xref: /dragonfly/sys/vfs/hammer/hammer_volume.c (revision 896f2e3a)
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 #include <sys/fcntl.h>
39 #include <sys/nlookup.h>
40 #include <sys/buf.h>
41 
42 #include <sys/buf2.h>
43 
44 static int
45 hammer_setup_device(struct vnode **devvpp, const char *dev_path, int ronly);
46 
47 static void
48 hammer_close_device(struct vnode **devvpp, int ronly);
49 
50 static int
51 hammer_format_volume_header(struct hammer_mount *hmp, struct vnode *devvp,
52 	const char *vol_name, int vol_no, int vol_count,
53 	int64_t vol_size, int64_t boot_area_size, int64_t mem_area_size);
54 
55 static int
56 hammer_clear_volume_header(struct vnode *devvp);
57 
58 struct bigblock_stat {
59 	uint64_t total_bigblocks;
60 	uint64_t total_free_bigblocks;
61 	uint64_t counter;
62 };
63 
64 static int
65 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume,
66 	struct bigblock_stat *stat);
67 
68 static int
69 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume,
70 	struct bigblock_stat *stat);
71 
72 int
73 hammer_ioc_volume_add(hammer_transaction_t trans, hammer_inode_t ip,
74 		struct hammer_ioc_volume *ioc)
75 {
76 	struct hammer_mount *hmp = trans->hmp;
77 	struct mount *mp = hmp->mp;
78 	hammer_volume_t volume;
79 	int error;
80 
81 	if (mp->mnt_flag & MNT_RDONLY) {
82 		kprintf("Cannot add volume to read-only HAMMER filesystem\n");
83 		return (EINVAL);
84 	}
85 
86 	if (hmp->nvolumes >= HAMMER_MAX_VOLUMES) {
87 		kprintf("Max number of HAMMER volumes exceeded\n");
88 		return (EINVAL);
89 	}
90 
91 	if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
92 		kprintf("Another volume operation is in progress!\n");
93 		return (EAGAIN);
94 	}
95 
96 	/*
97 	 * Find an unused volume number.
98 	 */
99 	int free_vol_no = 0;
100 	while (free_vol_no < HAMMER_MAX_VOLUMES &&
101 	       RB_LOOKUP(hammer_vol_rb_tree, &hmp->rb_vols_root, free_vol_no)) {
102 		++free_vol_no;
103 	}
104 	if (free_vol_no >= HAMMER_MAX_VOLUMES) {
105 		kprintf("Max number of HAMMER volumes exceeded\n");
106 		hammer_unlock(&hmp->volume_lock);
107 		return (EINVAL);
108 	}
109 
110 	struct vnode *devvp = NULL;
111 	error = hammer_setup_device(&devvp, ioc->device_name, 0);
112 	if (error)
113 		goto end;
114 	KKASSERT(devvp);
115 	error = hammer_format_volume_header(
116 		hmp,
117 		devvp,
118 		hmp->rootvol->ondisk->vol_name,
119 		free_vol_no,
120 		hmp->nvolumes+1,
121 		ioc->vol_size,
122 		ioc->boot_area_size,
123 		ioc->mem_area_size);
124 	hammer_close_device(&devvp, 0);
125 	if (error)
126 		goto end;
127 
128 	error = hammer_install_volume(hmp, ioc->device_name, NULL);
129 	if (error)
130 		goto end;
131 
132 	hammer_sync_lock_sh(trans);
133 	hammer_lock_ex(&hmp->blkmap_lock);
134 
135 	++hmp->nvolumes;
136 
137 	/*
138 	 * Set each volumes new value of the vol_count field.
139 	 */
140 	for (int vol_no = 0; vol_no < HAMMER_MAX_VOLUMES; ++vol_no) {
141 		volume = hammer_get_volume(hmp, vol_no, &error);
142 		if (volume == NULL && error == ENOENT) {
143 			/*
144 			 * Skip unused volume numbers
145 			 */
146 			error = 0;
147 			continue;
148 		}
149 		KKASSERT(volume != NULL && error == 0);
150 		hammer_modify_volume_field(trans, volume, vol_count);
151 		volume->ondisk->vol_count = hmp->nvolumes;
152 		hammer_modify_volume_done(volume);
153 
154 		/*
155 		 * Only changes to the header of the root volume
156 		 * are automatically flushed to disk. For all
157 		 * other volumes that we modify we do it here.
