xref: /dragonfly/sys/vfs/hammer/hammer_volume.c (revision 61c0377f)
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 + 1 >= 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_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
189 	mp->mnt_vstat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
190 	    (HAMMER_LARGEBLOCK_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 	 *
276 	 */
277 
278 	hmp->volume_to_remove = volume->vol_no;
279 
280 	struct hammer_ioc_reblock reblock;
281 	bzero(&reblock, sizeof(reblock));
282 
283 	reblock.key_beg.localization = HAMMER_MIN_LOCALIZATION;
284 	reblock.key_beg.obj_id = HAMMER_MIN_OBJID;
285 	reblock.key_end.localization = HAMMER_MAX_LOCALIZATION;
286 	reblock.key_end.obj_id = HAMMER_MAX_OBJID;
287 	reblock.head.flags = HAMMER_IOC_DO_FLAGS;
288 	reblock.free_level = 0;
289 
290 	error = hammer_ioc_reblock(trans, ip, &reblock);
291 
292 	if (reblock.head.flags & HAMMER_IOC_HEAD_INTR) {
293 		error = EINTR;
294 	}
295 
296 	if (error) {
297 		if (error == EINTR) {
298 			kprintf("reblock was interrupted\n");
299 		} else {
300 			kprintf("reblock failed: %d\n", error);
301 		}
302 		hmp->volume_to_remove = -1;
303 		hammer_rel_volume(volume, 0);
304 		goto end;
305 	}
306 
307 	/*
308 	 * Sync filesystem
309 	 */
310 	int count = 0;
311 	while (hammer_flusher_haswork(hmp)) {
312 		hammer_flusher_sync(hmp);
313 		++count;
314 		if (count >= 5) {
315 			if (count == 5)
316 				kprintf("HAMMER: flushing.");
317 			else
318 				kprintf(".");
319 			tsleep(&count, 0, "hmrufl", hz);
320 		}
321 		if (count == 30) {
322 			kprintf("giving up");
323 			break;
324 		}
325 	}
326 	kprintf("\n");
327 
328 	hammer_sync_lock_sh(trans);
329 	hammer_lock_ex(&hmp->blkmap_lock);
330 
331 	/*
332 	 * We use stat later to update rootvol's bigblock stats
333 	 */
334 	struct bigblock_stat stat;
335 	error = hammer_free_freemap(trans, volume, &stat);
336 	if (error) {
337 		kprintf("Failed to free volume. Volume not empty!\n");
338 		hmp->volume_to_remove = -1;
339 		hammer_rel_volume(volume, 0);
340 		hammer_unlock(&hmp->blkmap_lock);
341 		hammer_sync_unlock(trans);
342 		goto end;
343 	}
344 
345 	hmp->volume_to_remove = -1;
346 
347 	hammer_rel_volume(volume, 0);
348 
349 	/*
350 	 * Unload buffers
351 	 */
352         RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
353 		hammer_unload_buffer, volume);
354 
355 	error = hammer_unload_volume(volume, NULL);
356 	if (error == -1) {
357 		kprintf("Failed to unload volume\n");
358 		hammer_unlock(&hmp->blkmap_lock);
359 		hammer_sync_unlock(trans);
360 		goto end;
361 	}
362 
363 	volume = NULL;
364 	--hmp->nvolumes;
365 
366 	/*
367 	 * Set each volume's new value of the vol_count field.
368 	 */
369 	for (int vol_no = 0; vol_no < HAMMER_MAX_VOLUMES; ++vol_no) {
370 		volume = hammer_get_volume(hmp, vol_no, &error);
371 		if (volume == NULL && error == ENOENT) {
372 			/*
373 			 * Skip unused volume numbers
374 			 */
375 			error = 0;
376 			continue;
377 		}
378 
379 		KKASSERT(volume != NULL && error == 0);
380 		hammer_modify_volume_field(trans, volume, vol_count);
381 		volume->ondisk->vol_count = hmp->nvolumes;
382 		hammer_modify_volume_done(volume);
383 
384 		/*
385 		 * Only changes to the header of the root volume
386 		 * are automatically flushed to disk. For all
387 		 * other volumes that we modify we do it here.
