
2148 return (DCMD_ERR);
2149
2150 return (DCMD_OK);
2151 }
2152
2153 /*
2154 * [addr]::zio_state
2155 *
2156 * Print a summary of all zio_t structures on the system, or for a particular
2157 * pool. This is equivalent to '::walk zio_root | ::zio'.
2158 */
2159 /*ARGSUSED*/
2160 static int
2161 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2162 {
2163 /*
2164 * MDB will remember the last address of the pipeline, so if we don't
2165 * zero this we'll end up trying to walk zio structures for a
2166 * non-existent spa_t.
2167 */
2168 if (!(flags & DCMD_ADDRSPEC))
2169 addr = 0;
2170
2171 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr));
2172 }
2173
2174 typedef struct txg_list_walk_data {
2175 uintptr_t lw_head[TXG_SIZE];
2176 int lw_txgoff;
2177 int lw_maxoff;
2178 size_t lw_offset;
2179 void *lw_obj;
2180 } txg_list_walk_data_t;
2181
2182 static int
2183 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
2184 {
2185 txg_list_walk_data_t *lwd;
2186 txg_list_t list;
2187 int i;
2188
2189 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
2190 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
2191 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
2192 return (WALK_ERR);
2193 }
2194
2195 for (i = 0; i < TXG_SIZE; i++)
2196 lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
2197 lwd->lw_offset = list.tl_offset;
2198 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
2199 UM_SLEEP | UM_GC);
2200 lwd->lw_txgoff = txg;
2201 lwd->lw_maxoff = maxoff;
2202
2203 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
2204 wsp->walk_data = lwd;
2205
2206 return (WALK_NEXT);
2207 }
2208
2209 static int
2210 txg_list_walk_init(mdb_walk_state_t *wsp)
2211 {
2212 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
2213 }
2214
2215 static int
2216 txg_list0_walk_init(mdb_walk_state_t *wsp)
2217 {
2218 return (txg_list_walk_init_common(wsp, 0, 0));
2219 }
2220
2221 static int
2222 txg_list1_walk_init(mdb_walk_state_t *wsp)
2223 {
2224 return (txg_list_walk_init_common(wsp, 1, 1));
2225 }
2226
2227 static int
2228 txg_list2_walk_init(mdb_walk_state_t *wsp)
2229 {
2230 return (txg_list_walk_init_common(wsp, 2, 2));
2231 }
2232
2233 static int
2234 txg_list3_walk_init(mdb_walk_state_t *wsp)
2235 {
2236 return (txg_list_walk_init_common(wsp, 3, 3));
2237 }
2238
2239 static int
2240 txg_list_walk_step(mdb_walk_state_t *wsp)
2241 {
2242 txg_list_walk_data_t *lwd = wsp->walk_data;
2243 uintptr_t addr;
2244 txg_node_t *node;
2245 int status;
2246
2247 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
2248 lwd->lw_txgoff++;
2249 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
2250 }
2251
2252 if (wsp->walk_addr == NULL)
2253 return (WALK_DONE);
2254
2255 addr = wsp->walk_addr - lwd->lw_offset;
2256
2257 if (mdb_vread(lwd->lw_obj,
2258 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
2259 mdb_warn("failed to read list element at %#lx", addr);
2260 return (WALK_ERR);
2261 }
2262
2263 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
2264 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
2265 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
2266
2267 return (status);
2268 }
2269
2270 /*
2271 * ::walk spa
2272 *
2273 * Walk all named spa_t structures in the namespace. This is nothing more than
2274 * a layered avl walk.
2275 */
2276 static int
2277 spa_walk_init(mdb_walk_state_t *wsp)
2278 {
2279 GElf_Sym sym;
2280
2281 if (wsp->walk_addr != NULL) {
2282 mdb_warn("spa walk only supports global walks\n");
2283 return (WALK_ERR);
2284 }
2285
2286 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
2287 mdb_warn("failed to find symbol 'spa_namespace_avl'");
2288 return (WALK_ERR);
2289 }
2290
2291 wsp->walk_addr = (uintptr_t)sym.st_value;
2292
2293 if (mdb_layered_walk("avl", wsp) == -1) {
2294 mdb_warn("failed to walk 'avl'\n");
2295 return (WALK_ERR);
2296 }
2297
2298 return (WALK_NEXT);
2299 }
2300
2301 static int
2302 spa_walk_step(mdb_walk_state_t *wsp)
2303 {
2304 return (wsp->walk_callback(wsp->walk_addr, NULL, wsp->walk_cbdata));
2305 }
2306
2307 /*
2308 * [addr]::walk zio
2309 *
2310 * Walk all active zio_t structures on the system. This is simply a layered
2311 * walk on top of ::walk zio_cache, with the optional ability to limit the
2312 * structures to a particular pool.
