xref: /dragonfly/sys/vfs/hammer/hammer_volume.c (revision 7d462af5)
1 /*
2  * Copyright (c) 2009 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> and
6  * Michael Neumann <mneumann@ntecs.de>
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in
16  *    the documentation and/or other materials provided with the
17  *    distribution.
18  * 3. Neither the name of The DragonFly Project nor the names of its
19  *    contributors may be used to endorse or promote products derived
20  *    from this software without specific, prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  */
36 
37 #include "hammer.h"
38 
39 static int
40 hammer_format_volume_header(hammer_mount_t hmp,
41 	struct hammer_ioc_volume *ioc,
42 	hammer_volume_ondisk_t ondisk,
43 	int vol_no);
44 
45 static int
46 hammer_update_volumes_header(hammer_transaction_t trans,
47 	int64_t total_bigblocks, int64_t empty_bigblocks);
48 
49 static int
50 hammer_do_reblock(hammer_transaction_t trans, hammer_inode_t ip);
51 
52 static int
53 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume);
54 
55 static int
56 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume);
57 
58 static int
59 hammer_count_bigblocks(hammer_mount_t hmp, hammer_volume_t volume,
60 	int64_t *total_bigblocks, int64_t *empty_bigblocks);
61 
62 int
63 hammer_ioc_volume_add(hammer_transaction_t trans, hammer_inode_t ip,
64 		struct hammer_ioc_volume *ioc)
65 {
66 	hammer_mount_t hmp = trans->hmp;
67 	struct mount *mp = hmp->mp;
68 	struct hammer_volume_ondisk ondisk;
69 	hammer_volume_t volume;
70 	int64_t total_bigblocks, empty_bigblocks;
71 	int free_vol_no = 0;
72 	int error;
73 
74 	if (mp->mnt_flag & MNT_RDONLY) {
75 		hmkprintf(hmp, "Cannot add volume to read-only HAMMER filesystem\n");
76 		return (EINVAL);
77 	}
78 
79 	if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
80 		hmkprintf(hmp, "Another volume operation is in progress!\n");
81 		return (EAGAIN);
82 	}
83 
84 	if (hmp->nvolumes >= HAMMER_MAX_VOLUMES) {
85 		hammer_unlock(&hmp->volume_lock);
86 		hmkprintf(hmp, "Max number of HAMMER volumes exceeded\n");
87 		return (EINVAL);
88 	}
89 
90 	/*
91 	 * Find an unused volume number.
92 	 */
93 	while (free_vol_no < HAMMER_MAX_VOLUMES &&
94 		hammer_volume_number_test(hmp, free_vol_no)) {
95 		++free_vol_no;
96 	}
97 	if (free_vol_no >= HAMMER_MAX_VOLUMES) {
98 		hmkprintf(hmp, "Max number of HAMMER volumes exceeded\n");
99 		error = EINVAL;
100 		goto end;
101 	}
102 
103 	error = hammer_format_volume_header(hmp, ioc, &ondisk, free_vol_no);
104 	if (error)
105 		goto end;
106 
107 	error = hammer_install_volume(hmp, ioc->device_name, NULL, &ondisk);
108 	if (error)
109 		goto end;
110 
111 	hammer_sync_lock_sh(trans);
112 	hammer_lock_ex(&hmp->blkmap_lock);
113 
114 	volume = hammer_get_volume(hmp, free_vol_no, &error);
115 	KKASSERT(volume != NULL && error == 0);
116 
117 	error =	hammer_format_freemap(trans, volume);
118 	KKASSERT(error == 0);
119 
120 	error = hammer_count_bigblocks(hmp, volume,
121 			&total_bigblocks, &empty_bigblocks);
122 	KKASSERT(error == 0);
123 	KKASSERT(total_bigblocks == empty_bigblocks);
124 
125 	hammer_rel_volume(volume, 0);
126 
127 	++hmp->nvolumes;
128 	error = hammer_update_volumes_header(trans,
129 			total_bigblocks, empty_bigblocks);
130 	KKASSERT(error == 0);
131 
132 	hammer_unlock(&hmp->blkmap_lock);
133 	hammer_sync_unlock(trans);
134 
135 	KKASSERT(error == 0);
136 end:
137 	hammer_unlock(&hmp->volume_lock);
138 	if (error)
139 		hmkprintf(hmp, "An error occurred: %d\n", error);
140 	return (error);
141 }
142 
143 
144 /*
145  * Remove a volume.