388 		 *
389 		 * No interlock is needed, volume buffers are not
390 		 * messed with by bioops.
391 		 */
392 		if (volume != trans->rootvol && volume->io.modified) {
393 			hammer_crc_set_volume(volume->ondisk);
394 			hammer_io_flush(&volume->io, 0);
395 		}
396 
397 		hammer_rel_volume(volume, 0);
398 	}
399 
400 	/*
401 	 * Update the total number of bigblocks
402 	 */
403 	hammer_modify_volume_field(trans, trans->rootvol,
404 		vol0_stat_bigblocks);
405 	trans->rootvol->ondisk->vol0_stat_bigblocks -= stat.total_bigblocks;
406 	hammer_modify_volume_done(trans->rootvol);
407 
408 	/*
409 	 * Update the number of free bigblocks
410 	 * (including the copy in hmp)
411 	 */
412 	hammer_modify_volume_field(trans, trans->rootvol,
413 		vol0_stat_freebigblocks);
414 	trans->rootvol->ondisk->vol0_stat_freebigblocks -= stat.total_free_bigblocks;
415 	hmp->copy_stat_freebigblocks =
416 		trans->rootvol->ondisk->vol0_stat_freebigblocks;
417 	hammer_modify_volume_done(trans->rootvol);
418 	/*
419 	 * Bigblock count changed so recompute the total number of blocks.
420 	 */
421 	mp->mnt_stat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
422 	    (HAMMER_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
423 	mp->mnt_vstat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
424 	    (HAMMER_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
425 
426 	hammer_unlock(&hmp->blkmap_lock);
427 	hammer_sync_unlock(trans);
428 
429 	/*
430 	 * Erase the volume header of the removed device.
431 	 *
432 	 * This is to not accidentally mount the volume again.
433 	 */
434 	struct vnode *devvp = NULL;
435 	error = hammer_setup_device(&devvp, ioc->device_name, 0);
436 	if (error) {
437 		kprintf("Failed to open device: %s\n", ioc->device_name);
438 		goto end;
439 	}
440 	KKASSERT(devvp);
441 	error = hammer_clear_volume_header(devvp);
442 	if (error) {
443 		kprintf("Failed to clear volume header of device: %s\n",
444 			ioc->device_name);
445 		goto end;
446 	}
447 	hammer_close_device(&devvp, 0);
448 
449 	KKASSERT(error == 0);
450 end:
451 	hammer_unlock(&hmp->volume_lock);
452 	return (error);
453 }
454 
455 
456 int
457 hammer_ioc_volume_list(hammer_transaction_t trans, hammer_inode_t ip,
458     struct hammer_ioc_volume_list *ioc)
459 {
460 	struct hammer_mount *hmp = trans->hmp;
461 	hammer_volume_t volume;
462 	int error = 0;
463 	int i, cnt, len;
464 
465 	for (i = 0, cnt = 0; i < HAMMER_MAX_VOLUMES && cnt < ioc->nvols; i++) {
466 		volume = hammer_get_volume(hmp, i, &error);
467 		if (volume == NULL && error == ENOENT) {
468 			error = 0;
469 			continue;
470 		}
471 		KKASSERT(volume != NULL && error == 0);
472 
473 		len = strlen(volume->vol_name) + 1;
474 		KKASSERT(len <= MAXPATHLEN);
475 
476 		error = copyout(volume->vol_name, ioc->vols[cnt].device_name,
477 				len);
478 		if (error) {
479 			hammer_rel_volume(volume, 0);
480 			return (error);
481 		}
482 		cnt++;
483 		hammer_rel_volume(volume, 0);
484 	}
485 	ioc->nvols = cnt;
486 
487 	return (error);
488 }
489 
490 /*
491  * Iterate over all usable L1 entries of the volume and
492  * the corresponding L2 entries.