2313 */
2314 static int
2315 zio_walk_init(mdb_walk_state_t *wsp)
2316 {
2317 wsp->walk_data = (void *)wsp->walk_addr;
2318
2319 if (mdb_layered_walk("zio_cache", wsp) == -1) {
2320 mdb_warn("failed to walk 'zio_cache'\n");
2321 return (WALK_ERR);
2322 }
2323
2324 return (WALK_NEXT);
2325 }
2326
2327 static int
2328 zio_walk_step(mdb_walk_state_t *wsp)
2329 {
2330 mdb_zio_t zio;
2331 uintptr_t spa = (uintptr_t)wsp->walk_data;
2332
2333 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2334 wsp->walk_addr, 0) == -1)
2335 return (WALK_ERR);
2336
2337 if (spa != 0 && spa != zio.io_spa)
2338 return (WALK_NEXT);
2339
2340 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2341 }
2342
2343 /*
2344 * [addr]::walk zio_root
2345 *
2346 * Walk only root zio_t structures, optionally for a particular spa_t.
2347 */
2348 static int
2349 zio_walk_root_step(mdb_walk_state_t *wsp)
2350 {
2351 mdb_zio_t zio;
2352 uintptr_t spa = (uintptr_t)wsp->walk_data;
2353
2354 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2355 wsp->walk_addr, 0) == -1)
2356 return (WALK_ERR);
2357
2358 if (spa != 0 && spa != zio.io_spa)
2359 return (WALK_NEXT);
2360
2361 /* If the parent list is not empty, ignore */
2362 if (zio.io_parent_list.list_head.list_next !=
2363 wsp->walk_addr +
2364 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") +
2365 mdb_ctf_offsetof_by_name("struct list", "list_head"))
2366 return (WALK_NEXT);
2367
2368 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2369 }
2370
2371 /*
2372 * ::zfs_blkstats
2373 *
2374 * -v print verbose per-level information
2375 *
2376 */
2377 static int
2378 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2379 {
2380 boolean_t verbose = B_FALSE;
2381 zfs_all_blkstats_t stats;
2382 dmu_object_type_t t;
2383 zfs_blkstat_t *tzb;
2384 uint64_t ditto;
2385 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10];
2386 /* +10 in case it grew */
2387
2388 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) {
2389 mdb_warn("failed to read 'dmu_ot'");
2390 return (DCMD_ERR);
2391 }
2392
2393 if (mdb_getopts(argc, argv,
2394 'v', MDB_OPT_SETBITS, TRUE, &verbose,
2395 NULL) != argc)
2396 return (DCMD_USAGE);
2397
2398 if (!(flags & DCMD_ADDRSPEC))
2399 return (DCMD_USAGE);
2400
2401 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) ||
2402 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) ||
2403 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) {
2404 mdb_warn("failed to read data at %p;", addr);
2405 mdb_printf("maybe no stats? run \"zpool scrub\" first.");
2406 return (DCMD_ERR);
2407 }
2408
2409 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL];
2410 if (tzb->zb_gangs != 0) {
2411 mdb_printf("Ganged blocks: %llu\n",
2412 (longlong_t)tzb->zb_gangs);
2413 }
2414
2415 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev +
2416 tzb->zb_ditto_3_of_3_samevdev;
2417 if (ditto != 0) {
2418 mdb_printf("Dittoed blocks on same vdev: %llu\n",
2419 (longlong_t)ditto);
2420 }
2421
2422 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
2423 "\t avg\t comp\t%%Total\tType\n");
2424
2425 for (t = 0; t <= DMU_OT_TOTAL; t++) {
2426 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN];
2427 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN];
2428 char avg[NICENUM_BUFLEN];
2429 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN];
2430 char typename[64];
2431 int l;
2432
2433
2434 if (t == DMU_OT_DEFERRED)
2435 strcpy(typename, "deferred free");
2436 else if (t == DMU_OT_OTHER)
2437 strcpy(typename, "other");
2438 else if (t == DMU_OT_TOTAL)
2439 strcpy(typename, "Total");
2440 else if (mdb_readstr(typename, sizeof (typename),
2441 (uintptr_t)dmu_ot[t].ot_name) == -1) {
2442 mdb_warn("failed to read type name");
2443 return (DCMD_ERR);
2444 }
2445
2446 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0)
2447 continue;
2448
2449 for (l = -1; l < DN_MAX_LEVELS; l++) {
2450 int level = (l == -1 ? DN_MAX_LEVELS : l);
2451 zfs_blkstat_t *zb = &stats.zab_type[level][t];
2452
2453 if (zb->zb_asize == 0)
2454 continue;
2455
2456 /*
2457 * Don't print each level unless requested.
2458 */
2459 if (!verbose && level != DN_MAX_LEVELS)
2460 continue;
2461
2462 /*
2463 * If all the space is level 0, don't print the
2464 * level 0 separately.
