xref: /dflybsd-src/sbin/hammer/cmd_blockmap.c (revision e8721bf4b6e6b12c2025fbba27e7b9214bb19756)
1 /*
2  * Copyright (c) 2008 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Matthew Dillon <dillon@backplane.com>
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
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
16  *    distribution.
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.
20  *
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
32  * SUCH DAMAGE.
33  *
34  * $DragonFly: src/sbin/hammer/cmd_blockmap.c,v 1.4 2008/07/19 18:48:14 dillon Exp $
35  */
36 
37 #include "hammer.h"
38 
39 /*
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).
42  */
43 typedef struct collect {
44 	RB_ENTRY(collect) entry;
45 	hammer_off_t	phys_offset;  /* layer2 address pointed by layer1 */
46 	struct hammer_blockmap_layer2 *track2;  /* track of layer2 entries */
47 	struct hammer_blockmap_layer2 *layer2;  /* 1<<19 x 16 bytes entries */
48 	int error;  /* # of inconsistencies */
49 } *collect_t;
50 
51 static int
52 collect_compare(struct collect *c1, struct collect *c2)
53 {
54 	if (c1->phys_offset < c2->phys_offset)
55 		return(-1);
56 	if (c1->phys_offset > c2->phys_offset)
57 		return(1);
58 	return(0);
59 }
60 
61 RB_HEAD(collect_rb_tree, collect) CollectTree = RB_INITIALIZER(&CollectTree);
62 RB_PROTOTYPE2(collect_rb_tree, collect, entry, collect_compare, hammer_off_t);
63 RB_GENERATE2(collect_rb_tree, collect, entry, collect_compare, hammer_off_t,
64 	phys_offset);
65 
66 static void dump_blockmap(const char *label, int zone);
67 static void check_btree_node(hammer_off_t node_offset, int depth);
68 static void check_undo(hammer_blockmap_t rootmap);
69 static __inline void collect_btree_root(hammer_off_t node_offset);
70 static __inline void collect_btree_internal(hammer_btree_elm_t elm);
71 static __inline void collect_btree_leaf(hammer_btree_elm_t elm);
72 static __inline void collect_undo(hammer_off_t scan_offset,
73 	hammer_fifo_head_t head);
74 static void collect_blockmap(hammer_off_t offset, int32_t length, int zone);
75 static struct hammer_blockmap_layer2 *collect_get_track(
76 	collect_t collect, hammer_off_t offset, int zone,
77 	struct hammer_blockmap_layer2 *layer2);
78 static collect_t collect_get(hammer_off_t phys_offset);
79 static void dump_collect_table(void);
80 static void dump_collect(collect_t collect, int *stats);
81 
82 void
83 hammer_cmd_blockmap(void)
84 {
85 	dump_blockmap("freemap", HAMMER_ZONE_FREEMAP_INDEX);
86 }
87 
88 static
89 void
90 dump_blockmap(const char *label, int zone)
91 {
92 	struct volume_info *root_volume;
93 	hammer_blockmap_t rootmap;
94 	struct hammer_blockmap_layer1 *layer1;
95 	struct hammer_blockmap_layer2 *layer2;
96 	struct buffer_info *buffer1 = NULL;
97 	struct buffer_info *buffer2 = NULL;
98 	hammer_off_t layer1_offset;
99 	hammer_off_t layer2_offset;
100 	hammer_off_t scan1;
101 	hammer_off_t scan2;
102 	int xerr;
103 
104 	assert(RootVolNo >= 0);
105 	root_volume = get_volume(RootVolNo);
106 	rootmap = &root_volume->ondisk->vol0_blockmap[zone];
107 	assert(rootmap->phys_offset != 0);
108 
109 	printf("zone %-16s next %016jx alloc %016jx\n",
110 		label,
111 		(uintmax_t)rootmap->next_offset,
112 		(uintmax_t)rootmap->alloc_offset);
113 
114 	for (scan1 = HAMMER_ZONE_ENCODE(zone, 0);
115 	     scan1 < HAMMER_ZONE_ENCODE(zone, HAMMER_OFF_LONG_MASK);
116 	     scan1 += HAMMER_BLOCKMAP_LAYER2) {
117 		/*
118 		 * Dive layer 1.