158 		 *
159 		 * No interlock is needed, volume buffers are not
160 		 * messed with by bioops.
161 		 */
162 		if (volume != trans->rootvol && volume->io.modified) {
163 			hammer_crc_set_volume(volume->ondisk);
164 			hammer_io_flush(&volume->io, 0);
165 		}
166 
167 		hammer_rel_volume(volume, 0);
168 	}
169 
170 	volume = hammer_get_volume(hmp, free_vol_no, &error);
171 	KKASSERT(volume != NULL && error == 0);
172 
173 	struct bigblock_stat stat;
174 	error =	hammer_format_freemap(trans, volume, &stat);
175 	KKASSERT(error == 0);
176 
177 	/*
178 	 * Increase the total number of bigblocks and update stat/vstat totals.
179 	 */
180 	hammer_modify_volume_field(trans, trans->rootvol,
181 		vol0_stat_bigblocks);
182 	trans->rootvol->ondisk->vol0_stat_bigblocks += stat.total_bigblocks;
183 	hammer_modify_volume_done(trans->rootvol);
184 	/*
185 	 * Bigblock count changed so recompute the total number of blocks.
186 	 */
187 	mp->mnt_stat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
188 	    (HAMMER_BIGBLOCK_SIZE / HAMMER_BUFSIZE);
189 	mp->mnt_vstat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
190 	    (HAMMER_BIGBLOCK_SIZE / HAMMER_BUFSIZE);
191 
192 	/*
193 	 * Increase the number of free bigblocks
194 	 * (including the copy in hmp)
195 	 */
196 	hammer_modify_volume_field(trans, trans->rootvol,
197 		vol0_stat_freebigblocks);
198 	trans->rootvol->ondisk->vol0_stat_freebigblocks += stat.total_free_bigblocks;
199 	hmp->copy_stat_freebigblocks =
200 		trans->rootvol->ondisk->vol0_stat_freebigblocks;
201 	hammer_modify_volume_done(trans->rootvol);
202 
203 	hammer_rel_volume(volume, 0);
204 
205 	hammer_unlock(&hmp->blkmap_lock);
206 	hammer_sync_unlock(trans);
207 
208 	KKASSERT(error == 0);
209 end:
210 	hammer_unlock(&hmp->volume_lock);
211 	if (error)
212 		kprintf("An error occurred: %d\n", error);
213 	return (error);
214 }
215 
216 
217 /*
218  * Remove a volume.
219  */
220 int
221 hammer_ioc_volume_del(hammer_transaction_t trans, hammer_inode_t ip,
222 		struct hammer_ioc_volume *ioc)
223 {
224 	struct hammer_mount *hmp = trans->hmp;
225 	struct mount *mp = hmp->mp;
226 	hammer_volume_t volume;
227 	int error = 0;
228 
229 	if (mp->mnt_flag & MNT_RDONLY) {
230 		kprintf("Cannot del volume from read-only HAMMER filesystem\n");
231 		return (EINVAL);
232 	}
233 
234 	if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
235 		kprintf("Another volume operation is in progress!\n");
236 		return (EAGAIN);
237 	}
238 
239 	volume = NULL;
240 
241 	/*
242 	 * find volume by volname
243 	 */
244 	for (int vol_no = 0; vol_no < HAMMER_MAX_VOLUMES; ++vol_no) {
245 		volume = hammer_get_volume(hmp, vol_no, &error);
246 		if (volume == NULL && error == ENOENT) {
247 			/*
248 			 * Skip unused volume numbers
249 			 */
250 			error = 0;
251 			continue;
252 		}
253 		KKASSERT(volume != NULL && error == 0);
254 		if (strcmp(volume->vol_name, ioc->device_name) == 0) {
255 			break;
256 		}
257 		hammer_rel_volume(volume, 0);
258 		volume = NULL;
259 	}
260 
261 	if (volume == NULL) {
262 		kprintf("Couldn't find volume\n");
263 		error = EINVAL;
264 		goto end;
265 	}
266 
267 	if (volume == trans->rootvol) {