146  */
147 int
148 hammer_ioc_volume_del(hammer_transaction_t trans, hammer_inode_t ip,
149 		struct hammer_ioc_volume *ioc)
150 {
151 	hammer_mount_t hmp = trans->hmp;
152 	struct mount *mp = hmp->mp;
153 	struct hammer_volume_ondisk ondisk;
154 	hammer_volume_t volume;
155 	int64_t total_bigblocks, empty_bigblocks;
156 	int vol_no;
157 	int error = 0;
158 
159 	if (mp->mnt_flag & MNT_RDONLY) {
160 		hmkprintf(hmp, "Cannot del volume from read-only HAMMER filesystem\n");
161 		return (EINVAL);
162 	}
163 
164 	if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
165 		hmkprintf(hmp, "Another volume operation is in progress!\n");
166 		return (EAGAIN);
167 	}
168 
169 	if (hmp->nvolumes <= 1) {
170 		hammer_unlock(&hmp->volume_lock);
171 		hmkprintf(hmp, "No HAMMER volume to delete\n");
172 		return (EINVAL);
173 	}
174 
175 	/*
176 	 * find volume by volname
177 	 */
178 	volume = NULL;
179 	HAMMER_VOLUME_NUMBER_FOREACH(hmp, vol_no) {
180 		volume = hammer_get_volume(hmp, vol_no, &error);
181 		KKASSERT(volume != NULL && error == 0);
182 		if (strcmp(volume->vol_name, ioc->device_name) == 0) {
183 			break;
184 		}
185 		hammer_rel_volume(volume, 0);
186 		volume = NULL;
187 	}
188 
189 	if (volume == NULL) {
190 		hmkprintf(hmp, "Couldn't find volume\n");
191 		error = EINVAL;
192 		goto end;
193 	}
194 
195 	if (volume == trans->rootvol) {
196 		hmkprintf(hmp, "Cannot remove root-volume\n");
197 		hammer_rel_volume(volume, 0);
198 		error = EINVAL;
199 		goto end;
200 	}
201 
202 	/*
203 	 * Reblock filesystem if the volume is not empty
204 	 */
205 	hmp->volume_to_remove = volume->vol_no;
206 
207 	error = hammer_count_bigblocks(hmp, volume,
208 			&total_bigblocks, &empty_bigblocks);
209 	KKASSERT(error == 0);
210 
211 	if (total_bigblocks == empty_bigblocks) {
212 		hmkprintf(hmp, "%s is already empty\n", volume->vol_name);
213 	} else if (ioc->flag & HAMMER_IOC_VOLUME_REBLOCK) {
214 		error = hammer_do_reblock(trans, ip);
215 		if (error) {
216 			hmp->volume_to_remove = -1;
217 			hammer_rel_volume(volume, 0);
218 			goto end;
219 		}
220 	} else {
221 		hmkprintf(hmp, "%s is not empty\n", volume->vol_name);
222 		hammer_rel_volume(volume, 0);
223 		error = ENOTEMPTY;
224 		goto end;
225 	}
226 
227 	hammer_sync_lock_sh(trans);
228 	hammer_lock_ex(&hmp->blkmap_lock);
229 
230 	error = hammer_count_bigblocks(hmp, volume,
231 			&total_bigblocks, &empty_bigblocks);
232 	KKASSERT(error == 0);
233 
234 	error = hammer_free_freemap(trans, volume);
235 	if (error) {
236 		hmkprintf(hmp, "Failed to free volume: ");
237 		if (error == EBUSY)
238 			kprintf("Volume %d not empty\n", volume->vol_no);
239 		else
240 			kprintf("%d\n", error);
241 		hmp->volume_to_remove = -1;
242 		hammer_rel_volume(volume, 0);
243 		goto end1;
244 	}
245 	hammer_rel_volume(volume, 0);
246 
247 	/*
248 	 * XXX: Temporary solution for
249 	 * http://lists.dragonflybsd.org/pipermail/kernel/2015-August/175027.html
250 	 */
251 	hammer_unlock(&hmp->blkmap_lock);
252 	hammer_sync_unlock(trans);
253 	hammer_flusher_sync(hmp); /* 1 */
254 	hammer_flusher_sync(hmp); /* 2 */
255 	hammer_flusher_sync(hmp); /* 3 */
256 	hammer_sync_lock_sh(trans);
257 	hammer_lock_ex(&hmp->blkmap_lock);
258 
259 	/*
260 	 * Unload buffers
261 	 */
