2 * Copyright (c) 2009 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com> and
6 * Michael Neumann <mneumann@ntecs.de>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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
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.
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
37 #include <sys/fcntl.h>
38 #include <sys/nlookup.h>
42 struct bigblock_stat {
43 int64_t total_bigblocks;
44 int64_t total_free_bigblocks;
49 hammer_format_volume_header(struct hammer_mount *hmp,
50 struct hammer_volume_ondisk *ondisk,
51 const char *vol_name, int vol_no, int vol_count,
52 int64_t vol_size, int64_t boot_area_size, int64_t mem_area_size);
55 hammer_update_volumes_header(hammer_transaction_t trans,
56 struct bigblock_stat *stat);
59 hammer_do_reblock(hammer_transaction_t trans, hammer_inode_t ip);
62 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume,
63 struct bigblock_stat *stat);
66 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume,
67 struct bigblock_stat *stat);
70 hammer_test_free_freemap(hammer_transaction_t trans, hammer_volume_t volume);
73 hammer_ioc_volume_add(hammer_transaction_t trans, hammer_inode_t ip,
74 struct hammer_ioc_volume *ioc)
76 struct hammer_mount *hmp = trans->hmp;
77 struct mount *mp = hmp->mp;
78 struct hammer_volume_ondisk ondisk;
79 struct bigblock_stat stat;
80 hammer_volume_t volume;
84 if (mp->mnt_flag & MNT_RDONLY) {
85 kprintf("Cannot add volume to read-only HAMMER filesystem\n");
89 if (hmp->nvolumes >= HAMMER_MAX_VOLUMES) {
90 kprintf("Max number of HAMMER volumes exceeded\n");
94 if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
95 kprintf("Another volume operation is in progress!\n");
100 * Find an unused volume number.
102 while (free_vol_no < HAMMER_MAX_VOLUMES &&
103 HAMMER_VOLUME_NUMBER_IS_SET(hmp, free_vol_no)) {
106 if (free_vol_no >= HAMMER_MAX_VOLUMES) {
107 kprintf("Max number of HAMMER volumes exceeded\n");
112 error = hammer_format_volume_header(
115 hmp->rootvol->ondisk->vol_name,
124 error = hammer_install_volume(hmp, ioc->device_name, NULL, &ondisk);
128 hammer_sync_lock_sh(trans);
129 hammer_lock_ex(&hmp->blkmap_lock);
131 volume = hammer_get_volume(hmp, free_vol_no, &error);
132 KKASSERT(volume != NULL && error == 0);
134 error = hammer_format_freemap(trans, volume, &stat);
135 KKASSERT(error == 0);
136 hammer_rel_volume(volume, 0);
139 error = hammer_update_volumes_header(trans, &stat);
140 KKASSERT(error == 0);
142 hammer_unlock(&hmp->blkmap_lock);
143 hammer_sync_unlock(trans);
145 KKASSERT(error == 0);
147 hammer_unlock(&hmp->volume_lock);
149 kprintf("An error occurred: %d\n", error);
158 hammer_ioc_volume_del(hammer_transaction_t trans, hammer_inode_t ip,
159 struct hammer_ioc_volume *ioc)
161 struct hammer_mount *hmp = trans->hmp;
162 struct mount *mp = hmp->mp;
163 struct hammer_volume_ondisk *ondisk;
164 struct bigblock_stat stat;
165 hammer_volume_t volume;
170 if (mp->mnt_flag & MNT_RDONLY) {
171 kprintf("Cannot del volume from read-only HAMMER filesystem\n");
175 if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
176 kprintf("Another volume operation is in progress!\n");
181 * find volume by volname
184 HAMMER_VOLUME_NUMBER_FOREACH(hmp, vol_no) {
185 volume = hammer_get_volume(hmp, vol_no, &error);
186 KKASSERT(volume != NULL && error == 0);
187 if (strcmp(volume->vol_name, ioc->device_name) == 0) {
190 hammer_rel_volume(volume, 0);
194 if (volume == NULL) {
195 kprintf("Couldn't find volume\n");
200 if (volume == trans->rootvol) {
201 kprintf("Cannot remove root-volume\n");
202 hammer_rel_volume(volume, 0);