268 		kprintf("Cannot remove root-volume\n");
269 		hammer_rel_volume(volume, 0);
270 		error = EINVAL;
271 		goto end;
272 	}
273 
274 	/*
275 	 * Reblock filesystem
276 	 */
277 	hmp->volume_to_remove = volume->vol_no;
278 
279 	struct hammer_ioc_reblock reblock;
280 	bzero(&reblock, sizeof(reblock));
281 
282 	reblock.key_beg.localization = HAMMER_MIN_LOCALIZATION;
283 	reblock.key_beg.obj_id = HAMMER_MIN_OBJID;
284 	reblock.key_end.localization = HAMMER_MAX_LOCALIZATION;
285 	reblock.key_end.obj_id = HAMMER_MAX_OBJID;
286 	reblock.head.flags = HAMMER_IOC_DO_FLAGS;
287 	reblock.free_level = 0;
288 
289 	error = hammer_ioc_reblock(trans, ip, &reblock);
290 
291 	if (reblock.head.flags & HAMMER_IOC_HEAD_INTR) {
292 		error = EINTR;
293 	}
294 
295 	if (error) {
296 		if (error == EINTR) {
297 			kprintf("reblock was interrupted\n");
298 		} else {
299 			kprintf("reblock failed: %d\n", error);
300 		}
301 		hmp->volume_to_remove = -1;
302 		hammer_rel_volume(volume, 0);
303 		goto end;
304 	}
305 
306 	/*
307 	 * Sync filesystem
308 	 */
309 	int count = 0;
310 	while (hammer_flusher_haswork(hmp)) {
311 		hammer_flusher_sync(hmp);
312 		++count;
313 		if (count >= 5) {
314 			if (count == 5)
315 				kprintf("HAMMER: flushing.");
316 			else
317 				kprintf(".");
318 			tsleep(&count, 0, "hmrufl", hz);
319 		}
320 		if (count == 30) {
321 			kprintf("giving up");
322 			break;
323 		}
324 	}
325 	kprintf("\n");
326 
327 	hammer_sync_lock_sh(trans);
328 	hammer_lock_ex(&hmp->blkmap_lock);
329 
330 	/*
331 	 * We use stat later to update rootvol's bigblock stats
332 	 */
333 	struct bigblock_stat stat;
334 	error = hammer_free_freemap(trans, volume, &stat);
335 	if (error) {
336 		kprintf("Failed to free volume. Volume not empty!\n");
337 		hmp->volume_to_remove = -1;
338 		hammer_rel_volume(volume, 0);
339 		hammer_unlock(&hmp->blkmap_lock);
340 		hammer_sync_unlock(trans);
341 		goto end;
342 	}
343 
344 	hmp->volume_to_remove = -1;
345 
346 	hammer_rel_volume(volume, 0);
347 
348 	/*
349 	 * Unload buffers
350 	 */
351         RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
352 		hammer_unload_buffer, volume);
353 
354 	error = hammer_unload_volume(volume, NULL);
355 	if (error == -1) {
356 		kprintf("Failed to unload volume\n");
357 		hammer_unlock(&hmp->blkmap_lock);
358 		hammer_sync_unlock(trans);
359 		goto end;
360 	}
361 
362 	volume = NULL;
363 	--hmp->nvolumes;
364 
365 	/*
366 	 * Set each volume's new value of the vol_count field.
367 	 */
368 	for (int vol_no = 0; vol_no < HAMMER_MAX_VOLUMES; ++vol_no) {
369 		volume = hammer_get_volume(hmp, vol_no, &error);
370 		if (volume == NULL && error == ENOENT) {
371 			/*
372 			 * Skip unused volume numbers
373 			 */
374 			error = 0;
375 			continue;
376 		}
377 
378 		KKASSERT(volume != NULL && error == 0);
379 		hammer_modify_volume_field(trans, volume, vol_count);
380 		volume->ondisk->vol_count = hmp->nvolumes;
381 		hammer_modify_volume_done(volume);
382 
383 		/*
384 		 * Only changes to the header of the root volume
385 		 * are automatically flushed to disk. For all
386 		 * other volumes that we modify we do it here.
