2 * Copyright (c) 2008 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * $DragonFly: src/sbin/hammer/cmd_blockmap.c,v 1.4 2008/07/19 18:48:14 dillon Exp $
40 * Each collect covers 1<<(19+23) bytes address space of layer 1.
41 * (plus a copy of 1<<23 bytes that holds layer2 entries in layer 1).
43 typedef struct collect
{
44 RB_ENTRY(collect
) entry
;
45 hammer_off_t phys_offset
; /* layer2 address pointed by layer1 */
46 hammer_off_t
*offsets
; /* big-block offset for layer2[i] */
47 struct hammer_blockmap_layer2
*track2
; /* track of layer2 entries */
48 struct hammer_blockmap_layer2
*layer2
; /* 1<<19 x 16 bytes entries */
49 int error
; /* # of inconsistencies */
53 collect_compare(struct collect
*c1
, struct collect
*c2
)
55 if (c1
->phys_offset
< c2
->phys_offset
)
57 if (c1
->phys_offset
> c2
->phys_offset
)
62 RB_HEAD(collect_rb_tree
, collect
) CollectTree
= RB_INITIALIZER(&CollectTree
);
63 RB_PROTOTYPE2(collect_rb_tree
, collect
, entry
, collect_compare
, hammer_off_t
);
64 RB_GENERATE2(collect_rb_tree
, collect
, entry
, collect_compare
, hammer_off_t
,
67 static void dump_blockmap(const char *label
, int zone
);
68 static void check_freemap(hammer_blockmap_t freemap
);
69 static void check_btree_node(hammer_off_t node_offset
, int depth
);
70 static void check_undo(hammer_blockmap_t undomap
);
71 static __inline
void collect_btree_root(hammer_off_t node_offset
);
72 static __inline
void collect_btree_internal(hammer_btree_elm_t elm
);
73 static __inline
void collect_btree_leaf(hammer_btree_elm_t elm
);
74 static __inline
void collect_freemap_layer1(hammer_blockmap_t freemap
);
75 static __inline
void collect_freemap_layer2(struct hammer_blockmap_layer1
*layer1
);
76 static __inline
void collect_undo(hammer_off_t scan_offset
,
77 hammer_fifo_head_t head
);
78 static void collect_blockmap(hammer_off_t offset
, int32_t length
, int zone
);
79 static struct hammer_blockmap_layer2
*collect_get_track(
80 collect_t collect
, hammer_off_t offset
, int zone
,
81 struct hammer_blockmap_layer2
*layer2
);
82 static collect_t
collect_get(hammer_off_t phys_offset
);
83 static void dump_collect_table(void);
84 static void dump_collect(collect_t collect
, struct zone_stat
*stats
);
86 static int num_bad_layer1
= 0;
87 static int num_bad_layer2
= 0;
88 static int num_bad_node
= 0;
91 hammer_cmd_blockmap(void)
93 dump_blockmap("freemap", HAMMER_ZONE_FREEMAP_INDEX
);
98 dump_blockmap(const char *label
, int zone
)
100 struct volume_info
*root_volume
;
101 hammer_blockmap_t rootmap
;
102 hammer_blockmap_t blockmap
;
103 struct hammer_blockmap_layer1
*layer1
;
104 struct hammer_blockmap_layer2
*layer2
;
105 struct buffer_info
*buffer1
= NULL
;
106 struct buffer_info
*buffer2
= NULL
;
107 hammer_off_t layer1_offset
;
108 hammer_off_t layer2_offset
;
111 struct zone_stat
*stats
= NULL
;
115 assert(RootVolNo
>= 0);
116 root_volume
= get_volume(RootVolNo
);
117 rootmap
= &root_volume
->ondisk
->vol0_blockmap
[zone
];