2465 */
2466 if (level == 0 && zb->zb_asize ==
2467 stats.zab_type[DN_MAX_LEVELS][t].zb_asize)
2468 continue;
2469
2470 mdb_nicenum(zb->zb_count, csize);
2471 mdb_nicenum(zb->zb_lsize, lsize);
2472 mdb_nicenum(zb->zb_psize, psize);
2473 mdb_nicenum(zb->zb_asize, asize);
2474 mdb_nicenum(zb->zb_asize / zb->zb_count, avg);
2475 (void) snprintfrac(comp, NICENUM_BUFLEN,
2476 zb->zb_lsize, zb->zb_psize, 2);
2477 (void) snprintfrac(pct, NICENUM_BUFLEN,
2478 100 * zb->zb_asize, tzb->zb_asize, 2);
2479
2480 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s"
2481 "\t%5s\t%6s\t",
2482 csize, lsize, psize, asize, avg, comp, pct);
2483
2484 if (level == DN_MAX_LEVELS)
2485 mdb_printf("%s\n", typename);
2486 else
2487 mdb_printf(" L%d %s\n",
2488 level, typename);
2489 }
2490 }
2491
2492 return (DCMD_OK);
2493 }
2494
2495 typedef struct mdb_reference {
2496 uintptr_t ref_holder;
2497 uintptr_t ref_removed;
2498 uint64_t ref_number;
2499 } mdb_reference_t;
2500
2501 /* ARGSUSED */
2502 static int
2503 reference_cb(uintptr_t addr, const void *ignored, void *arg)
2504 {
2505 mdb_reference_t ref;
2506 boolean_t holder_is_str = B_FALSE;
2507 char holder_str[128];
2508 boolean_t removed = (boolean_t)arg;
2509
2510 if (mdb_ctf_vread(&ref, "reference_t", "mdb_reference_t", addr,
2511 0) == -1)
2512 return (DCMD_ERR);
2513
2514 if (mdb_readstr(holder_str, sizeof (holder_str),
2515 ref.ref_holder) != -1)
2516 holder_is_str = strisprint(holder_str);
2517
2518 if (removed)
2519 mdb_printf("removed ");
2520 mdb_printf("reference ");
2521 if (ref.ref_number != 1)
2522 mdb_printf("with count=%llu ", ref.ref_number);
2523 mdb_printf("with tag %lx", ref.ref_holder);
2524 if (holder_is_str)
2525 mdb_printf(" \"%s\"", holder_str);
2526 mdb_printf(", held at:\n");
2527
2528 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
2529
2530 if (removed) {
2531 mdb_printf("removed at:\n");
2532 (void) mdb_call_dcmd("whatis", ref.ref_removed,
2533 DCMD_ADDRSPEC, 0, NULL);
2534 }
2535
2536 mdb_printf("\n");
2537
2538 return (WALK_NEXT);
2539 }
2540
2541 typedef struct mdb_refcount {
2542 uint64_t rc_count;
2543 } mdb_refcount_t;
2544
2545 typedef struct mdb_refcount_removed {
2546 uint64_t rc_removed_count;
2547 } mdb_refcount_removed_t;
2548
2549 typedef struct mdb_refcount_tracked {
2550 boolean_t rc_tracked;
2551 } mdb_refcount_tracked_t;
2552
2553 /* ARGSUSED */
2554 static int
2555 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2556 {
2557 mdb_refcount_t rc;
2558 mdb_refcount_removed_t rcr;
2559 mdb_refcount_tracked_t rct;
2560 int off;
2561 boolean_t released = B_FALSE;
2562
2563 if (!(flags & DCMD_ADDRSPEC))
2564 return (DCMD_USAGE);
2565
2566 if (mdb_getopts(argc, argv,
2567 'r', MDB_OPT_SETBITS, B_TRUE, &released,
2568 NULL) != argc)
2569 return (DCMD_USAGE);
2570
2571 if (mdb_ctf_vread(&rc, "refcount_t", "mdb_refcount_t", addr,
2572 0) == -1)
2573 return (DCMD_ERR);
2574
2575 if (mdb_ctf_vread(&rcr, "refcount_t", "mdb_refcount_removed_t", addr,
2576 MDB_CTF_VREAD_QUIET) == -1) {
2577 mdb_printf("refcount_t at %p has %llu holds (untracked)\n",
2578 addr, (longlong_t)rc.rc_count);
2579 return (DCMD_OK);
2580 }
2581
2582 if (mdb_ctf_vread(&rct, "refcount_t", "mdb_refcount_tracked_t", addr,
2583 MDB_CTF_VREAD_QUIET) == -1) {
2584 /* If this is an old target, it might be tracked. */
2585 rct.rc_tracked = B_TRUE;
2586 }
2587
2588 mdb_printf("refcount_t at %p has %llu current holds, "
2589 "%llu recently released holds\n",
2590 addr, (longlong_t)rc.rc_count, (longlong_t)rcr.rc_removed_count);
2591
2592 if (rct.rc_tracked && rc.rc_count > 0)
2593 mdb_printf("current holds:\n");
2594 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_list");
2595 if (off == -1)
2596 return (DCMD_ERR);
2597 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off);
2598
2599 if (released && rcr.rc_removed_count > 0) {
2600 mdb_printf("released holds:\n");
2601
2602 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_removed");
2603 if (off == -1)
2604 return (DCMD_ERR);
2605 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off);
2606 }
2607
2608 return (DCMD_OK);
2609 }
2610
2611 /* ARGSUSED */
2612 static int
2613 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2614 {
2615 sa_attr_table_t *table;
2616 sa_os_t sa_os;
2617 char *name;
2618 int i;
2619