262         RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
263 		hammer_unload_buffer, volume);
264 
265 	bzero(&ondisk, sizeof(ondisk));
266 	error = hammer_unload_volume(volume, &ondisk);
267 	if (error == -1) {
268 		hmkprintf(hmp, "Failed to unload volume\n");
269 		goto end1;
270 	}
271 
272 	--hmp->nvolumes;
273 	error = hammer_update_volumes_header(trans,
274 			-total_bigblocks, -empty_bigblocks);
275 	KKASSERT(error == 0);
276 	hmp->volume_to_remove = -1;
277 
278 end1:
279 	hammer_unlock(&hmp->blkmap_lock);
280 	hammer_sync_unlock(trans);
281 
282 end:
283 	hammer_unlock(&hmp->volume_lock);
284 	if (error)
285 		hmkprintf(hmp, "An error occurred: %d\n", error);
286 	return (error);
287 }
288 
289 
290 int
291 hammer_ioc_volume_list(hammer_transaction_t trans, hammer_inode_t ip,
292     struct hammer_ioc_volume_list *ioc)
293 {
294 	hammer_mount_t hmp = trans->hmp;
295 	hammer_volume_t volume;
296 	int error = 0;
297 	int i, len, cnt = 0;
298 
299 	if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
300 		hmkprintf(hmp, "Another volume operation is in progress!\n");
301 		return (EAGAIN);
302 	}
303 
304 	HAMMER_VOLUME_NUMBER_FOREACH(hmp, i) {
305 		if (cnt >= ioc->nvols)
306 			break;
307 		volume = hammer_get_volume(hmp, i, &error);
308 		KKASSERT(volume != NULL && error == 0);
309 
310 		len = strlen(volume->vol_name) + 1;
311 		KKASSERT(len <= MAXPATHLEN);
312 
313 		ioc->vols[cnt].vol_no = volume->vol_no;
314 		error = copyout(volume->vol_name, ioc->vols[cnt].device_name,
315 				len);
316 		hammer_rel_volume(volume, 0);
317 		if (error)
318 			goto end;
319 		cnt++;
320 	}
321 	ioc->nvols = cnt;
322 
323 end:
324 	hammer_unlock(&hmp->volume_lock);
325 	return (error);
326 }
327 
328 static
329 int
330 hammer_do_reblock(hammer_transaction_t trans, hammer_inode_t ip)
331 {
332 	hammer_mount_t hmp = trans->hmp;
333 	int error;
334 	int vol_no;
335 
336 	struct hammer_ioc_reblock reblock;
337 	bzero(&reblock, sizeof(reblock));
338 
339 	vol_no = trans->hmp->volume_to_remove;
340 	KKASSERT(vol_no != -1);
341 
342 	reblock.key_beg.localization = HAMMER_MIN_LOCALIZATION;
343 	reblock.key_beg.obj_id = HAMMER_MIN_OBJID;
344 	reblock.key_end.localization = HAMMER_MAX_LOCALIZATION;
345 	reblock.key_end.obj_id = HAMMER_MAX_OBJID;
346 	reblock.head.flags = HAMMER_IOC_DO_FLAGS;
347 	reblock.free_level = 0;	/* reblock all big-blocks */
348 	reblock.allpfs = 1;	/* reblock all PFS */
349 	reblock.vol_no = vol_no;
350 
351 	hmkprintf(hmp, "reblock started\n");
352 	error = hammer_ioc_reblock(trans, ip, &reblock);
353 
354 	if (reblock.head.flags & HAMMER_IOC_HEAD_INTR) {
355 		error = EINTR;
356 	}
357 
358 	if (error) {
359 		if (error == EINTR) {
360 			hmkprintf(hmp, "reblock was interrupted\n");
361 		} else {
362 			hmkprintf(hmp, "reblock failed: %d\n", error);
363 		}
364 		return(error);
365 	}
366 
367 	return(0);
368 }
369 
370 /*
371  * XXX This somehow needs to stop doing hammer_modify_buffer() for
372  * layer2 entries.  In theory adding a large block device could
373  * blow away UNDO fifo.  The best way is to format layer2 entries
374  * in userspace without UNDO getting involved before the device is
375  * safely added to the filesystem.  HAMMER has no interest in what
376  * has happened to the device before it safely joins the filesystem.