118 assert(rootmap
->phys_offset
!= 0);
121 "phys first next alloc\n");
122 for (i
= 0; i
< HAMMER_MAX_ZONES
; i
++) {
123 blockmap
= &root_volume
->ondisk
->vol0_blockmap
[i
];
124 if (VerboseOpt
|| i
== zone
) {
125 printf("zone %-2d %-10s %016jx %016jx %016jx %016jx\n",
126 i
, (i
== zone
? label
: ""),
127 (uintmax_t)blockmap
->phys_offset
,
128 (uintmax_t)blockmap
->first_offset
,
129 (uintmax_t)blockmap
->next_offset
,
130 (uintmax_t)blockmap
->alloc_offset
);
135 stats
= hammer_init_zone_stat();
137 for (scan1
= HAMMER_ZONE_ENCODE(zone
, 0);
138 scan1
< HAMMER_ZONE_ENCODE(zone
, HAMMER_OFF_LONG_MASK
);
139 scan1
+= HAMMER_BLOCKMAP_LAYER2
) {
143 layer1_offset
= rootmap
->phys_offset
+
144 HAMMER_BLOCKMAP_LAYER1_OFFSET(scan1
);
145 layer1
= get_buffer_data(layer1_offset
, &buffer1
, 0);
147 xerr
= ' '; /* good */
148 if (layer1
->layer1_crc
!=
149 crc32(layer1
, HAMMER_LAYER1_CRCSIZE
)) {
154 layer1
->phys_offset
== HAMMER_BLOCKMAP_UNAVAIL
) {
157 printf("%c layer1 %016jx @%016jx blocks-free %jd\n",
160 (uintmax_t)layer1
->phys_offset
,
161 (intmax_t)layer1
->blocks_free
);
164 scan2
< scan1
+ HAMMER_BLOCKMAP_LAYER2
;
165 scan2
+= HAMMER_BIGBLOCK_SIZE
) {
167 * Dive layer 2, each entry represents a big-block.
169 layer2_offset
= layer1
->phys_offset
+
170 HAMMER_BLOCKMAP_LAYER2_OFFSET(scan2
);
171 layer2
= get_buffer_data(layer2_offset
, &buffer2
, 0);
173 xerr
= ' '; /* good */
174 if (layer2
->entry_crc
!=
175 crc32(layer2
, HAMMER_LAYER2_CRCSIZE
)) {
179 printf("%c %016jx zone=%-2d ",
184 printf("vol=%-3d L1=%-7lu L2=%-7lu ",
185 HAMMER_VOL_DECODE(scan2
),
186 HAMMER_BLOCKMAP_LAYER1_OFFSET(scan2
),
187 HAMMER_BLOCKMAP_LAYER2_OFFSET(scan2
));
188 else if (VerboseOpt
> 0)
189 printf("vol=%-3d L1=%-6lu L2=%-6lu ",
190 HAMMER_VOL_DECODE(scan2
),
191 HAMMER_BLOCKMAP_LAYER1_INDEX(scan2
),
192 HAMMER_BLOCKMAP_LAYER2_INDEX(scan2
));
193 printf("app=%-7d free=%-7d",
197 printf(" crc=%04x-%04x\n",
204 hammer_add_zone_stat_layer2(stats
, layer2
);
209 rel_volume(root_volume
);
212 hammer_print_zone_stat(stats
);
213 hammer_cleanup_zone_stat(stats
);
216 if (num_bad_layer1
|| VerboseOpt
) {
217 printf("%d bad layer1\n", num_bad_layer1
);
219 if (num_bad_layer2
|| VerboseOpt
) {
220 printf("%d bad layer2\n", num_bad_layer1
);
225 hammer_cmd_checkmap(void)
227 struct volume_info
*volume
;
228 hammer_blockmap_t freemap
;
229 hammer_blockmap_t undomap
;
230 hammer_off_t node_offset
;
232 volume
= get_volume(RootVolNo
);
233 node_offset
= volume
->ondisk
->vol0_btree_root
;
234 freemap
= &volume
->ondisk
->vol0_blockmap
[HAMMER_ZONE_FREEMAP_INDEX
];
235 undomap
= &volume
->ondisk
->vol0_blockmap
[HAMMER_ZONE_UNDO_INDEX
];
238 printf("Volume header\trecords=%jd next_tid=%016jx\n",
239 (intmax_t)volume
->ondisk
->vol0_stat_records
,
240 (uintmax_t)volume
->ondisk
->vol0_next_tid
);
241 printf("\t\tbufoffset=%016jx\n",
242 (uintmax_t)volume
->ondisk
->vol_buf_beg
);
243 printf("\t\tundosize=%jdMB\n",
244 (intmax_t)((undomap