2620 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) {
2621 mdb_warn("failed to read sa_os at %p", addr);
2622 return (DCMD_ERR);
2623 }
2624
2625 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2626 UM_SLEEP | UM_GC);
2627 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC);
2628
2629 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2630 (uintptr_t)sa_os.sa_attr_table) == -1) {
2631 mdb_warn("failed to read sa_os at %p", addr);
2632 return (DCMD_ERR);
2633 }
2634
2635 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n",
2636 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME");
2637 for (i = 0; i != sa_os.sa_num_attrs; i++) {
2638 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name);
2639 mdb_printf("%5x %8x %8x %8x %-s\n",
2640 (int)table[i].sa_attr, (int)table[i].sa_registered,
2641 (int)table[i].sa_length, table[i].sa_byteswap, name);
2642 }
2643
2644 return (DCMD_OK);
2645 }
2646
2647 static int
2648 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count)
2649 {
2650 uintptr_t idx_table;
2651
2652 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) {
2653 mdb_printf("can't find offset table in sa_idx_tab\n");
2654 return (-1);
2655 }
2656
2657 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t),
2658 UM_SLEEP | UM_GC);
2659
2660 if (mdb_vread(*off_tab,
2661 attr_count * sizeof (uint32_t), idx_table) == -1) {
2662 mdb_warn("failed to attribute offset table %p", idx_table);
2663 return (-1);
2664 }
2665
2666 return (DCMD_OK);
2667 }
2668
2669 /*ARGSUSED*/
2670 static int
2671 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2672 {
2673 uint32_t *offset_tab;
2674 int attr_count;
2675 uint64_t attr_id;
2676 uintptr_t attr_addr;
2677 uintptr_t bonus_tab, spill_tab;
2678 uintptr_t db_bonus, db_spill;
2679 uintptr_t os, os_sa;
2680 uintptr_t db_data;
2681
2682 if (argc != 1)
2683 return (DCMD_USAGE);
2684
2685 if (argv[0].a_type == MDB_TYPE_STRING)
2686 attr_id = mdb_strtoull(argv[0].a_un.a_str);
2687 else
2688 return (DCMD_USAGE);
2689
2690 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) ||
2691 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) ||
2692 GETMEMB(addr, "sa_handle", sa_os, os) ||
2693 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) ||
2694 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) {
2695 mdb_printf("Can't find necessary information in sa_handle "
2696 "in sa_handle\n");
2697 return (DCMD_ERR);
2698 }
2699
2700 if (GETMEMB(os, "objset", os_sa, os_sa)) {
2701 mdb_printf("Can't find os_sa in objset\n");
2702 return (DCMD_ERR);
2703 }
2704
2705 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) {
2706 mdb_printf("Can't find sa_num_attrs\n");
2707 return (DCMD_ERR);
2708 }
2709
2710 if (attr_id > attr_count) {
2711 mdb_printf("attribute id number is out of range\n");
2712 return (DCMD_ERR);
2713 }
2714
2715 if (bonus_tab) {
2716 if (sa_get_off_table(bonus_tab, &offset_tab,
2717 attr_count) == -1) {
2718 return (DCMD_ERR);
2719 }
2720
2721 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) {
2722 mdb_printf("can't find db_data in bonus dbuf\n");
2723 return (DCMD_ERR);
2724 }
2725 }
2726
2727 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) &&
2728 spill_tab == NULL) {
2729 mdb_printf("Attribute does not exist\n");
2730 return (DCMD_ERR);
2731 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) {
2732 if (sa_get_off_table(spill_tab, &offset_tab,
2733 attr_count) == -1) {
2734 return (DCMD_ERR);
2735 }
2736 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) {
2737 mdb_printf("can't find db_data in spill dbuf\n");
2738 return (DCMD_ERR);
2739 }
2740 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) {
2741 mdb_printf("Attribute does not exist\n");
2742 return (DCMD_ERR);
2743 }
2744 }
2745 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]);
2746 mdb_printf("%p\n", attr_addr);
2747 return (DCMD_OK);
2748 }
2749
2750 /* ARGSUSED */
2751 static int
2752 zfs_ace_print_common(uintptr_t addr, uint_t flags,
2753 uint64_t id, uint32_t access_mask, uint16_t ace_flags,
2754 uint16_t ace_type, int verbose)
2755 {
2756 if (DCMD_HDRSPEC(flags) && !verbose)
2757 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n",
2758 "ADDR", "FLAGS", "MASK", "TYPE", "ID");
2759
2760 if (!verbose) {
2761 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr,