493  */
494 static int
495 hammer_iterate_l1l2_entries(hammer_transaction_t trans, hammer_volume_t volume,
496 	int (*callback)(hammer_transaction_t, hammer_volume_t, hammer_buffer_t*,
497 		struct hammer_blockmap_layer1*, struct hammer_blockmap_layer2*,
498 		hammer_off_t, hammer_off_t, void*),
499 	void *data)
500 {
501 	struct hammer_mount *hmp = trans->hmp;
502 	hammer_blockmap_t freemap = &hmp->blockmap[HAMMER_ZONE_FREEMAP_INDEX];
503 	hammer_buffer_t buffer = NULL;
504 	int error = 0;
505 
506 	hammer_off_t phys_off;
507 	hammer_off_t block_off;
508 	hammer_off_t layer1_off;
509 	hammer_off_t layer2_off;
510 	hammer_off_t aligned_buf_end_off;
511 	struct hammer_blockmap_layer1 *layer1;
512 	struct hammer_blockmap_layer2 *layer2;
513 
514 	/*
515 	 * Calculate the usable size of the volume, which
516 	 * must be aligned at a bigblock (8 MB) boundary.
517 	 */
518 	aligned_buf_end_off = (HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no,
519 		(volume->ondisk->vol_buf_end - volume->ondisk->vol_buf_beg)
520 		& ~HAMMER_LARGEBLOCK_MASK64));
521 
522 	/*
523 	 * Iterate the volume's address space in chunks of 4 TB, where each
524 	 * chunk consists of at least one physically available 8 MB bigblock.
525 	 *
526 	 * For each chunk we need one L1 entry and one L2 bigblock.
527 	 * We use the first bigblock of each chunk as L2 block.
528 	 */
529 	for (phys_off = HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no, 0);
530 	     phys_off < aligned_buf_end_off;
531 	     phys_off += HAMMER_BLOCKMAP_LAYER2) {
532 		for (block_off = 0;
533 		     block_off < HAMMER_BLOCKMAP_LAYER2;
534 		     block_off += HAMMER_LARGEBLOCK_SIZE) {
535 			layer2_off = phys_off +
536 				HAMMER_BLOCKMAP_LAYER2_OFFSET(block_off);
537 			layer2 = hammer_bread(hmp, layer2_off, &error, &buffer);
538 			if (error)
539 				goto end;
540 
541 			error = callback(trans, volume, &buffer, NULL,
542 					 layer2, phys_off, block_off, data);
543 			if (error)
544 				goto end;
545 		}
546 
547 		layer1_off = freemap->phys_offset +
548 				HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_off);
549 		layer1 = hammer_bread(hmp, layer1_off, &error, &buffer);
550 		if (error)
551 			goto end;
552 
553 		error = callback(trans, volume, &buffer, layer1, NULL,
554 				 phys_off, 0, data);
555 		if (error)
556 			goto end;
557 	}
558 
559 end:
560 	if (buffer) {
561 		hammer_rel_buffer(buffer, 0);
562 		buffer = NULL;
563 	}
564 
565 	return error;
566 }
567 
568 
569 static int
570 format_callback(hammer_transaction_t trans, hammer_volume_t volume,
571 	hammer_buffer_t *bufferp,
572 	struct hammer_blockmap_layer1 *layer1,
573 	struct hammer_blockmap_layer2 *layer2,
574 	hammer_off_t phys_off,
575 	hammer_off_t block_off,
576 	void *data)
577 {
578 	struct bigblock_stat *stat = (struct bigblock_stat*)data;
579 
580 	/*
581 	 * Calculate the usable size of the volume, which must be aligned
582 	 * at a bigblock (8 MB) boundary.