377  */
378 static int
379 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume)
380 {
381 	hammer_mount_t hmp = trans->hmp;
382 	hammer_volume_ondisk_t ondisk;
383 	hammer_blockmap_t freemap;
384 	hammer_off_t alloc_offset;
385 	hammer_off_t phys_offset;
386 	hammer_off_t block_offset;
387 	hammer_off_t layer1_offset;
388 	hammer_off_t layer2_offset;
389 	hammer_off_t vol_free_end;
390 	hammer_off_t aligned_vol_free_end;
391 	hammer_blockmap_layer1_t layer1;
392 	hammer_blockmap_layer2_t layer2;
393 	hammer_buffer_t buffer1 = NULL;
394 	hammer_buffer_t buffer2 = NULL;
395 	int64_t vol_buf_size;
396 	int64_t layer1_count = 0;
397 	int error = 0;
398 
399 	KKASSERT(volume->vol_no != HAMMER_ROOT_VOLNO);
400 
401 	ondisk = volume->ondisk;
402 	vol_buf_size = HAMMER_VOL_BUF_SIZE(ondisk);
403 	KKASSERT((vol_buf_size & ~HAMMER_OFF_SHORT_MASK) == 0);
404 	vol_free_end = HAMMER_ENCODE_RAW_BUFFER(ondisk->vol_no,
405 			vol_buf_size & ~HAMMER_BIGBLOCK_MASK64);
406 	aligned_vol_free_end = HAMMER_BLOCKMAP_LAYER2_DOALIGN(vol_free_end);
407 
408 	freemap = &hmp->blockmap[HAMMER_ZONE_FREEMAP_INDEX];
409 	alloc_offset = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no, 0);
410 
411 	hmkprintf(hmp, "Initialize freemap volume %d\n", volume->vol_no);
412 
413 	for (phys_offset = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no, 0);
414 	     phys_offset < aligned_vol_free_end;
415 	     phys_offset += HAMMER_BLOCKMAP_LAYER2) {
416 		layer1_offset = freemap->phys_offset +
417 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_offset);
418 		layer1 = hammer_bread(hmp, layer1_offset, &error, &buffer1);
419 		if (error)
420 			goto end;
421 		if (layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL) {
422 			hammer_modify_buffer(trans, buffer1, layer1, sizeof(*layer1));
423 			bzero(layer1, sizeof(*layer1));
424 			layer1->phys_offset = alloc_offset;
425 			layer1->blocks_free = 0;
426 			hammer_crc_set_layer1(layer1);
427 			hammer_modify_buffer_done(buffer1);
428 			alloc_offset += HAMMER_BIGBLOCK_SIZE;
429 		}
430 	}
431 
432 	for (phys_offset = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no, 0);
433 	     phys_offset < aligned_vol_free_end;
434 	     phys_offset += HAMMER_BLOCKMAP_LAYER2) {
435 		layer1_count = 0;
436 		layer1_offset = freemap->phys_offset +
437 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_offset);
438 		layer1 = hammer_bread(hmp, layer1_offset, &error, &buffer1);
439 		if (error)
440 			goto end;
441 		KKASSERT(layer1->phys_offset != HAMMER_BLOCKMAP_UNAVAIL);
442 
443 		for (block_offset = 0;
444 		     block_offset < HAMMER_BLOCKMAP_LAYER2;
445 		     block_offset += HAMMER_BIGBLOCK_SIZE) {
446 			layer2_offset = layer1->phys_offset +
447 				        HAMMER_BLOCKMAP_LAYER2_OFFSET(block_offset);
448 			layer2 = hammer_bread(hmp, layer2_offset, &error, &buffer2);
449 			if (error)
450 				goto end;
451 