2762 ace_flags, access_mask, ace_type, id);
2763 return (DCMD_OK);
2764 }
2765
2766 switch (ace_flags & ACE_TYPE_FLAGS) {
2767 case ACE_OWNER:
2768 mdb_printf("owner@:");
2769 break;
2770 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2771 mdb_printf("group@:");
2772 break;
2773 case ACE_EVERYONE:
2774 mdb_printf("everyone@:");
2775 break;
2776 case ACE_IDENTIFIER_GROUP:
2777 mdb_printf("group:%llx:", (u_longlong_t)id);
2778 break;
2779 case 0: /* User entry */
2780 mdb_printf("user:%llx:", (u_longlong_t)id);
2781 break;
2782 }
2783
2784 /* print out permission mask */
2785 if (access_mask & ACE_READ_DATA)
2786 mdb_printf("r");
2787 else
2788 mdb_printf("-");
2789 if (access_mask & ACE_WRITE_DATA)
2790 mdb_printf("w");
2791 else
2792 mdb_printf("-");
2793 if (access_mask & ACE_EXECUTE)
2794 mdb_printf("x");
2795 else
2796 mdb_printf("-");
2797 if (access_mask & ACE_APPEND_DATA)
2798 mdb_printf("p");
2799 else
2800 mdb_printf("-");
2801 if (access_mask & ACE_DELETE)
2802 mdb_printf("d");
2803 else
2804 mdb_printf("-");
2805 if (access_mask & ACE_DELETE_CHILD)
2806 mdb_printf("D");
2807 else
2808 mdb_printf("-");
2809 if (access_mask & ACE_READ_ATTRIBUTES)
2810 mdb_printf("a");
2811 else
2812 mdb_printf("-");
2813 if (access_mask & ACE_WRITE_ATTRIBUTES)
2814 mdb_printf("A");
2815 else
2816 mdb_printf("-");
2817 if (access_mask & ACE_READ_NAMED_ATTRS)
2818 mdb_printf("R");
2819 else
2820 mdb_printf("-");
2821 if (access_mask & ACE_WRITE_NAMED_ATTRS)
2822 mdb_printf("W");
2823 else
2824 mdb_printf("-");
2825 if (access_mask & ACE_READ_ACL)
2826 mdb_printf("c");
2827 else
2828 mdb_printf("-");
2829 if (access_mask & ACE_WRITE_ACL)
2830 mdb_printf("C");
2831 else
2832 mdb_printf("-");
2833 if (access_mask & ACE_WRITE_OWNER)
2834 mdb_printf("o");
2835 else
2836 mdb_printf("-");
2837 if (access_mask & ACE_SYNCHRONIZE)
2838 mdb_printf("s");
2839 else
2840 mdb_printf("-");
2841
2842 mdb_printf(":");
2843
2844 /* Print out inheritance flags */
2845 if (ace_flags & ACE_FILE_INHERIT_ACE)
2846 mdb_printf("f");
2847 else
2848 mdb_printf("-");
2849 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE)
2850 mdb_printf("d");
2851 else
2852 mdb_printf("-");
2853 if (ace_flags & ACE_INHERIT_ONLY_ACE)
2854 mdb_printf("i");
2855 else
2856 mdb_printf("-");
2857 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE)
2858 mdb_printf("n");
2859 else
2860 mdb_printf("-");
2861 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG)
2862 mdb_printf("S");
2863 else
2864 mdb_printf("-");
2865 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG)
2866 mdb_printf("F");
2867 else
2868 mdb_printf("-");
2869 if (ace_flags & ACE_INHERITED_ACE)
2870 mdb_printf("I");
2871 else
2872 mdb_printf("-");
2873
2874 switch (ace_type) {
2875 case ACE_ACCESS_ALLOWED_ACE_TYPE:
2876 mdb_printf(":allow\n");
2877 break;
2878 case ACE_ACCESS_DENIED_ACE_TYPE:
2879 mdb_printf(":deny\n");
2880 break;
2881 case ACE_SYSTEM_AUDIT_ACE_TYPE:
2882 mdb_printf(":audit\n");
2883 break;
2884 case ACE_SYSTEM_ALARM_ACE_TYPE:
2885 mdb_printf(":alarm\n");
2886 break;
2887 default:
2888 mdb_printf(":?\n");
2889 }
2890 return (DCMD_OK);
2891 }
2892
2893 /* ARGSUSED */
2894 static int
2895 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2896 {
2897 zfs_ace_t zace;
2898 int verbose = FALSE;
2899 uint64_t id;
2900
2901 if (!(flags & DCMD_ADDRSPEC))
2902 return (DCMD_USAGE);
2903
2904 if (mdb_getopts(argc, argv,
2905 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2906 return (DCMD_USAGE);
2907
2908 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) {
2909 mdb_warn("failed to read zfs_ace_t");
2910 return (DCMD_ERR);
2911 }
2912
2913 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 ||
2914 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2915 id = zace.z_fuid;
2916 else
2917 id = -1;
2918
2919 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask,
2920 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose));
2921 }
2922
2923 /* ARGSUSED */
2924 static int
2925 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2926 {
2927 ace_t ace;
2928 uint64_t id;
2929 int verbose = FALSE;
2930
2931 if (!(flags & DCMD_ADDRSPEC))
2932 return (DCMD_USAGE);
2933
2934 if (mdb_getopts(argc, argv,
2935 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2936 return (DCMD_USAGE);
2937