387 		 *
388 		 * No interlock is needed, volume buffers are not
389 		 * messed with by bioops.
390 		 */
391 		if (volume != trans->rootvol && volume->io.modified) {
392 			hammer_crc_set_volume(volume->ondisk);
393 			hammer_io_flush(&volume->io, 0);
394 		}
395 
396 		hammer_rel_volume(volume, 0);
397 	}
398 
399 	/*
400 	 * Update the total number of bigblocks
401 	 */
402 	hammer_modify_volume_field(trans, trans->rootvol,
403 		vol0_stat_bigblocks);
404 	trans->rootvol->ondisk->vol0_stat_bigblocks -= stat.total_bigblocks;
405 	hammer_modify_volume_done(trans->rootvol);
406 
407 	/*
408 	 * Update the number of free bigblocks
409 	 * (including the copy in hmp)
410 	 */
411 	hammer_modify_volume_field(trans, trans->rootvol,
412 		vol0_stat_freebigblocks);
413 	trans->rootvol->ondisk->vol0_stat_freebigblocks -= stat.total_free_bigblocks;
414 	hmp->copy_stat_freebigblocks =
415 		trans->rootvol->ondisk->vol0_stat_freebigblocks;
416 	hammer_modify_volume_done(trans->rootvol);
417 	/*
418 	 * Bigblock count changed so recompute the total number of blocks.
419 	 */
420 	mp->mnt_stat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
421 	    (HAMMER_BIGBLOCK_SIZE / HAMMER_BUFSIZE);
422 	mp->mnt_vstat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
423 	    (HAMMER_BIGBLOCK_SIZE / HAMMER_BUFSIZE);
424 
425 	hammer_unlock(&hmp->blkmap_lock);
426 	hammer_sync_unlock(trans);
427 
428 	/*
429 	 * Erase the volume header of the removed device.
430 	 *
431 	 * This is to not accidentally mount the volume again.
432 	 */
433 	struct vnode *devvp = NULL;
434 	error = hammer_setup_device(&devvp, ioc->device_name, 0);
435 	if (error) {
436 		kprintf("Failed to open device: %s\n", ioc->device_name);
437 		goto end;
438 	}
439 	KKASSERT(devvp);
440 	error = hammer_clear_volume_header(devvp);
441 	if (error) {
442 		kprintf("Failed to clear volume header of device: %s\n",
443 			ioc->device_name);
444 		goto end;
445 	}
446 	hammer_close_device(&devvp, 0);
447 
448 	KKASSERT(error == 0);
449 end:
450 	hammer_unlock(&hmp->volume_lock);
451 	return (error);
452 }
453 
454 
455 int
456 hammer_ioc_volume_list(hammer_transaction_t trans, hammer_inode_t ip,
457     struct hammer_ioc_volume_list *ioc)
458 {
459 	struct hammer_mount *hmp = trans->hmp;
460 	hammer_volume_t volume;
461 	int error = 0;
462 	int i, cnt, len;
463 
464 	for (i = 0, cnt = 0; i < HAMMER_MAX_VOLUMES && cnt < ioc->nvols; i++) {
465 		volume = hammer_get_volume(hmp, i, &error);
466 		if (volume == NULL && error == ENOENT) {
467 			error = 0;
468 			continue;
469 		}
470 		KKASSERT(volume != NULL && error == 0);
471 
472 		len = strlen(volume->vol_name) + 1;
473 		KKASSERT(len <= MAXPATHLEN);
474 
475 		error = copyout(volume->vol_name, ioc->vols[cnt].device_name,
476 				len);
477 		if (error) {
478 			hammer_rel_volume(volume, 0);
479 			return (error);
480 		}
481 		cnt++;
482 		hammer_rel_volume(volume, 0);
483 	}
484 	ioc->nvols = cnt;
485 
486 	return (error);
487 }
488 
489 /*
490  * Iterate over all usable L1 entries of the volume and
491  * the corresponding L2 entries.