452 			hammer_modify_buffer(trans, buffer2, layer2, sizeof(*layer2));
453 			bzero(layer2, sizeof(*layer2));
454 
455 			if (phys_offset + block_offset < alloc_offset) {
456 				layer2->zone = HAMMER_ZONE_FREEMAP_INDEX;
457 				layer2->append_off = HAMMER_BIGBLOCK_SIZE;
458 				layer2->bytes_free = 0;
459 			} else if (phys_offset + block_offset < vol_free_end) {
460 				layer2->zone = 0;
461 				layer2->append_off = 0;
462 				layer2->bytes_free = HAMMER_BIGBLOCK_SIZE;
463 				++layer1_count;
464 			} else {
465 				layer2->zone = HAMMER_ZONE_UNAVAIL_INDEX;
466 				layer2->append_off = HAMMER_BIGBLOCK_SIZE;
467 				layer2->bytes_free = 0;
468 			}
469 
470 			hammer_crc_set_layer2(layer2);
471 			hammer_modify_buffer_done(buffer2);
472 		}
473 
474 		hammer_modify_buffer(trans, buffer1, layer1, sizeof(*layer1));
475 		layer1->blocks_free += layer1_count;
476 		hammer_crc_set_layer1(layer1);
477 		hammer_modify_buffer_done(buffer1);
478 	}
479 
480 end:
481 	if (buffer1)
482 		hammer_rel_buffer(buffer1, 0);
483 	if (buffer2)
484 		hammer_rel_buffer(buffer2, 0);
485 
486 	return error;
487 }
488 
489 /*
490  * XXX This somehow needs to stop doing hammer_modify_buffer() for
491  * layer2 entries.  In theory removing a large block device could
492  * blow away UNDO fifo.  The best way is to erase layer2 entries
493  * in userspace without UNDO getting involved after the device has
494  * been safely removed from the filesystem.  HAMMER has no interest
495  * in what happens to the device once it's safely removed.
496  */
497 static int
498 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume)
499 {
500 	hammer_mount_t hmp = trans->hmp;
501 	hammer_volume_ondisk_t ondisk;
502 	hammer_blockmap_t freemap;
503 	hammer_off_t phys_offset;
504 	hammer_off_t block_offset;
505 	hammer_off_t layer1_offset;
506 	hammer_off_t layer2_offset;
507 	hammer_off_t vol_free_end;
508 	hammer_off_t aligned_vol_free_end;
509 	hammer_blockmap_layer1_t layer1;
510 	hammer_blockmap_layer2_t layer2;
511 	hammer_buffer_t buffer1 = NULL;
512 	hammer_buffer_t buffer2 = NULL;
513 	int64_t vol_buf_size;
514 	int error = 0;
515 
516 	KKASSERT(volume->vol_no != HAMMER_ROOT_VOLNO);
517 
518 	ondisk = volume->ondisk;
519 	vol_buf_size = HAMMER_VOL_BUF_SIZE(ondisk);
520 	KKASSERT((vol_buf_size & ~HAMMER_OFF_SHORT_MASK) == 0);
521 	vol_free_end = HAMMER_ENCODE_RAW_BUFFER(ondisk->vol_no,
522 			vol_buf_size & ~HAMMER_BIGBLOCK_MASK64);
523 	aligned_vol_free_end = HAMMER_BLOCKMAP_LAYER2_DOALIGN(vol_free_end);
524 
525 	freemap = &hmp->blockmap[HAMMER_ZONE_FREEMAP_INDEX];
526 
527 	hmkprintf(hmp, "Free freemap volume %d\n", volume->vol_no);
528 
529 	for (phys_offset = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no, 0);
530 	     phys_offset < aligned_vol_free_end;
531 	     phys_offset += HAMMER_BLOCKMAP_LAYER2) {
532 		layer1_offset = freemap->phys_offset +
533 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_offset);