2938 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) {
2939 mdb_warn("failed to read ace_t");
2940 return (DCMD_ERR);
2941 }
2942
2943 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 ||
2944 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2945 id = ace.a_who;
2946 else
2947 id = -1;
2948
2949 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask,
2950 ace.a_flags, ace.a_type, verbose));
2951 }
2952
2953 typedef struct acl_dump_args {
2954 int a_argc;
2955 const mdb_arg_t *a_argv;
2956 uint16_t a_version;
2957 int a_flags;
2958 } acl_dump_args_t;
2959
2960 /* ARGSUSED */
2961 static int
2962 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg)
2963 {
2964 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2965
2966 if (acl_args->a_version == 1) {
2967 if (mdb_call_dcmd("zfs_ace", addr,
2968 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2969 acl_args->a_argv) != DCMD_OK) {
2970 return (WALK_ERR);
2971 }
2972 } else {
2973 if (mdb_call_dcmd("zfs_ace0", addr,
2974 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2975 acl_args->a_argv) != DCMD_OK) {
2976 return (WALK_ERR);
2977 }
2978 }
2979 acl_args->a_flags = DCMD_LOOP;
2980 return (WALK_NEXT);
2981 }
2982
2983 /* ARGSUSED */
2984 static int
2985 acl_cb(uintptr_t addr, const void *unknown, void *arg)
2986 {
2987 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2988
2989 if (acl_args->a_version == 1) {
2990 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb,
2991 arg, addr) != 0) {
2992 mdb_warn("can't walk ACEs");
2993 return (DCMD_ERR);
2994 }
2995 } else {
2996 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb,
2997 arg, addr) != 0) {
2998 mdb_warn("can't walk ACEs");
2999 return (DCMD_ERR);
3000 }
3001 }
3002 return (WALK_NEXT);
3003 }
3004
3005 /* ARGSUSED */
3006 static int
3007 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
3008 {
3009 zfs_acl_t zacl;
3010 int verbose = FALSE;
3011 acl_dump_args_t acl_args;
3012
3013 if (!(flags & DCMD_ADDRSPEC))
3014 return (DCMD_USAGE);
3015
3016 if (mdb_getopts(argc, argv,
3017 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
3018 return (DCMD_USAGE);
3019
3020 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) {
3021 mdb_warn("failed to read zfs_acl_t");
3022 return (DCMD_ERR);
3023 }
3024
3025 acl_args.a_argc = argc;
3026 acl_args.a_argv = argv;
3027 acl_args.a_version = zacl.z_version;
3028 acl_args.a_flags = DCMD_LOOPFIRST;
3029
3030 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) {
3031 mdb_warn("can't walk ACL");
3032 return (DCMD_ERR);
3033 }
3034
3035 return (DCMD_OK);
3036 }
3037
3038 /* ARGSUSED */
3039 static int
3040 zfs_acl_node_walk_init(mdb_walk_state_t *wsp)
3041 {
3042 if (wsp->walk_addr == NULL) {
3043 mdb_warn("must supply address of zfs_acl_node_t\n");
3044 return (WALK_ERR);
3045 }
3046
3047 wsp->walk_addr +=
3048 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl");
3049
3050 if (mdb_layered_walk("list", wsp) == -1) {
3051 mdb_warn("failed to walk 'list'\n");
3052 return (WALK_ERR);
3053 }
3054
3055 return (WALK_NEXT);
3056 }
3057
3058 static int
3059 zfs_acl_node_walk_step(mdb_walk_state_t *wsp)
3060 {
3061 zfs_acl_node_t aclnode;
3062
3063 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t),
3064 wsp->walk_addr) == -1) {
3065 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr);
3066 return (WALK_ERR);
3067 }
3068
3069 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata));
3070 }
3071
3072 typedef struct ace_walk_data {
3073 int ace_count;
3074 int ace_version;
3075 } ace_walk_data_t;
3076
3077 static int
3078 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version,
3079 int ace_count, uintptr_t ace_data)
3080 {
3081 ace_walk_data_t *ace_walk_data;
3082
3083 if (wsp->walk_addr == NULL) {
3084 mdb_warn("must supply address of zfs_acl_node_t\n");
3085 return (WALK_ERR);
3086 }
3087
3088 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC);
3089
3090 ace_walk_data->ace_count = ace_count;
3091 ace_walk_data->ace_version = version;
3092
3093 wsp->walk_addr = ace_data;
3094 wsp->walk_data = ace_walk_data;
3095
3096 return (WALK_NEXT);
3097 }
3098
3099 static int
3100 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version)
3101 {
3102 static int gotid;
3103 static mdb_ctf_id_t acl_id;
3104 int z_ace_count;
3105 uintptr_t z_acldata;
3106