583 	 */
584 	hammer_off_t aligned_buf_end_off;
585 	aligned_buf_end_off = (HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no,
586 		(volume->ondisk->vol_buf_end - volume->ondisk->vol_buf_beg)
587 		& ~HAMMER_LARGEBLOCK_MASK64));
588 
589 	if (layer1) {
590 		KKASSERT(layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL);
591 
592 		hammer_modify_buffer(trans, *bufferp, layer1, sizeof(*layer1));
593 		bzero(layer1, sizeof(*layer1));
594 		layer1->phys_offset = phys_off;
595 		layer1->blocks_free = stat->counter;
596 		layer1->layer1_crc = crc32(layer1, HAMMER_LAYER1_CRCSIZE);
597 		hammer_modify_buffer_done(*bufferp);
598 
599 		stat->total_free_bigblocks += stat->counter;
600 		stat->counter = 0; /* reset */
601 	} else if (layer2) {
602 		hammer_modify_buffer(trans, *bufferp, layer2, sizeof(*layer2));
603 		bzero(layer2, sizeof(*layer2));
604 
605 		if (block_off == 0) {
606 			/*
607 			 * The first entry represents the L2 bigblock itself.
608 			 */
609 			layer2->zone = HAMMER_ZONE_FREEMAP_INDEX;
610 			layer2->append_off = HAMMER_LARGEBLOCK_SIZE;
611 			layer2->bytes_free = 0;
612 			++stat->total_bigblocks;
613 		} else if (phys_off + block_off < aligned_buf_end_off) {
614 			/*
615 			 * Available bigblock
616 			 */
617 			layer2->zone = 0;
618 			layer2->append_off = 0;
619 			layer2->bytes_free = HAMMER_LARGEBLOCK_SIZE;
620 			++stat->total_bigblocks;
621 			++stat->counter;
622 		} else {
623 			/*
624 			 * Bigblock outside of physically available
625 			 * space
626 			 */
627 			layer2->zone = HAMMER_ZONE_UNAVAIL_INDEX;
628 			layer2->append_off = HAMMER_LARGEBLOCK_SIZE;
629 			layer2->bytes_free = 0;
630 		}
631 
632 		layer2->entry_crc = crc32(layer2, HAMMER_LAYER2_CRCSIZE);
633 		hammer_modify_buffer_done(*bufferp);
634 	} else {
635 		KKASSERT(0);
636 	}
637 
638 	return 0;
639 }
640 
641 static int
642 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume,
643 	struct bigblock_stat *stat)
644 {
645 	stat->total_bigblocks = 0;
646 	stat->total_free_bigblocks = 0;
647 	stat->counter = 0;
648 	return hammer_iterate_l1l2_entries(trans, volume, format_callback, stat);
649 }
650 
651 static int
652 free_callback(hammer_transaction_t trans, hammer_volume_t volume __unused,
653 	hammer_buffer_t *bufferp,
654 	struct hammer_blockmap_layer1 *layer1,
655 	struct hammer_blockmap_layer2 *layer2,
656 	hammer_off_t phys_off,
657 	hammer_off_t block_off __unused,
658 	void *data)
659 {
660 	struct bigblock_stat *stat = (struct bigblock_stat*)data;
661 
662 	/*
663 	 * No modifications to ondisk structures
664 	 */
665 	int testonly = (stat == NULL);
666 
667 	if (layer1) {
668 		if (layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL) {
669 			/*
670 			 * This layer1 entry is already free.
671 			 */
672 			return 0;
673 		}
674 
675 		KKASSERT((int)HAMMER_VOL_DECODE(layer1->phys_offset) ==
676 			trans->hmp->volume_to_remove);
677 
678 		if (testonly)
679 			return 0;
680 
681 		/*
682 		 * Free the L1 entry
683 		 */
684 		hammer_modify_buffer(trans, *bufferp, layer1, sizeof(*layer1));
685 		bzero(layer1, sizeof(*layer1));
686 		layer1->phys_offset = HAMMER_BLOCKMAP_UNAVAIL;
687 		layer1->layer1_crc = crc32(layer1, HAMMER_LAYER1_CRCSIZE);
688 		hammer_modify_buffer_done(*bufferp);
689 
690 		return 0;
691 	} else if (layer2) {
692 		if (layer2->zone == HAMMER_ZONE_UNAVAIL_INDEX) {
693 			return 0;
694 		}
695 
696 		if (layer2->zone == HAMMER_ZONE_FREEMAP_INDEX) {
697 			if (stat) {
698 				++stat->total_bigblocks;
699 			}
700 			return 0;
701 		}
702 
703 		if (layer2->append_off == 0 &&
704 		    layer2->bytes_free == HAMMER_LARGEBLOCK_SIZE) {
705 			if (stat) {
706 				++stat->total_bigblocks;
707 				++stat->total_free_bigblocks;
708 			}
709 			return 0;
710 		}
711 
712 		/*
713 		 * We found a layer2 entry that is not empty!