492  */
493 static int
494 hammer_iterate_l1l2_entries(hammer_transaction_t trans, hammer_volume_t volume,
495 	int (*callback)(hammer_transaction_t, hammer_volume_t, hammer_buffer_t*,
496 		struct hammer_blockmap_layer1*, struct hammer_blockmap_layer2*,
497 		hammer_off_t, hammer_off_t, void*),
498 	void *data)
499 {
500 	struct hammer_mount *hmp = trans->hmp;
501 	hammer_blockmap_t freemap = &hmp->blockmap[HAMMER_ZONE_FREEMAP_INDEX];
502 	hammer_buffer_t buffer = NULL;
503 	int error = 0;
504 
505 	hammer_off_t phys_off;
506 	hammer_off_t block_off;
507 	hammer_off_t layer1_off;
508 	hammer_off_t layer2_off;
509 	hammer_off_t aligned_buf_end_off;
510 	struct hammer_blockmap_layer1 *layer1;
511 	struct hammer_blockmap_layer2 *layer2;
512 
513 	/*
514 	 * Calculate the usable size of the volume, which
515 	 * must be aligned at a bigblock (8 MB) boundary.
516 	 */
517 	aligned_buf_end_off = (HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no,
518 		(volume->ondisk->vol_buf_end - volume->ondisk->vol_buf_beg)
519 		& ~HAMMER_BIGBLOCK_MASK64));
520 
521 	/*
522 	 * Iterate the volume's address space in chunks of 4 TB, where each
523 	 * chunk consists of at least one physically available 8 MB bigblock.
524 	 *
525 	 * For each chunk we need one L1 entry and one L2 bigblock.
526 	 * We use the first bigblock of each chunk as L2 block.
527 	 */
528 	for (phys_off = HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no, 0);
529 	     phys_off < aligned_buf_end_off;
530 	     phys_off += HAMMER_BLOCKMAP_LAYER2) {
531 		for (block_off = 0;
532 		     block_off < HAMMER_BLOCKMAP_LAYER2;
533 		     block_off += HAMMER_BIGBLOCK_SIZE) {
534 			layer2_off = phys_off +
535 				HAMMER_BLOCKMAP_LAYER2_OFFSET(block_off);
536 			layer2 = hammer_bread(hmp, layer2_off, &error, &buffer);
537 			if (error)
538 				goto end;
539 
540 			error = callback(trans, volume, &buffer, NULL,
541 					 layer2, phys_off, block_off, data);
542 			if (error)
543 				goto end;
544 		}
545 
546 		layer1_off = freemap->phys_offset +
547 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_off);
548 		layer1 = hammer_bread(hmp, layer1_off, &error, &buffer);
549 		if (error)
550 			goto end;
551 
552 		error = callback(trans, volume, &buffer, layer1, NULL,
553 				 phys_off, 0, data);
554 		if (error)
555 			goto end;
556 	}
557 
558 end:
559 	if (buffer) {
560 		hammer_rel_buffer(buffer, 0);
561 		buffer = NULL;
562 	}
563 
564 	return error;
565 }
566 
567 
568 static int
569 format_callback(hammer_transaction_t trans, hammer_volume_t volume,
570 	hammer_buffer_t *bufferp,
571 	struct hammer_blockmap_layer1 *layer1,
572 	struct hammer_blockmap_layer2 *layer2,
573 	hammer_off_t phys_off,
574 	hammer_off_t block_off,
575 	void *data)
576 {
577 	struct bigblock_stat *stat = (struct bigblock_stat*)data;
578 
579 	/*
580 	 * Calculate the usable size of the volume, which must be aligned
581 	 * at a bigblock (8 MB) boundary.