534 		layer1 = hammer_bread(hmp, layer1_offset, &error, &buffer1);
535 		if (error)
536 			goto end;
537 		KKASSERT(layer1->phys_offset != HAMMER_BLOCKMAP_UNAVAIL);
538 
539 		for (block_offset = 0;
540 		     block_offset < HAMMER_BLOCKMAP_LAYER2;
541 		     block_offset += HAMMER_BIGBLOCK_SIZE) {
542 			layer2_offset = layer1->phys_offset +
543 				        HAMMER_BLOCKMAP_LAYER2_OFFSET(block_offset);
544 			layer2 = hammer_bread(hmp, layer2_offset, &error, &buffer2);
545 			if (error)
546 				goto end;
547 
548 			switch (layer2->zone) {
549 			case HAMMER_ZONE_UNDO_INDEX:
550 				KKASSERT(0);
551 			case HAMMER_ZONE_FREEMAP_INDEX:
552 			case HAMMER_ZONE_UNAVAIL_INDEX:
553 				continue;
554 			default:
555 				KKASSERT(phys_offset + block_offset < aligned_vol_free_end);
556 				if (layer2->append_off == 0 &&
557 				    layer2->bytes_free == HAMMER_BIGBLOCK_SIZE)
558 					continue;
559 				break;
560 			}
561 			return EBUSY;  /* Not empty */
562 		}
563 	}
564 
565 	for (phys_offset = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no, 0);
566 	     phys_offset < aligned_vol_free_end;
567 	     phys_offset += HAMMER_BLOCKMAP_LAYER2) {
568 		layer1_offset = freemap->phys_offset +
569 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_offset);
570 		layer1 = hammer_bread(hmp, layer1_offset, &error, &buffer1);
571 		if (error)
572 			goto end;
573 		KKASSERT(layer1->phys_offset != HAMMER_BLOCKMAP_UNAVAIL);
574 
575 		for (block_offset = 0;
576 		     block_offset < HAMMER_BLOCKMAP_LAYER2;
577 		     block_offset += HAMMER_BIGBLOCK_SIZE) {
578 			layer2_offset = layer1->phys_offset +
579 				        HAMMER_BLOCKMAP_LAYER2_OFFSET(block_offset);
580 			layer2 = hammer_bread(hmp, layer2_offset, &error, &buffer2);
581 			if (error)
582 				goto end;
583 
584 			switch (layer2->zone) {
585 			case HAMMER_ZONE_UNDO_INDEX:
586 				KKASSERT(0);
587 			default:
588 				KKASSERT(phys_offset + block_offset < aligned_vol_free_end);
589 				hammer_modify_buffer(trans, buffer2, layer2, sizeof(*layer2));
590 				bzero(layer2, sizeof(*layer2));
591 				hammer_modify_buffer_done(buffer2);
592 				break;
593 			}
594 		}
595 
596 		hammer_modify_buffer(trans, buffer1, layer1, sizeof(*layer1));
597 		bzero(layer1, sizeof(*layer1));
598 		layer1->phys_offset = HAMMER_BLOCKMAP_UNAVAIL;
599 		hammer_crc_set_layer1(layer1);
600 		hammer_modify_buffer_done(buffer1);
601 	}
602 
603 end:
604 	if (buffer1)
605 		hammer_rel_buffer(buffer1, 0);
606 	if (buffer2)
607 		hammer_rel_buffer(buffer2, 0);
608 
609 	return error;
610 }
611 
612 static int
613 hammer_format_volume_header(hammer_mount_t hmp,
614 	struct hammer_ioc_volume *ioc,
615 	hammer_volume_ondisk_t ondisk,
616 	int vol_no)
617 {
618 	hammer_volume_ondisk_t root_ondisk;
619 	int64_t vol_alloc;
620 
621 	KKASSERT(HAMMER_BUFSIZE >= sizeof(struct hammer_volume_ondisk));
622 
623 	/*
624 	 * Just copy from the root volume header.