208 * Reblock filesystem if the volume is not empty
210 hmp->volume_to_remove = volume->vol_no;
212 if (hammer_test_free_freemap(trans, volume)) {
213 error = hammer_do_reblock(trans, ip);
215 hmp->volume_to_remove = -1;
216 hammer_rel_volume(volume, 0);
224 while (hammer_flusher_haswork(hmp)) {
225 hammer_flusher_sync(hmp);
229 kprintf("HAMMER: flushing.");
232 tsleep(&count, 0, "hmrufl", hz);
235 kprintf("giving up");
242 hammer_sync_lock_sh(trans);
243 hammer_lock_ex(&hmp->blkmap_lock);
246 * We use stat later to update rootvol's big-block stats
248 error = hammer_free_freemap(trans, volume, &stat);
250 kprintf("Failed to free volume: ");
252 kprintf("Volume %d not empty\n", volume->vol_no);
254 kprintf("%d\n", error);
255 hmp->volume_to_remove = -1;
256 hammer_rel_volume(volume, 0);
260 hmp->volume_to_remove = -1;
261 hammer_rel_volume(volume, 0);
266 RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
267 hammer_unload_buffer, volume);
269 bzero(&ondisk, sizeof(ondisk));
270 error = hammer_unload_volume(volume, &ondisk);
272 kprintf("Failed to unload volume\n");
277 error = hammer_update_volumes_header(trans, &stat);
278 KKASSERT(error == 0);
281 hammer_unlock(&hmp->blkmap_lock);
282 hammer_sync_unlock(trans);
285 hammer_unlock(&hmp->volume_lock);
287 kprintf("An error occurred: %d\n", error);
293 hammer_ioc_volume_list(hammer_transaction_t trans, hammer_inode_t ip,
294 struct hammer_ioc_volume_list *ioc)
296 struct hammer_mount *hmp = trans->hmp;
297 hammer_volume_t volume;
301 if (hammer_lock_ex_try(&hmp->volume_lock) != 0) {
302 kprintf("Another volume operation is in progress!\n");
306 HAMMER_VOLUME_NUMBER_FOREACH(hmp, i) {
307 if (cnt >= ioc->nvols)
309 volume = hammer_get_volume(hmp, i, &error);
310 KKASSERT(volume != NULL && error == 0);
312 len = strlen(volume->vol_name) + 1;
313 KKASSERT(len <= MAXPATHLEN);
315 error = copyout(volume->vol_name, ioc->vols[cnt].device_name,
317 hammer_rel_volume(volume, 0);
325 hammer_unlock(&hmp->volume_lock);
331 hammer_do_reblock(hammer_transaction_t trans, hammer_inode_t ip)
335 struct hammer_ioc_reblock reblock;
336 bzero(&reblock, sizeof(reblock));
338 reblock.key_beg.localization = HAMMER_MIN_LOCALIZATION;
339 reblock.key_beg.obj_id = HAMMER_MIN_OBJID;
340 reblock.key_end.localization = HAMMER_MAX_LOCALIZATION;
341 reblock.key_end.obj_id = HAMMER_MAX_OBJID;
342 reblock.head.flags = HAMMER_IOC_DO_FLAGS;
343 reblock.free_level = 0;
346 kprintf("reblock started\n");
347 error = hammer_ioc_reblock(trans, ip, &reblock);
349 if (reblock.head.flags & HAMMER_IOC_HEAD_INTR) {
354 if (error == EINTR) {
355 kprintf("reblock was interrupted\n");
357 kprintf("reblock failed: %d\n", error);
366 * Iterate over all usable L1 entries of the volume and
367 * the corresponding L2 entries.
370 hammer_iterate_l1l2_entries(hammer_transaction_t trans, hammer_volume_t volume,
371 int (*callback)(hammer_transaction_t, hammer_volume_t, hammer_buffer_t*,
372 struct hammer_blockmap_layer1*, struct hammer_blockmap_layer2*,
373 hammer_off_t, hammer_off_t, void*),
376 struct hammer_mount *hmp = trans->hmp;
377 hammer_blockmap_t freemap = &hmp->blockmap[HAMMER_ZONE_FREEMAP_INDEX];
379 hammer_off_t phys_off;
380 hammer_off_t block_off;
381 hammer_off_t layer1_off;
382 hammer_off_t layer2_off;
383 hammer_off_t aligned_buf_end_off;
384 hammer_off_t aligned_vol_end_off;
385 struct hammer_blockmap_layer1 *layer1;
386 struct hammer_blockmap_layer2 *layer2;
387 hammer_buffer_t buffer1 = NULL;
388 hammer_buffer_t buffer2 = NULL;
391 * Calculate the usable size of the volume, which
392 * must be aligned at a big-block (8 MB) boundary.