582 	 */
583 	hammer_off_t aligned_buf_end_off;
584 	aligned_buf_end_off = (HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no,
585 		(volume->ondisk->vol_buf_end - volume->ondisk->vol_buf_beg)
586 		& ~HAMMER_BIGBLOCK_MASK64));
587 
588 	if (layer1) {
589 		KKASSERT(layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL);
590 
591 		hammer_modify_buffer(trans, *bufferp, layer1, sizeof(*layer1));
592 		bzero(layer1, sizeof(*layer1));
593 		layer1->phys_offset = phys_off;
594 		layer1->blocks_free = stat->counter;
595 		layer1->layer1_crc = crc32(layer1, HAMMER_LAYER1_CRCSIZE);
596 		hammer_modify_buffer_done(*bufferp);
597 
598 		stat->total_free_bigblocks += stat->counter;
599 		stat->counter = 0; /* reset */
600 	} else if (layer2) {
601 		hammer_modify_buffer(trans, *bufferp, layer2, sizeof(*layer2));
602 		bzero(layer2, sizeof(*layer2));
603 
604 		if (block_off == 0) {
605 			/*
606 			 * The first entry represents the L2 bigblock itself.
607 			 */
608 			layer2->zone = HAMMER_ZONE_FREEMAP_INDEX;
609 			layer2->append_off = HAMMER_BIGBLOCK_SIZE;
610 			layer2->bytes_free = 0;
611 			++stat->total_bigblocks;
612 		} else if (phys_off + block_off < aligned_buf_end_off) {
613 			/*
614 			 * Available bigblock
615 			 */
616 			layer2->zone = 0;
617 			layer2->append_off = 0;
618 			layer2->bytes_free = HAMMER_BIGBLOCK_SIZE;
619 			++stat->total_bigblocks;
620 			++stat->counter;
621 		} else {
622 			/*
623 			 * Bigblock outside of physically available
624 			 * space
625 			 */
626 			layer2->zone = HAMMER_ZONE_UNAVAIL_INDEX;
627 			layer2->append_off = HAMMER_BIGBLOCK_SIZE;
628 			layer2->bytes_free = 0;
629 		}
630 
631 		layer2->entry_crc = crc32(layer2, HAMMER_LAYER2_CRCSIZE);
632 		hammer_modify_buffer_done(*bufferp);
633 	} else {
634 		KKASSERT(0);
635 	}
636 
637 	return 0;
638 }
639 
640 static int
641 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume,
642 	struct bigblock_stat *stat)
643 {
644 	stat->total_bigblocks = 0;
645 	stat->total_free_bigblocks = 0;
646 	stat->counter = 0;
647 	return hammer_iterate_l1l2_entries(trans, volume, format_callback, stat);
648 }
649 
650 static int
651 free_callback(hammer_transaction_t trans, hammer_volume_t volume __unused,
652 	hammer_buffer_t *bufferp,
653 	struct hammer_blockmap_layer1 *layer1,
654 	struct hammer_blockmap_layer2 *layer2,
655 	hammer_off_t phys_off,
656 	hammer_off_t block_off __unused,
657 	void *data)
658 {
659 	struct bigblock_stat *stat = (struct bigblock_stat*)data;
660 
661 	/*
662 	 * No modifications to ondisk structures
663 	 */
664 	int testonly = (stat == NULL);
665 
666 	if (layer1) {
667 		if (layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL) {
668 			/*
669 			 * This layer1 entry is already free.
670 			 */
671 			return 0;
672 		}
673 
674 		KKASSERT((int)HAMMER_VOL_DECODE(layer1->phys_offset) ==
675 			trans->hmp->volume_to_remove);
676 
677 		if (testonly)
678 			return 0;
679 
680 		/*
681 		 * Free the L1 entry
682 		 */
683 		hammer_modify_buffer(trans, *bufferp, layer1, sizeof(*layer1));
684 		bzero(layer1, sizeof(*layer1));
685 		layer1->phys_offset = HAMMER_BLOCKMAP_UNAVAIL;
686 		layer1->layer1_crc = crc32(layer1, HAMMER_LAYER1_CRCSIZE);
687 		hammer_modify_buffer_done(*bufferp);
688 
689 		return 0;
690 	} else if (layer2) {
691 		if (layer2->zone == HAMMER_ZONE_UNAVAIL_INDEX) {
692 			return 0;
693 		}
694 
695 		if (layer2->zone == HAMMER_ZONE_FREEMAP_INDEX) {
696 			if (stat) {
697 				++stat->total_bigblocks;
698 			}
699 			return 0;
700 		}
701 
702 		if (layer2->append_off == 0 &&
703 		    layer2->bytes_free == HAMMER_BIGBLOCK_SIZE) {
704 			if (stat) {
705 				++stat->total_bigblocks;
706 				++stat->total_free_bigblocks;
707 			}
708 			return 0;
709 		}
710 
711 		/*
712 		 * We found a layer2 entry that is not empty!