625 	 */
626 	root_ondisk = hmp->rootvol->ondisk;
627 	bzero(ondisk, sizeof(struct hammer_volume_ondisk));
628 	ondisk->vol_fsid = root_ondisk->vol_fsid;
629 	ondisk->vol_fstype = root_ondisk->vol_fstype;
630 	ksnprintf(ondisk->vol_label, sizeof(ondisk->vol_label), "%s",
631 		root_ondisk->vol_label);
632 	ondisk->vol_version = root_ondisk->vol_version;
633 	ondisk->vol_rootvol = root_ondisk->vol_no;
634 	ondisk->vol_signature = root_ondisk->vol_signature;
635 
636 	KKASSERT(ondisk->vol_rootvol == HAMMER_ROOT_VOLNO);
637 	KKASSERT(ondisk->vol_signature == HAMMER_FSBUF_VOLUME);
638 
639 	/*
640 	 * Assign the new vol_no and vol_count.
641 	 */
642 	ondisk->vol_no = vol_no;
643 	ondisk->vol_count = root_ondisk->vol_count + 1;
644 
645 	/*
646 	 * Reserve space for (future) header junk.
647 	 */
648 	vol_alloc = root_ondisk->vol_bot_beg;
649 	ondisk->vol_bot_beg = vol_alloc;
650 	vol_alloc += ioc->boot_area_size;
651 	ondisk->vol_mem_beg = vol_alloc;
652 	vol_alloc += ioc->memory_log_size;
653 
654 	/*
655 	 * The remaining area is the zone 2 buffer allocation area.
656 	 */
657 	ondisk->vol_buf_beg = vol_alloc;
658 	ondisk->vol_buf_end = ioc->vol_size & ~(int64_t)HAMMER_BUFMASK;
659 
660 	if (HAMMER_VOL_BUF_SIZE(ondisk) < 0) { /* int64_t */
661 		hmkprintf(hmp, "volume %d is too small to hold the volume header\n",
662 			ondisk->vol_no);
663 		return(EFTYPE);
664 	}
665 
666 	return(0);
667 }
668 
669 static int
670 hammer_update_volumes_header(hammer_transaction_t trans,
671 	int64_t total_bigblocks, int64_t empty_bigblocks)
672 {
673 	hammer_mount_t hmp = trans->hmp;
674 	struct mount *mp = hmp->mp;
675 	hammer_volume_t volume;
676 	int vol_no;
677 	int error = 0;
678 
679 	/*
680 	 * Set each volume's new value of the vol_count field.
681 	 */
682 	HAMMER_VOLUME_NUMBER_FOREACH(hmp, vol_no) {
683 		volume = hammer_get_volume(hmp, vol_no, &error);
684 		KKASSERT(volume != NULL && error == 0);
685 		hammer_modify_volume_field(trans, volume, vol_count);
686 		volume->ondisk->vol_count = hmp->nvolumes;
687 		hammer_modify_volume_done(volume);
688 
689 		/*
690 		 * Only changes to the header of the root volume
691 		 * are automatically flushed to disk. For all
692 		 * other volumes that we modify we do it here.
693 		 *
694 		 * No interlock is needed, volume buffers are not
695 		 * messed with by bioops.
696 		 */
697 		if (volume != trans->rootvol && volume->io.modified) {
698 			hammer_crc_set_volume(volume->ondisk);
699 			hammer_io_flush(&volume->io, 0);
700 		}
701 
702 		hammer_rel_volume(volume, 0);
703 	}
704 
705 	/*
706 	 * Update the total number of big-blocks.
707 	 */
708 	hammer_modify_volume_field(trans, trans->rootvol, vol0_stat_bigblocks);
709 	trans->rootvol->ondisk->vol0_stat_bigblocks += total_bigblocks;
710 	hammer_modify_volume_done(trans->rootvol);
711 
712 	/*
713 	 * Big-block count changed so recompute the total number of blocks.
714 	 */
715 	mp->mnt_stat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
716 				HAMMER_BUFFERS_PER_BIGBLOCK;
717 	mp->mnt_vstat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
718 				HAMMER_BUFFERS_PER_BIGBLOCK;
719 
720 	/*
721 	 * Update the total number of free big-blocks.
722 	 */
723 	hammer_modify_volume_field(trans, trans->rootvol,
724 		vol0_stat_freebigblocks);
725 	trans->rootvol->ondisk->vol0_stat_freebigblocks += empty_bigblocks;
726 	hammer_modify_volume_done(trans->rootvol);
727 
728 	/*
729 	 * Update the copy in hmp.