3107 if (!gotid) {
3108 if (mdb_ctf_lookup_by_name("struct zfs_acl_node",
3109 &acl_id) == -1) {
3110 mdb_warn("couldn't find struct zfs_acl_node");
3111 return (DCMD_ERR);
3112 }
3113 gotid = TRUE;
3114 }
3115
3116 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) {
3117 return (DCMD_ERR);
3118 }
3119 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) {
3120 return (DCMD_ERR);
3121 }
3122
3123 return (zfs_aces_walk_init_common(wsp, version,
3124 z_ace_count, z_acldata));
3125 }
3126
3127 /* ARGSUSED */
3128 static int
3129 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp)
3130 {
3131 return (zfs_acl_node_aces_walk_init_common(wsp, 1));
3132 }
3133
3134 /* ARGSUSED */
3135 static int
3136 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp)
3137 {
3138 return (zfs_acl_node_aces_walk_init_common(wsp, 0));
3139 }
3140
3141 static int
3142 zfs_aces_walk_step(mdb_walk_state_t *wsp)
3143 {
3144 ace_walk_data_t *ace_data = wsp->walk_data;
3145 zfs_ace_t zace;
3146 ace_t *acep;
3147 int status;
3148 int entry_type;
3149 int allow_type;
3150 uintptr_t ptr;
3151
3152 if (ace_data->ace_count == 0)
3153 return (WALK_DONE);
3154
3155 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) {
3156 mdb_warn("failed to read zfs_ace_t at %#lx",
3157 wsp->walk_addr);
3158 return (WALK_ERR);
3159 }
3160
3161 switch (ace_data->ace_version) {
3162 case 0:
3163 acep = (ace_t *)&zace;
3164 entry_type = acep->a_flags & ACE_TYPE_FLAGS;
3165 allow_type = acep->a_type;
3166 break;
3167 case 1:
3168 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS;
3169 allow_type = zace.z_hdr.z_type;
3170 break;
3171 default:
3172 return (WALK_ERR);
3173 }
3174
3175 ptr = (uintptr_t)wsp->walk_addr;
3176 switch (entry_type) {
3177 case ACE_OWNER:
3178 case ACE_EVERYONE:
3179 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
3180 ptr += ace_data->ace_version == 0 ?
3181 sizeof (ace_t) : sizeof (zfs_ace_hdr_t);
3182 break;
3183 case ACE_IDENTIFIER_GROUP:
3184 default:
3185 switch (allow_type) {
3186 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
3187 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
3188 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
3189 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
3190 ptr += ace_data->ace_version == 0 ?
3191 sizeof (ace_t) : sizeof (zfs_object_ace_t);
3192 break;
3193 default:
3194 ptr += ace_data->ace_version == 0 ?
3195 sizeof (ace_t) : sizeof (zfs_ace_t);
3196 break;
3197 }
3198 }
3199
3200 ace_data->ace_count--;
3201 status = wsp->walk_callback(wsp->walk_addr,
3202 (void *)(uintptr_t)&zace, wsp->walk_cbdata);
3203
3204 wsp->walk_addr = ptr;
3205 return (status);
3206 }
3207
3208 typedef struct mdb_zfs_rrwlock {
3209 uintptr_t rr_writer;
3210 boolean_t rr_writer_wanted;
3211 } mdb_zfs_rrwlock_t;
3212
3213 static uint_t rrw_key;
3214
3215 /* ARGSUSED */
3216 static int
3217 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
3218 {
3219 mdb_zfs_rrwlock_t rrw;
3220
3221 if (rrw_key == 0) {
3222 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1)
3223 return (DCMD_ERR);
3224 }
3225
3226 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr,
3227 0) == -1)
3228 return (DCMD_ERR);
3229
3230 if (rrw.rr_writer != 0) {
3231 mdb_printf("write lock held by thread %lx\n", rrw.rr_writer);
3232 return (DCMD_OK);
3233 }
3234
3235 if (rrw.rr_writer_wanted) {
3236 mdb_printf("writer wanted\n");
3237 }
3238
3239 mdb_printf("anonymous references:\n");
3240 (void) mdb_call_dcmd("refcount", addr +
3241 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"),
3242 DCMD_ADDRSPEC, 0, NULL);
3243
3244 mdb_printf("linked references:\n");
3245 (void) mdb_call_dcmd("refcount", addr +
3246 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"),
3247 DCMD_ADDRSPEC, 0, NULL);
3248
3249 /*
3250 * XXX This should find references from
3251 * "::walk thread | ::tsd -v <rrw_key>", but there is no support
3252 * for programmatic consumption of dcmds, so this would be
3253 * difficult, potentially requiring reimplementing ::tsd (both
3254 * user and kernel versions) in this MDB module.
3255 */
3256
3257 return (DCMD_OK);
3258 }
3259
3260 /*
3261 * MDB module linkage information:
3262 *
3263 * We declare a list of structures describing our dcmds, and a function
3264 * named _mdb_init to return a pointer to our module information.