394 aligned_buf_end_off = HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no,
395 (volume->ondisk->vol_buf_end - volume->ondisk->vol_buf_beg)
396 & ~HAMMER_BIGBLOCK_MASK64);
397 aligned_vol_end_off = (aligned_buf_end_off + HAMMER_BLOCKMAP_LAYER2_MASK)
398 & ~HAMMER_BLOCKMAP_LAYER2_MASK;
401 * Iterate the volume's address space in chunks of 4 TB, where each
402 * chunk consists of at least one physically available 8 MB big-block.
404 * For each chunk we need one L1 entry and one L2 big-block.
405 * We use the first big-block of each chunk as L2 block.
407 for (phys_off = HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no, 0);
408 phys_off < aligned_vol_end_off;
409 phys_off += HAMMER_BLOCKMAP_LAYER2) {
411 block_off < HAMMER_BLOCKMAP_LAYER2;
412 block_off += HAMMER_BIGBLOCK_SIZE) {
413 layer2_off = phys_off +
414 HAMMER_BLOCKMAP_LAYER2_OFFSET(block_off);
415 layer2 = hammer_bread(hmp, layer2_off, &error, &buffer2);
419 error = callback(trans, volume, &buffer2, NULL,
420 layer2, phys_off, block_off, data);
425 layer1_off = freemap->phys_offset +
426 HAMMER_BLOCKMAP_LAYER1_OFFSET(phys_off);
427 layer1 = hammer_bread(hmp, layer1_off, &error, &buffer1);
431 error = callback(trans, volume, &buffer1, layer1, NULL,
439 hammer_rel_buffer(buffer1, 0);
441 hammer_rel_buffer(buffer2, 0);
448 format_callback(hammer_transaction_t trans, hammer_volume_t volume,
449 hammer_buffer_t *bufferp,
450 struct hammer_blockmap_layer1 *layer1,
451 struct hammer_blockmap_layer2 *layer2,
452 hammer_off_t phys_off,
453 hammer_off_t block_off,
456 struct bigblock_stat *stat = (struct bigblock_stat*)data;
459 * Calculate the usable size of the volume, which must be aligned
460 * at a big-block (8 MB) boundary.
462 hammer_off_t aligned_buf_end_off;
463 aligned_buf_end_off = HAMMER_ENCODE_RAW_BUFFER(volume->ondisk->vol_no,
464 (volume->ondisk->vol_buf_end - volume->ondisk->vol_buf_beg)
465 & ~HAMMER_BIGBLOCK_MASK64);
468 KKASSERT(layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL);
470 hammer_modify_buffer(trans, *bufferp, layer1, sizeof(*layer1));
471 bzero(layer1, sizeof(*layer1));
472 layer1->phys_offset = phys_off;
473 layer1->blocks_free = stat->counter;
474 layer1->layer1_crc = crc32(layer1, HAMMER_LAYER1_CRCSIZE);
475 hammer_modify_buffer_done(*bufferp);
476 stat->counter = 0; /* reset */
478 hammer_modify_buffer(trans, *bufferp, layer2, sizeof(*layer2));
479 bzero(layer2, sizeof(*layer2));
481 if (block_off == 0) {
483 * The first entry represents the L2 big-block itself.
484 * Note that the first entry represents the L1 big-block
485 * and the second entry represents the L2 big-block for
486 * root volume, but this function assumes the volume is
487 * non-root given that we can't add a new root volume.