119 		 */
120 		layer1_offset = rootmap->phys_offset +
121 				HAMMER_BLOCKMAP_LAYER1_OFFSET(scan1);
122 		layer1 = get_buffer_data(layer1_offset, &buffer1, 0);
123 		xerr = ' ';
124 		if (layer1->layer1_crc != crc32(layer1, HAMMER_LAYER1_CRCSIZE))
125 			xerr = 'B';
126 		if (xerr == ' ' &&
127 		    layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL) {
128 			continue;
129 		}
130 		printf("%c layer1 %016jx @%016jx blocks-free %jd\n",
131 			xerr,
132 			(uintmax_t)scan1,
133 			(uintmax_t)layer1->phys_offset,
134 			(intmax_t)layer1->blocks_free);
135 		if (layer1->phys_offset == HAMMER_BLOCKMAP_FREE)
136 			continue;
137 		for (scan2 = scan1;
138 		     scan2 < scan1 + HAMMER_BLOCKMAP_LAYER2;
139 		     scan2 += HAMMER_BIGBLOCK_SIZE
140 		) {
141 			/*
142 			 * Dive layer 2, each entry represents a big-block.
143 			 */
144 			layer2_offset = layer1->phys_offset +
145 					HAMMER_BLOCKMAP_LAYER2_OFFSET(scan2);
146 			layer2 = get_buffer_data(layer2_offset, &buffer2, 0);
147 			xerr = ' ';
148 			if (layer2->entry_crc != crc32(layer2, HAMMER_LAYER2_CRCSIZE))
149 				xerr = 'B';
150 			printf("%c       %016jx zone=%d app=%-7d free=%-7d\n",
151 				xerr,
152 				(uintmax_t)scan2,
153 				layer2->zone,
154 				layer2->append_off,
155 				layer2->bytes_free);
156 		}
157 	}
158 	rel_buffer(buffer1);
159 	rel_buffer(buffer2);
160 	rel_volume(root_volume);
161 }
162 
163 void
164 hammer_cmd_checkmap(void)
165 {
166 	struct volume_info *volume;
167 	hammer_blockmap_t rootmap;
168 	hammer_off_t node_offset;
169 
170 	volume = get_volume(RootVolNo);
171 	node_offset = volume->ondisk->vol0_btree_root;
172 	rootmap = &volume->ondisk->vol0_blockmap[HAMMER_ZONE_UNDO_INDEX];
173 
174 	if (QuietOpt < 3) {
175 		printf("Volume header\trecords=%jd next_tid=%016jx\n",
176 		       (intmax_t)volume->ondisk->vol0_stat_records,
177 		       (uintmax_t)volume->ondisk->vol0_next_tid);
178 		printf("\t\tbufoffset=%016jx\n",
179 		       (uintmax_t)volume->ondisk->vol_buf_beg);
180 		printf("\t\tundosize=%jdMB\n",
181 		       (intmax_t)((rootmap->alloc_offset & HAMMER_OFF_LONG_MASK)
182 			/ (1024 * 1024)));
183 	}
184 	rel_volume(volume);
185 
186 	assert(HAMMER_ZONE_UNDO_INDEX < HAMMER_ZONE2_MAPPED_INDEX);
187 	assert(HAMMER_ZONE2_MAPPED_INDEX < HAMMER_MAX_ZONES);
188 	AssertOnFailure = 0;
189 
190 	printf("Collecting allocation info from B-Tree: ");
191 	fflush(stdout);
192 	collect_btree_root(node_offset);
193 	check_btree_node(node_offset, 0);
194 	printf("done\n");
195 
196 	printf("Collecting allocation info from UNDO: ");
197 	fflush(stdout);
198 	check_undo(rootmap);
199 	printf("done\n");
200 
201 	dump_collect_table();
202 	AssertOnFailure = 1;
203 }
204 
205 static void
206 check_btree_node(hammer_off_t node_offset, int depth)
207 {
208 	struct buffer_info *buffer = NULL;
209 	hammer_node_ondisk_t node;
210 	hammer_btree_elm_t elm;
211 	int i;
212 	char badc;
213 
214 	node = get_node(node_offset, &buffer);
215 
216 	if (crc32(&node->crc + 1, HAMMER_BTREE_CRCSIZE) == node->crc)
217 		badc = ' ';
218 	else
219 		badc = 'B';
220 