->alloc_offset
& HAMMER_OFF_LONG_MASK
)
249 assert(HAMMER_ZONE_UNDO_INDEX
< HAMMER_ZONE2_MAPPED_INDEX
);
250 assert(HAMMER_ZONE2_MAPPED_INDEX
< HAMMER_MAX_ZONES
);
251 AssertOnFailure
= (DebugOpt
!= 0);
253 printf("Collecting allocation info from freemap: ");
255 check_freemap(freemap
);
258 printf("Collecting allocation info from B-Tree: ");
260 check_btree_node(node_offset
, 0);
263 printf("Collecting allocation info from UNDO: ");
268 dump_collect_table();
273 check_freemap(hammer_blockmap_t freemap
)
276 struct buffer_info
*buffer1
= NULL
;
277 struct hammer_blockmap_layer1
*layer1
;
280 collect_freemap_layer1(freemap
);
282 for (i
= 0; i
< (int)HAMMER_BLOCKMAP_RADIX1
; ++i
) {
283 offset
= freemap
->phys_offset
+ i
* sizeof(*layer1
);
284 layer1
= get_buffer_data(offset
, &buffer1
, 0);
285 if (layer1
->phys_offset
!= HAMMER_BLOCKMAP_UNAVAIL
)
286 collect_freemap_layer2(layer1
);
292 check_btree_node(hammer_off_t node_offset
, int depth
)
294 struct buffer_info
*buffer
= NULL
;
295 hammer_node_ondisk_t node
;
296 hammer_btree_elm_t elm
;
298 char badc
= ' '; /* good */
299 char badm
= ' '; /* good */
302 collect_btree_root(node_offset
);
303 node
= get_node(node_offset
, &buffer
);
308 } else if (crc32(&node
->crc
+ 1, HAMMER_BTREE_CRCSIZE
) != node
->crc
) {
312 if (badm
!= ' ' || badc
!= ' ') { /* not good */
314 printf("%c%c NODE %016jx ",
315 badc
, badm
, (uintmax_t)node_offset
);
317 printf("(IO ERROR)\n");
320 printf("cnt=%02d p=%016jx type=%c depth=%d mirror=%016jx\n",
322 (uintmax_t)node
->parent
,
323 (node
->type
? node
->type
: '?'),
325 (uintmax_t)node
->mirror_tid
);
329 for (i
= 0; i
< node
->count
; ++i
) {
330 elm
= &node
->elms
[i
];
333 case HAMMER_BTREE_TYPE_INTERNAL
:
334 if (elm
->internal
.subtree_offset
) {
335 collect_btree_internal(elm
);
336 check_btree_node(elm
->internal
.subtree_offset
,
340 case HAMMER_BTREE_TYPE_LEAF
:
341 if (elm
->leaf
.data_offset
)
342 collect_btree_leaf(elm
);
354 check_undo(hammer_blockmap_t undomap
)
356 struct buffer_info
*buffer
= NULL
;
357 hammer_off_t scan_offset
;
358 hammer_fifo_head_t head
;
360 scan_offset
= HAMMER_ZONE_ENCODE(HAMMER_ZONE_UNDO_INDEX
, 0);
361 while (scan_offset
< undomap
->alloc_offset
) {
362 head
= get_buffer_data(scan_offset
, &buffer
, 0);
363 switch (head
->hdr_type
) {
364 case HAMMER_HEAD_TYPE_PAD
:
365 case HAMMER_HEAD_TYPE_DUMMY
:
366 case HAMMER_HEAD_TYPE_UNDO
:
367 case HAMMER_HEAD_TYPE_REDO
:
368 collect_undo(scan_offset
, head
);
375 if ((head
->hdr_size
& HAMMER_HEAD_ALIGN_MASK
) ||
376 head
->hdr_size
== 0 ||
377 head
->hdr_size
> HAMMER_UNDO_ALIGN
-
378 ((u_int
)scan_offset
& HAMMER_UNDO_MASK
)) {
379 printf("Illegal size, skipping to next boundary\n");
380 scan_offset
= (scan_offset
+ HAMMER_UNDO_MASK
) &
383 scan_offset
+= head
->hdr_size
;
391 collect_freemap_layer1(hammer_blockmap_t freemap
)
394 * This translation is necessary to do checkmap properly
395 * as zone4 is really just zone2 address space.