713 		 */
714 		return EBUSY;
715 	} else {
716 		KKASSERT(0);
717 	}
718 
719 	return EINVAL;
720 }
721 
722 static int
723 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume,
724 	struct bigblock_stat *stat)
725 {
726 	int error;
727 
728 	stat->total_bigblocks = 0;
729 	stat->total_free_bigblocks = 0;
730 	stat->counter = 0;
731 
732 	error = hammer_iterate_l1l2_entries(trans, volume, free_callback, NULL);
733 	if (error)
734 		return error;
735 
736 	error = hammer_iterate_l1l2_entries(trans, volume, free_callback, stat);
737 	return error;
738 }
739 
740 /************************************************************************
741  *				MISC					*
742  ************************************************************************
743  */
744 
745 static int
746 hammer_setup_device(struct vnode **devvpp, const char *dev_path, int ronly)
747 {
748 	int error;
749 	struct nlookupdata nd;
750 
751 	/*
752 	 * Get the device vnode
753 	 */
754 	if (*devvpp == NULL) {
755 		error = nlookup_init(&nd, dev_path, UIO_SYSSPACE, NLC_FOLLOW);
756 		if (error == 0)
757 			error = nlookup(&nd);
758 		if (error == 0)
759 			error = cache_vref(&nd.nl_nch, nd.nl_cred, devvpp);
760 		nlookup_done(&nd);
761 	} else {
762 		error = 0;
763 	}
764 
765 	if (error == 0) {
766 		if (vn_isdisk(*devvpp, &error)) {
767 			error = vfs_mountedon(*devvpp);
768 		}
769 	}
770 	if (error == 0 && vcount(*devvpp) > 0)
771 		error = EBUSY;
772 	if (error == 0) {
773 		vn_lock(*devvpp, LK_EXCLUSIVE | LK_RETRY);
774 		error = vinvalbuf(*devvpp, V_SAVE, 0, 0);
775 		if (error == 0) {
776 			error = VOP_OPEN(*devvpp,
777 					 (ronly ? FREAD : FREAD|FWRITE),
778 					 FSCRED, NULL);
779 		}
780 		vn_unlock(*devvpp);
781 	}
782 	if (error && *devvpp) {
783 		vrele(*devvpp);
784 		*devvpp = NULL;
785 	}
786 	return (error);
787 }
788 
789 static void
790 hammer_close_device(struct vnode **devvpp, int ronly)
791 {
792 	if (*devvpp) {
793 		vn_lock(*devvpp, LK_EXCLUSIVE | LK_RETRY);
794 		vinvalbuf(*devvpp, ronly ? 0 : V_SAVE, 0, 0);
795 		VOP_CLOSE(*devvpp, (ronly ? FREAD : FREAD|FWRITE), NULL);
796 		vn_unlock(*devvpp);
797 		vrele(*devvpp);
798 		*devvpp = NULL;
799 	}
800 }
801 
802 static int
803 hammer_format_volume_header(struct hammer_mount *hmp, struct vnode *devvp,
804 	const char *vol_name, int vol_no, int vol_count,
805 	int64_t vol_size, int64_t boot_area_size, int64_t mem_area_size)
806 {
807 	struct buf *bp = NULL;
808 	struct hammer_volume_ondisk *ondisk;
809 	int error;
810 
811 	/*
812 	 * Extract the volume number from the volume header and do various
813 	 * sanity checks.