714 		 */
715 		return EBUSY;
716 	} else {
717 		KKASSERT(0);
718 	}
719 
720 	return EINVAL;
721 }
722 
723 static int
724 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume,
725 	struct bigblock_stat *stat)
726 {
727 	int error;
728 
729 	stat->total_bigblocks = 0;
730 	stat->total_free_bigblocks = 0;
731 	stat->counter = 0;
732 
733 	error = hammer_iterate_l1l2_entries(trans, volume, free_callback, NULL);
734 	if (error)
735 		return error;
736 
737 	error = hammer_iterate_l1l2_entries(trans, volume, free_callback, stat);
738 	return error;
739 }
740 
741 /************************************************************************
742  *				MISC					*
743  ************************************************************************
744  */
745 
746 static int
747 hammer_setup_device(struct vnode **devvpp, const char *dev_path, int ronly)
748 {
749 	int error;
750 	struct nlookupdata nd;
751 
752 	/*
753 	 * Get the device vnode
754 	 */
755 	if (*devvpp == NULL) {
756 		error = nlookup_init(&nd, dev_path, UIO_SYSSPACE, NLC_FOLLOW);
757 		if (error == 0)
758 			error = nlookup(&nd);
759 		if (error == 0)
760 			error = cache_vref(&nd.nl_nch, nd.nl_cred, devvpp);
761 		nlookup_done(&nd);
762 	} else {
763 		error = 0;
764 	}
765 
766 	if (error == 0) {
767 		if (vn_isdisk(*devvpp, &error)) {
768 			error = vfs_mountedon(*devvpp);
769 		}
770 	}
771 	if (error == 0 && vcount(*devvpp) > 0)
772 		error = EBUSY;
773 	if (error == 0) {
774 		vn_lock(*devvpp, LK_EXCLUSIVE | LK_RETRY);
775 		error = vinvalbuf(*devvpp, V_SAVE, 0, 0);
776 		if (error == 0) {
777 			error = VOP_OPEN(*devvpp,
778 					 (ronly ? FREAD : FREAD|FWRITE),
779 					 FSCRED, NULL);
780 		}
781 		vn_unlock(*devvpp);
782 	}
783 	if (error && *devvpp) {
784 		vrele(*devvpp);
785 		*devvpp = NULL;
786 	}
787 	return (error);
788 }
789 
790 static void
791 hammer_close_device(struct vnode **devvpp, int ronly)
792 {
793 	if (*devvpp) {
794 		vn_lock(*devvpp, LK_EXCLUSIVE | LK_RETRY);
795 		vinvalbuf(*devvpp, ronly ? 0 : V_SAVE, 0, 0);
796 		VOP_CLOSE(*devvpp, (ronly ? FREAD : FREAD|FWRITE), NULL);
797 		vn_unlock(*devvpp);
798 		vrele(*devvpp);
799 		*devvpp = NULL;
800 	}
801 }
802 
803 static int
804 hammer_format_volume_header(struct hammer_mount *hmp, struct vnode *devvp,
805 	const char *vol_name, int vol_no, int vol_count,
806 	int64_t vol_size, int64_t boot_area_size, int64_t mem_area_size)
807 {
808 	struct buf *bp = NULL;
809 	struct hammer_volume_ondisk *ondisk;
810 	int error;
811 
812 	/*
813 	 * Extract the volume number from the volume header and do various
814 	 * sanity checks.