221 	if (badc != ' ') {
222 		printf("%c    NODE %016jx cnt=%02d p=%016jx "
223 		       "type=%c depth=%d",
224 		       badc,
225 		       (uintmax_t)node_offset, node->count,
226 		       (uintmax_t)node->parent,
227 		       (node->type ? node->type : '?'), depth);
228 		printf(" mirror %016jx\n", (uintmax_t)node->mirror_tid);
229 	}
230 
231 	for (i = 0; i < node->count; ++i) {
232 		elm = &node->elms[i];
233 
234 		switch(node->type) {
235 		case HAMMER_BTREE_TYPE_INTERNAL:
236 			if (elm->internal.subtree_offset) {
237 				collect_btree_internal(elm);
238 				check_btree_node(elm->internal.subtree_offset,
239 						 depth + 1);
240 			}
241 			break;
242 		case HAMMER_BTREE_TYPE_LEAF:
243 			if (elm->leaf.data_offset)
244 				collect_btree_leaf(elm);
245 			break;
246 		default:
247 			if (AssertOnFailure)
248 				assert(0);
249 			break;
250 		}
251 	}
252 	rel_buffer(buffer);
253 }
254 
255 static void
256 check_undo(hammer_blockmap_t rootmap)
257 {
258 	struct buffer_info *buffer = NULL;
259 	hammer_off_t scan_offset;
260 	hammer_fifo_head_t head;
261 
262 	scan_offset = HAMMER_ZONE_ENCODE(HAMMER_ZONE_UNDO_INDEX, 0);
263 	while (scan_offset < rootmap->alloc_offset) {
264 		head = get_buffer_data(scan_offset, &buffer, 0);
265 		switch (head->hdr_type) {
266 		case HAMMER_HEAD_TYPE_PAD:
267 		case HAMMER_HEAD_TYPE_DUMMY:
268 		case HAMMER_HEAD_TYPE_UNDO:
269 		case HAMMER_HEAD_TYPE_REDO:
270 			collect_undo(scan_offset, head);
271 			break;
272 		default:
273 			if (AssertOnFailure)
274 				assert(0);
275 			break;
276 		}
277 		if ((head->hdr_size & HAMMER_HEAD_ALIGN_MASK) ||
278 		     head->hdr_size == 0 ||
279 		     head->hdr_size > HAMMER_UNDO_ALIGN -
280 			((u_int)scan_offset & HAMMER_UNDO_MASK)) {
281 			printf("Illegal size, skipping to next boundary\n");
282 			scan_offset = (scan_offset + HAMMER_UNDO_MASK) &
283 					~HAMMER_UNDO_MASK64;
284 		} else {
285 			scan_offset += head->hdr_size;
286 		}
287 	}
288 	rel_buffer(buffer);
289 }
290 
291 static __inline
292 void
293 collect_btree_root(hammer_off_t node_offset)
294 {
295 	collect_blockmap(node_offset,
296 		sizeof(struct hammer_node_ondisk),  /* 4KB */
297 		HAMMER_ZONE_BTREE_INDEX);
298 }
299 
300 static __inline
301 void
302 collect_btree_internal(hammer_btree_elm_t elm)
303 {
304 	collect_blockmap(elm->internal.subtree_offset,
305 		sizeof(struct hammer_node_ondisk),  /* 4KB */
306 		HAMMER_ZONE_BTREE_INDEX);
307 }
308 
309 static __inline
310 void
311 collect_btree_leaf(hammer_btree_elm_t elm)
312 {
313 	int zone;
314 
315 	switch (elm->base.rec_type) {
316 	case HAMMER_RECTYPE_INODE:
317 	case HAMMER_RECTYPE_DIRENTRY:
318 	case HAMMER_RECTYPE_EXT:
319 	case HAMMER_RECTYPE_FIX:
320 	case HAMMER_RECTYPE_PFS:
321 	case HAMMER_RECTYPE_SNAPSHOT:
322 	case HAMMER_RECTYPE_CONFIG:
323 		zone = HAMMER_ZONE_META_INDEX;
324 		break;
325 	case HAMMER_RECTYPE_DATA:
326 	case HAMMER_RECTYPE_DB:
327 		/*
328 		 * There is an exceptional case where HAMMER uses
329 		 * HAMMER_ZONE_LARGE_DATA when the data length is
330 		 * >HAMMER_BUFSIZE/2 (not >=HAMMER_BUFSIZE).