397 hammer_off_t zone4_offset
= hammer_xlate_to_zoneX(
398 HAMMER_ZONE_FREEMAP_INDEX
, freemap
->phys_offset
);
399 collect_blockmap(zone4_offset
, HAMMER_BIGBLOCK_SIZE
,
400 HAMMER_ZONE_FREEMAP_INDEX
);
405 collect_freemap_layer2(struct hammer_blockmap_layer1
*layer1
)
408 * This translation is necessary to do checkmap properly
409 * as zone4 is really just zone2 address space.
411 hammer_off_t zone4_offset
= hammer_xlate_to_zoneX(
412 HAMMER_ZONE_FREEMAP_INDEX
, layer1
->phys_offset
);
413 collect_blockmap(zone4_offset
, HAMMER_BIGBLOCK_SIZE
,
414 HAMMER_ZONE_FREEMAP_INDEX
);
419 collect_btree_root(hammer_off_t node_offset
)
421 collect_blockmap(node_offset
,
422 sizeof(struct hammer_node_ondisk
), /* 4KB */
423 HAMMER_ZONE_BTREE_INDEX
);
428 collect_btree_internal(hammer_btree_elm_t elm
)
430 collect_blockmap(elm
->internal
.subtree_offset
,
431 sizeof(struct hammer_node_ondisk
), /* 4KB */
432 HAMMER_ZONE_BTREE_INDEX
);
437 collect_btree_leaf(hammer_btree_elm_t elm
)
441 switch (elm
->base
.rec_type
) {
442 case HAMMER_RECTYPE_INODE
:
443 case HAMMER_RECTYPE_DIRENTRY
:
444 case HAMMER_RECTYPE_EXT
:
445 case HAMMER_RECTYPE_FIX
:
446 case HAMMER_RECTYPE_PFS
:
447 case HAMMER_RECTYPE_SNAPSHOT
:
448 case HAMMER_RECTYPE_CONFIG
:
449 zone
= HAMMER_ZONE_META_INDEX
;
451 case HAMMER_RECTYPE_DATA
:
452 case HAMMER_RECTYPE_DB
:
454 * There is an exceptional case where HAMMER uses
455 * HAMMER_ZONE_LARGE_DATA when the data length is
456 * >HAMMER_BUFSIZE/2 (not >=HAMMER_BUFSIZE).
457 * This exceptional case is currently being used
458 * by mirror write code, however the following code
459 * can ignore that and simply use the normal way
460 * of selecting a zone using >=HAMMER_BUFSIZE.
461 * See hammer_alloc_data() for details.
463 zone
= elm
->leaf
.data_len
>= HAMMER_BUFSIZE
?
464 HAMMER_ZONE_LARGE_DATA_INDEX
:
465 HAMMER_ZONE_SMALL_DATA_INDEX
;
468 zone
= HAMMER_ZONE_UNAVAIL_INDEX
;
471 collect_blockmap(elm
->leaf
.data_offset
,
472 (elm
->leaf
.data_len
+ 15) & ~15, zone
);
477 collect_undo(hammer_off_t scan_offset
, hammer_fifo_head_t head
)
479 collect_blockmap(scan_offset
, head
->hdr_size
,
480 HAMMER_ZONE_UNDO_INDEX
);
485 collect_blockmap(hammer_off_t offset
, int32_t length
, int zone
)
487 struct hammer_blockmap_layer1 layer1
;
488 struct hammer_blockmap_layer2 layer2
;
489 struct hammer_blockmap_layer2
*track2
;
490 hammer_off_t result_offset
;
494 result_offset
= blockmap_lookup(offset
, &layer1
, &layer2
, &error
);
495 if (AssertOnFailure
) {
496 assert(HAMMER_ZONE_DECODE(offset
) == zone
);
497 assert(HAMMER_ZONE_DECODE(result_offset
) ==
498 HAMMER_ZONE_RAW_BUFFER_INDEX
);
501 collect
= collect_get(layer1
.phys_offset
); /* layer2 address */
502 track2
= collect_get_track(collect
, result_offset
, zone
, &layer2
);
503 track2
->bytes_free
-= length
;
508 collect_get(hammer_off_t phys_offset