815 	 */
816 	KKASSERT(HAMMER_BUFSIZE >= sizeof(struct hammer_volume_ondisk));
817 	error = bread(devvp, 0LL, HAMMER_BUFSIZE, &bp);
818 	if (error || bp->b_bcount < sizeof(struct hammer_volume_ondisk))
819 		goto late_failure;
820 
821 	ondisk = (struct hammer_volume_ondisk*) bp->b_data;
822 
823 	/*
824 	 * Note that we do NOT allow to use a device that contains
825 	 * a valid HAMMER signature. It has to be cleaned up with dd
826 	 * before.
827 	 */
828 	if (ondisk->vol_signature == HAMMER_FSBUF_VOLUME) {
829 		kprintf("hammer_volume_add: Formatting of valid HAMMER volume "
830 			"%s denied. Erase with dd!\n", vol_name);
831 		error = EFTYPE;
832 		goto late_failure;
833 	}
834 
835 	bzero(ondisk, sizeof(struct hammer_volume_ondisk));
836 	ksnprintf(ondisk->vol_name, sizeof(ondisk->vol_name), "%s", vol_name);
837 	ondisk->vol_fstype = hmp->rootvol->ondisk->vol_fstype;
838 	ondisk->vol_signature = HAMMER_FSBUF_VOLUME;
839 	ondisk->vol_fsid = hmp->fsid;
840 	ondisk->vol_rootvol = hmp->rootvol->vol_no;
841 	ondisk->vol_no = vol_no;
842 	ondisk->vol_count = vol_count;
843 	ondisk->vol_version = hmp->version;
844 
845 	/*
846 	 * Reserve space for (future) header junk, setup our poor-man's
847 	 * bigblock allocator.
848 	 */
849 	int64_t vol_alloc = HAMMER_BUFSIZE * 16;
850 
851 	ondisk->vol_bot_beg = vol_alloc;
852 	vol_alloc += boot_area_size;
853 	ondisk->vol_mem_beg = vol_alloc;
854 	vol_alloc += mem_area_size;
855 
856 	/*
857 	 * The remaining area is the zone 2 buffer allocation area.  These
858 	 * buffers
859 	 */
860 	ondisk->vol_buf_beg = vol_alloc;
861 	ondisk->vol_buf_end = vol_size & ~(int64_t)HAMMER_BUFMASK;
862 
863 	if (ondisk->vol_buf_end < ondisk->vol_buf_beg) {
864 		kprintf("volume %d %s is too small to hold the volume header",
865 		     ondisk->vol_no, ondisk->vol_name);
866 		error = EFTYPE;
867 		goto late_failure;
868 	}
869 
870 	ondisk->vol_nblocks = (ondisk->vol_buf_end - ondisk->vol_buf_beg) /
871 			      HAMMER_BUFSIZE;
872 	ondisk->vol_blocksize = HAMMER_BUFSIZE;
873 
874 	/*
875 	 * Write volume header to disk
876 	 */
877 	error = bwrite(bp);
878 	bp = NULL;
879 
880 late_failure:
881 	if (bp)
882 		brelse(bp);
883 	return (error);
884 }
885 
886 /*
887  * Invalidates the volume header. Used by volume-del.
888  */
889 static int
890 hammer_clear_volume_header(struct vnode *devvp)
891 {
892 	struct buf *bp = NULL;
893 	struct hammer_volume_ondisk *ondisk;
894 	int error;
895 
896 	KKASSERT(HAMMER_BUFSIZE >= sizeof(struct hammer_volume_ondisk));
897 	error = bread(devvp, 0LL, HAMMER_BUFSIZE, &bp);
898 	if (error || bp->b_bcount < sizeof(struct hammer_volume_ondisk))
899 		goto late_failure;
900 
901 	ondisk = (struct hammer_volume_ondisk*) bp->b_data;
902 	bzero(ondisk, sizeof(struct hammer_volume_ondisk));
903 
904 	error = bwrite(bp);
905 	bp = NULL;
906 
907 late_failure:
908 	if (bp)
909 		brelse(bp);
910 	return (error);
911 }
912