730 	 */
731 	hmp->copy_stat_freebigblocks =
732 		trans->rootvol->ondisk->vol0_stat_freebigblocks;
733 
734 	return(error);
735 }
736 
737 /*
738  * Count total big-blocks and empty big-blocks within the volume.
739  * The volume must be a non-root volume.
740  *
741  * Note that total big-blocks doesn't include big-blocks for layer2
742  * (and obviously layer1 and undomap).  This is requirement of the
743  * volume header and this function is to retrieve that information.
744  */
745 static int
746 hammer_count_bigblocks(hammer_mount_t hmp, hammer_volume_t volume,
747 	int64_t *total_bigblocks, int64_t *empty_bigblocks)
748 {
749 	hammer_volume_ondisk_t ondisk;
750 	hammer_blockmap_t freemap;
751 	hammer_off_t phys_offset;
752 	hammer_off_t block_offset;
753 	hammer_off_t layer1_offset;
754 	hammer_off_t layer2_offset;
755 	hammer_off_t vol_free_end;
756 	hammer_off_t aligned_vol_free_end;
757 	hammer_blockmap_layer1_t layer1;
758 	hammer_blockmap_layer2_t layer2;
759 	hammer_buffer_t buffer1 = NULL;
760 	hammer_buffer_t buffer2 = NULL;
761 	int64_t vol_buf_size;
762 	int64_t total = 0;
763 	int64_t empty = 0;
764 	int error = 0;
765 
766 	KKASSERT(volume->vol_no != HAMMER_ROOT_VOLNO);
767 
768 	*total_bigblocks = 0;	/* avoid gcc warnings */
769 	*empty_bigblocks = 0;	/* avoid gcc warnings */
770 
771 	ondisk = volume->ondisk;
772 	vol_buf_size = HAMMER_VOL_BUF_SIZE(ondisk);
773 	KKASSERT((vol_buf_size & ~HAMMER_OFF_SHORT_MASK) == 0);
774 	vol_free_end = HAMMER_ENCODE_RAW_BUFFER(ondisk->vol_no,
775 			vol_buf_size & ~HAMMER_BIGBLOCK_MASK64);
776 	aligned_vol_free_end = HAMMER_BLOCKMAP_LAYER2_DOALIGN(vol_free_end);
777 
778 	freemap = &hmp->blockmap[HAMMER_ZONE_FREEMAP_INDEX];
779 
780 	for (phys_offset = HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no, 0);
781 	     phys_offset < aligned_vol_free_end;
782 	     phys_offset += HAMMER_BLOCKMAP_LAYER2) {
783 		layer1_offset = freemap->phys_offset +
784 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_offset);
785 		layer1 = hammer_bread(hmp, layer1_offset, &error, &buffer1);
786 		if (error)
787 			goto end;
788 
789 		for (block_offset = 0;
790 		     block_offset < HAMMER_BLOCKMAP_LAYER2;
791 		     block_offset += HAMMER_BIGBLOCK_SIZE) {
792 			layer2_offset = layer1->phys_offset +
793 					HAMMER_BLOCKMAP_LAYER2_OFFSET(block_offset);
794 			layer2 = hammer_bread(hmp, layer2_offset, &error, &buffer2);
795 			if (error)
796 				goto end;
797 
798 			switch (layer2->zone) {
799 			case HAMMER_ZONE_UNDO_INDEX:
800 				KKASSERT(0);
801 			case HAMMER_ZONE_FREEMAP_INDEX:
802 			case HAMMER_ZONE_UNAVAIL_INDEX:
803 				continue;
804 			default:
805 				KKASSERT(phys_offset + block_offset < aligned_vol_free_end);
806 				total++;
807 				if (layer2->append_off == 0 &&
808 				    layer2->bytes_free == HAMMER_BIGBLOCK_SIZE)
809 					empty++;
810 				break;
811 			}
812 		}
813 	}
814 
815 	hmkprintf(hmp, "big-blocks total=%jd empty=%jd\n", total, empty);
816 	*total_bigblocks = total;
817 	*empty_bigblocks = empty;
818 end:
819 	if (buffer1)
820 		hammer_rel_buffer(buffer1, 0);
821 	if (buffer2)
822 		hammer_rel_buffer(buffer2, 0);
823 
824 	return error;
825 }
826