3265 */
3266
3267 static const mdb_dcmd_t dcmds[] = {
3268 { "arc", "[-bkmg]", "print ARC variables", arc_print },
3269 { "blkptr", ":", "print blkptr_t", blkptr },
3270 { "dbuf", ":", "print dmu_buf_impl_t", dbuf },
3271 { "dbuf_stats", ":", "dbuf stats", dbuf_stats },
3272 { "dbufs",
3273 "\t[-O objset_t*] [-n objset_name | \"mos\"] "
3274 "[-o object | \"mdn\"] \n"
3275 "\t[-l level] [-b blkid | \"bonus\"]",
3276 "find dmu_buf_impl_t's that match specified criteria", dbufs },
3277 { "abuf_find", "dva_word[0] dva_word[1]",
3278 "find arc_buf_hdr_t of a specified DVA",
3279 abuf_find },
3280 { "spa", "?[-cevmMh]\n"
3281 "\t-c display spa config\n"
3282 "\t-e display vdev statistics\n"
3283 "\t-v display vdev information\n"
3284 "\t-m display metaslab statistics\n"
3285 "\t-M display metaslab group statistics\n"
3286 "\t-h display histogram (requires -m or -M)\n",
3287 "spa_t summary", spa_print },
3288 { "spa_config", ":", "print spa_t configuration", spa_print_config },
3289 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
3290 { "spa_vdevs", ":[-emMh]\n"
3291 "\t-e display vdev statistics\n"
3292 "\t-m dispaly metaslab statistics\n"
3293 "\t-M display metaslab group statistic\n"
3294 "\t-h display histogram (requires -m or -M)\n",
3295 "given a spa_t, print vdev summary", spa_vdevs },
3296 { "vdev", ":[-re]\n"
3297 "\t-r display recursively\n"
3298 "\t-e display statistics\n"
3299 "\t-m display metaslab statistics\n"
3300 "\t-M display metaslab group statistics\n"
3301 "\t-h display histogram (requires -m or -M)\n",
3302 "vdev_t summary", vdev_print },
3303 { "zio", ":[-cpr]\n"
3304 "\t-c display children\n"
3305 "\t-p display parents\n"
3306 "\t-r display recursively",
3307 "zio_t summary", zio_print },
3308 { "zio_state", "?", "print out all zio_t structures on system or "
3309 "for a particular pool", zio_state },
3310 { "zfs_blkstats", ":[-v]",
3311 "given a spa_t, print block type stats from last scrub",
3312 zfs_blkstats },
3313 { "zfs_params", "", "print zfs tunable parameters", zfs_params },
3314 { "refcount", ":[-r]\n"
3315 "\t-r display recently removed references",
3316 "print refcount_t holders", refcount },
3317 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf },
3318 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t",
3319 zfs_acl_dump },
3320 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print },
3321 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print },
3322 { "sa_attr_table", ":", "print SA attribute table from sa_os_t",
3323 sa_attr_table},
3324 { "sa_attr", ": attr_id",
3325 "print SA attribute address when given sa_handle_t", sa_attr_print},
3326 { "zfs_dbgmsg", ":[-va]",
3327 "print zfs debug log", dbgmsg},
3328 { "rrwlock", ":",
3329 "print rrwlock_t, including readers", rrwlock},
3330 { NULL }
3331 };
3332
3333 static const mdb_walker_t walkers[] = {
3334 { "zms_freelist", "walk ZFS metaslab freelist",
3335 freelist_walk_init, freelist_walk_step, NULL },
3336 { "txg_list", "given any txg_list_t *, walk all entries in all txgs",
3337 txg_list_walk_init, txg_list_walk_step, NULL },
3338 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
3339 txg_list0_walk_init, txg_list_walk_step, NULL },
3340 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
3341 txg_list1_walk_init, txg_list_walk_step, NULL },
3342 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
3343 txg_list2_walk_init, txg_list_walk_step, NULL },
3344 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
3345 txg_list3_walk_init, txg_list_walk_step, NULL },
3346 { "zio", "walk all zio structures, optionally for a particular spa_t",
3347 zio_walk_init, zio_walk_step, NULL },
3348 { "zio_root",
3349 "walk all root zio_t structures, optionally for a particular spa_t",
3350 zio_walk_init, zio_walk_root_step, NULL },
3351 { "spa", "walk all spa_t entries in the namespace",
3352 spa_walk_init, spa_walk_step, NULL },
3353 { "metaslab", "given a spa_t *, walk all metaslab_t structures",
3354 metaslab_walk_init, metaslab_walk_step, NULL },
3355 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes",
3356 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL },
3357 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs",
3358 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL },
3359 { "zfs_acl_node_aces0",
3360 "given a zfs_acl_node_t, walk all ACEs as ace_t",
3361 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL },
3362 { NULL }
3363 };
3364
3365 static const mdb_modinfo_t modinfo = {
3366 MDB_API_VERSION, dcmds, walkers
3367 };
3368
3369 const mdb_modinfo_t *
3370 _mdb_init(void)
3371 {
3372 return (&modinfo);
3373 }
Unleashed OS