331 		 * This exceptional case is currently being used
332 		 * by mirror write code, however the following code
333 		 * can ignore that and simply use the normal way
334 		 * of selecting a zone using >=HAMMER_BUFSIZE.
335 		 * See hammer_alloc_data() for details.
336 		 */
337 		zone = elm->leaf.data_len >= HAMMER_BUFSIZE ?
338 		       HAMMER_ZONE_LARGE_DATA_INDEX :
339 		       HAMMER_ZONE_SMALL_DATA_INDEX;
340 		break;
341 	default:
342 		zone = HAMMER_ZONE_UNAVAIL_INDEX;
343 		break;
344 	}
345 	collect_blockmap(elm->leaf.data_offset,
346 		(elm->leaf.data_len + 15) & ~15, zone);
347 }
348 
349 static __inline
350 void
351 collect_undo(hammer_off_t scan_offset, hammer_fifo_head_t head)
352 {
353 	collect_blockmap(scan_offset, head->hdr_size,
354 		HAMMER_ZONE_UNDO_INDEX);
355 }
356 
357 static
358 void
359 collect_blockmap(hammer_off_t offset, int32_t length, int zone)
360 {
361 	struct hammer_blockmap_layer1 layer1;
362 	struct hammer_blockmap_layer2 layer2;
363 	struct hammer_blockmap_layer2 *track2;
364 	hammer_off_t result_offset;
365 	collect_t collect;
366 	int error;
367 
368 	result_offset = blockmap_lookup(offset, &layer1, &layer2, &error);
369 	if (AssertOnFailure) {
370 		assert(HAMMER_ZONE_DECODE(offset) == zone);
371 		assert(HAMMER_ZONE_DECODE(result_offset) ==
372 			HAMMER_ZONE_RAW_BUFFER_INDEX);
373 		assert(error == 0);
374 	}
375 	collect = collect_get(layer1.phys_offset); /* layer2 address */
376 	track2 = collect_get_track(collect, offset, zone, &layer2);
377 	track2->bytes_free -= length;
378 }
379 
380 static
381 collect_t
382 collect_get(hammer_off_t phys_offset)
383 {
384 	collect_t collect;
385 
386 	collect = RB_LOOKUP(collect_rb_tree, &CollectTree, phys_offset);
387 	if (collect)
388 		return(collect);
389 
390 	collect = calloc(sizeof(*collect), 1);
391 	collect->track2 = malloc(HAMMER_BIGBLOCK_SIZE);  /* 1<<23 bytes */
392 	collect->layer2 = malloc(HAMMER_BIGBLOCK_SIZE);  /* 1<<23 bytes */
393 	collect->phys_offset = phys_offset;
394 	RB_INSERT(collect_rb_tree, &CollectTree, collect);
395 	bzero(collect->track2, HAMMER_BIGBLOCK_SIZE);
396 	bzero(collect->layer2, HAMMER_BIGBLOCK_SIZE);
397 
398 	return (collect);
399 }
400 
401 static
402 void
403 collect_rel(collect_t collect)
404 {
405 	free(collect->layer2);
406 	free(collect->track2);
407 	free(collect);
408 }
409 
410 static
411 struct hammer_blockmap_layer2 *
412 collect_get_track(collect_t collect, hammer_off_t offset, int zone,
413 		  struct hammer_blockmap_layer2 *layer2)
414 {
415 	struct hammer_blockmap_layer2 *track2;
416 	size_t i;
417 
418 	i = HAMMER_BLOCKMAP_LAYER2_OFFSET(offset) / sizeof(*track2);