814 	 */
815 	KKASSERT(HAMMER_BUFSIZE >= sizeof(struct hammer_volume_ondisk));
816 	error = bread(devvp, 0LL, HAMMER_BUFSIZE, &bp);
817 	if (error || bp->b_bcount < sizeof(struct hammer_volume_ondisk))
818 		goto late_failure;
819 
820 	ondisk = (struct hammer_volume_ondisk*) bp->b_data;
821 
822 	/*
823 	 * Note that we do NOT allow to use a device that contains
824 	 * a valid HAMMER signature. It has to be cleaned up with dd
825 	 * before.
826 	 */
827 	if (ondisk->vol_signature == HAMMER_FSBUF_VOLUME) {
828 		kprintf("hammer_volume_add: Formatting of valid HAMMER volume "
829 			"%s denied. Erase with dd!\n", vol_name);
830 		error = EFTYPE;
831 		goto late_failure;
832 	}
833 
834 	bzero(ondisk, sizeof(struct hammer_volume_ondisk));
835 	ksnprintf(ondisk->vol_name, sizeof(ondisk->vol_name), "%s", vol_name);
836 	ondisk->vol_fstype = hmp->rootvol->ondisk->vol_fstype;
837 	ondisk->vol_signature = HAMMER_FSBUF_VOLUME;
838 	ondisk->vol_fsid = hmp->fsid;
839 	ondisk->vol_rootvol = hmp->rootvol->vol_no;
840 	ondisk->vol_no = vol_no;
841 	ondisk->vol_count = vol_count;
842 	ondisk->vol_version = hmp->version;
843 
844 	/*
845 	 * Reserve space for (future) header junk, setup our poor-man's
846 	 * bigblock allocator.
847 	 */
848 	int64_t vol_alloc = HAMMER_BUFSIZE * 16;
849 
850 	ondisk->vol_bot_beg = vol_alloc;
851 	vol_alloc += boot_area_size;
852 	ondisk->vol_mem_beg = vol_alloc;
853 	vol_alloc += mem_area_size;
854 
855 	/*
856 	 * The remaining area is the zone 2 buffer allocation area.  These
857 	 * buffers
858 	 */
859 	ondisk->vol_buf_beg = vol_alloc;
860 	ondisk->vol_buf_end = vol_size & ~(int64_t)HAMMER_BUFMASK;
861 
862 	if (ondisk->vol_buf_end < ondisk->vol_buf_beg) {
863 		kprintf("volume %d %s is too small to hold the volume header\n",
864 		     ondisk->vol_no, ondisk->vol_name);
865 		error = EFTYPE;
866 		goto late_failure;
867 	}
868 
869 	ondisk->vol_nblocks = (ondisk->vol_buf_end - ondisk->vol_buf_beg) /
870 			      HAMMER_BUFSIZE;
871 	ondisk->vol_blocksize = HAMMER_BUFSIZE;
872 
873 	/*
874 	 * Write volume header to disk
875 	 */
876 	error = bwrite(bp);
877 	bp = NULL;
878 
879 late_failure:
880 	if (bp)
881 		brelse(bp);
882 	return (error);
883 }
884 
885 /*
886  * Invalidates the volume header. Used by volume-del.
887  */
888 static int
889 hammer_clear_volume_header(struct vnode *devvp)
890 {
891 	struct buf *bp = NULL;
892 	struct hammer_volume_ondisk *ondisk;
893 	int error;
894 
895 	KKASSERT(HAMMER_BUFSIZE >= sizeof(struct hammer_volume_ondisk));
896 	error = bread(devvp, 0LL, HAMMER_BUFSIZE, &bp);
897 	if (error || bp->b_bcount < sizeof(struct hammer_volume_ondisk))
898 		goto late_failure;
899 
900 	ondisk = (struct hammer_volume_ondisk*) bp->b_data;
901 	bzero(ondisk, sizeof(struct hammer_volume_ondisk));
902 
903 	error = bwrite(bp);
904 	bp = NULL;
905 
906 late_failure:
907 	if (bp)
908 		brelse(bp);
909 	return (error);
910 }
911