)
512 collect
= RB_LOOKUP(collect_rb_tree
, &CollectTree
, phys_offset
);
516 collect
= calloc(sizeof(*collect
), 1);
517 collect
->track2
= malloc(HAMMER_BIGBLOCK_SIZE
); /* 1<<23 bytes */
518 collect
->layer2
= malloc(HAMMER_BIGBLOCK_SIZE
); /* 1<<23 bytes */
519 collect
->offsets
= malloc(sizeof(hammer_off_t
) * HAMMER_BLOCKMAP_RADIX2
);
520 collect
->phys_offset
= phys_offset
;
521 RB_INSERT(collect_rb_tree
, &CollectTree
, collect
);
522 bzero(collect
->track2
, HAMMER_BIGBLOCK_SIZE
);
523 bzero(collect
->layer2
, HAMMER_BIGBLOCK_SIZE
);
530 collect_rel(collect_t collect
)
532 free(collect
->offsets
);
533 free(collect
->layer2
);
534 free(collect
->track2
);
539 struct hammer_blockmap_layer2
*
540 collect_get_track(collect_t collect
, hammer_off_t offset
, int zone
,
541 struct hammer_blockmap_layer2
*layer2
)
543 struct hammer_blockmap_layer2
*track2
;
546 i
= HAMMER_BLOCKMAP_LAYER2_INDEX(offset
);
547 track2
= &collect
->track2
[i
];
548 if (track2
->entry_crc
== 0) {
549 collect
->layer2
[i
] = *layer2
;
550 collect
->offsets
[i
] = offset
& ~HAMMER_BIGBLOCK_MASK64
;
552 track2
->bytes_free
= HAMMER_BIGBLOCK_SIZE
;
553 track2
->entry_crc
= 1; /* steal field to tag track load */
560 dump_collect_table(void)
564 struct zone_stat
*stats
= NULL
;
567 stats
= hammer_init_zone_stat();
569 RB_FOREACH(collect
, collect_rb_tree
, &CollectTree
) {
570 dump_collect(collect
, stats
);
571 error
+= collect
->error
;
574 while ((collect
= RB_ROOT(&CollectTree
)) != NULL
) {
575 RB_REMOVE(collect_rb_tree
, &CollectTree
, collect
);
576 collect_rel(collect
);
578 assert(RB_EMPTY(&CollectTree
));
581 hammer_print_zone_stat(stats
);
582 hammer_cleanup_zone_stat(stats
);
585 if (num_bad_node
|| VerboseOpt
) {
586 printf("%d bad nodes\n", num_bad_node
);
588 if (error
|| VerboseOpt
) {
589 printf("%d errors\n", error
);
595 dump_collect(collect_t collect
, struct zone_stat
*stats
)
597 struct hammer_blockmap_layer2
*track2
;
598 struct hammer_blockmap_layer2
*layer2
;
603 for (i
= 0; i
< HAMMER_BLOCKMAP_RADIX2
; ++i
) {
604 track2
= &collect
->track2
[i
];
605 layer2
= &collect
->layer2
[i
];
606 offset
= collect
->offsets
[i
];
609 * Check big-blocks referenced by freemap, data,
610 * B-Tree nodes and UNDO fifo.
612 if (track2
->entry_crc
== 0)
616 if (AssertOnFailure
) {
617 assert((zone
== HAMMER_ZONE_UNDO_INDEX
) ||
618 (zone
== HAMMER_ZONE_FREEMAP_INDEX
) ||
619 (zone
>= HAMMER_ZONE2_MAPPED_INDEX
&&
620 zone
< HAMMER_MAX_ZONES
));
623 hammer_add_zone_stat_layer2(stats
, layer2
);
625 if (track2
->zone
!= layer2
->zone
) {
626 printf("BZ\tblock=%016jx calc zone=%-2d, got zone=%-2d\n",
631 } else if (track2
->bytes_free
!= layer2
->bytes_free
) {
632 printf("BM\tblock=%016jx zone=%-2d calc %d free, got %d\n",
638 } else if (VerboseOpt
) {
639 printf("\tblock=%016jx zone=%-2d %d free (correct)\n",