419 	track2 = &collect->track2[i];
420 	if (track2->entry_crc == 0) {
421 		collect->layer2[i] = *layer2;
422 		track2->zone = zone;
423 		track2->bytes_free = HAMMER_BIGBLOCK_SIZE;
424 		track2->entry_crc = 1;	/* steal field to tag track load */
425 	}
426 	return (track2);
427 }
428 
429 static
430 void
431 dump_collect_table(void)
432 {
433 	collect_t collect;
434 	int i;
435 	int error = 0;
436 	int total = 0;
437 	int stats[HAMMER_MAX_ZONES];
438 	bzero(stats, sizeof(stats));
439 
440 	RB_FOREACH(collect, collect_rb_tree, &CollectTree) {
441 		dump_collect(collect, stats);
442 		error += collect->error;
443 	}
444 
445 	while ((collect = RB_ROOT(&CollectTree)) != NULL) {
446 		RB_REMOVE(collect_rb_tree, &CollectTree, collect);
447 		collect_rel(collect);
448 	}
449 	assert(RB_EMPTY(&CollectTree));
450 
451 	if (VerboseOpt) {
452 		printf("Statistics\n");
453 		printf("\tzone #\tbig-blocks\n");
454 		for (i = 0; i < HAMMER_MAX_ZONES; i++) {
455 			printf("\tzone %d\t%d\n", i, stats[i]);
456 			total += stats[i];
457 		}
458 		printf("\t---------------\n");
459 		printf("\ttotal\t%d\n", total);
460 	}
461 
462 	if (error || VerboseOpt)
463 		printf("%d errors\n", error);
464 }
465 
466 static
467 void
468 dump_collect(collect_t collect, int *stats)
469 {
470 	struct hammer_blockmap_layer2 *track2;
471 	struct hammer_blockmap_layer2 *layer2;
472 	hammer_off_t offset;
473 	size_t i;
474 	int zone;
475 
476 	for (i = 0; i < HAMMER_BLOCKMAP_RADIX2; ++i) {
477 		track2 = &collect->track2[i];
478 		layer2 = &collect->layer2[i];
479 		offset = collect->phys_offset + i * HAMMER_BIGBLOCK_SIZE;
480 
481 		/*
482 		 * Check big-blocks referenced by data, B-Tree nodes
483 		 * and UNDO fifo.
484 		 */
485 		if (track2->entry_crc == 0)
486 			continue;
487 
488 		zone = layer2->zone;
489 		if (AssertOnFailure) {
490 			assert((zone == HAMMER_ZONE_UNDO_INDEX) ||
491 				(zone >= HAMMER_ZONE2_MAPPED_INDEX &&
492 				 zone < HAMMER_MAX_ZONES));
493 		}
494 		stats[zone]++;
495 
496 		if (track2->zone != layer2->zone) {
497 			printf("BZ\tblock=%016jx calc zone=%2d, got zone=%2d\n",
498 				(intmax_t)offset,
499 				track2->zone,
500 				layer2->zone);
501 			collect->error++;
502 		} else if (track2->bytes_free != layer2->bytes_free) {
503 			printf("BM\tblock=%016jx zone=%2d calc %d free, got %d\n",
504 				(intmax_t)offset,
505 				layer2->zone,
506 				track2->bytes_free,
507 				layer2->bytes_free);
508 			collect->error++;
509 		} else if (VerboseOpt) {
510 			printf("\tblock=%016jx zone=%2d %d free (correct)\n",
511 				(intmax_t)offset,
512 				layer2->zone,
513 				track2->bytes_free);
514 		}
515 	}
516 }
517