489 KKASSERT(trans->rootvol && trans->rootvol != volume);
490 layer2->zone = HAMMER_ZONE_FREEMAP_INDEX;
491 layer2->append_off = HAMMER_BIGBLOCK_SIZE;
492 layer2->bytes_free = 0;
493 } else if (phys_off + block_off < aligned_buf_end_off) {
495 * Available big-block
498 layer2->append_off = 0;
499 layer2->bytes_free = HAMMER_BIGBLOCK_SIZE;
500 ++stat->total_bigblocks;
501 ++stat->total_free_bigblocks;
505 * Big-block outside of physically available
508 layer2->zone = HAMMER_ZONE_UNAVAIL_INDEX;
509 layer2->append_off = HAMMER_BIGBLOCK_SIZE;
510 layer2->bytes_free = 0;
513 layer2->entry_crc = crc32(layer2, HAMMER_LAYER2_CRCSIZE);
514 hammer_modify_buffer_done(*bufferp);
523 hammer_format_freemap(hammer_transaction_t trans, hammer_volume_t volume,
524 struct bigblock_stat *stat)
526 stat->total_bigblocks = 0;
527 stat->total_free_bigblocks = 0;
529 return hammer_iterate_l1l2_entries(trans, volume, format_callback, stat);
533 free_callback(hammer_transaction_t trans, hammer_volume_t volume __unused,
534 hammer_buffer_t *bufferp,
535 struct hammer_blockmap_layer1 *layer1,
536 struct hammer_blockmap_layer2 *layer2,
537 hammer_off_t phys_off,
538 hammer_off_t block_off __unused,
541 struct bigblock_stat *stat = (struct bigblock_stat*)data;
544 if (layer1->phys_offset == HAMMER_BLOCKMAP_UNAVAIL) {
546 * This layer1 entry is already free.
551 KKASSERT((int)HAMMER_VOL_DECODE(layer1->phys_offset) ==
552 trans->hmp->volume_to_remove);
557 hammer_modify_buffer(trans, *bufferp, layer1, sizeof(*layer1));
558 bzero(layer1, sizeof(*layer1));
559 layer1->phys_offset = HAMMER_BLOCKMAP_UNAVAIL;
560 layer1->layer1_crc = crc32(layer1, HAMMER_LAYER1_CRCSIZE);
561 hammer_modify_buffer_done(*bufferp);
565 if (layer2->zone == HAMMER_ZONE_UNAVAIL_INDEX) {
569 if (layer2->zone == HAMMER_ZONE_FREEMAP_INDEX) {
573 if (layer2->append_off == 0 &&
574 layer2->bytes_free == HAMMER_BIGBLOCK_SIZE) {
575 --stat->total_bigblocks;
576 --stat->total_free_bigblocks;
581 * We found a layer2 entry that is not empty!
592 * Non-zero return value means we can't free the volume.
595 test_free_callback(hammer_transaction_t trans, hammer_volume_t volume __unused,
596 hammer_buffer_t *bufferp,
597 struct hammer_blockmap_layer1 *layer1,
598 struct hammer_blockmap_layer2 *layer2,
599 hammer_off_t phys_off,
600 hammer_off_t block_off __unused,
603 if (layer2 == NULL) {
604 return(0); /* only layer2 needs to be tested */
607 if (layer2->zone == HAMMER_ZONE_UNAVAIL_INDEX) {
608 return(0); /* beyond physically available space */
610 if (layer2->zone == HAMMER_ZONE_FREEMAP_INDEX) {
611 return(0); /* big-block for layer1/2 */
613 if (layer2->append_off == 0 &&
614 layer2->bytes_free == HAMMER_BIGBLOCK_SIZE) {
615 return(0); /* big-block is 0% used */
618 return(EBUSY); /* big-block has data */
622 hammer_free_freemap(hammer_transaction_t trans, hammer_volume_t volume,
623 struct bigblock_stat *stat)
627 error = hammer_test_free_freemap(trans, volume);
629 return error; /* not ready to free */
631 stat->total_bigblocks = 0;
632 stat->total_free_bigblocks = 0;
634 return hammer_iterate_l1l2_entries(trans, volume, free_callback, stat);
638 hammer_test_free_freemap(hammer_transaction_t trans, hammer_volume_t volume)
640 return hammer_iterate_l1l2_entries(trans, volume, test_free_callback, NULL);
644 hammer_format_volume_header(struct hammer_mount *hmp,
645 struct hammer_volume_ondisk *ondisk,
646 const char *vol_name, int vol_no, int vol_count,
647 int64_t vol_size, int64_t boot_area_size, int64_t mem_area_size)
651 KKASSERT(HAMMER_BUFSIZE >= sizeof(struct hammer_volume_ondisk));
653 bzero(ondisk, sizeof(struct hammer_volume_ondisk));
654 ksnprintf(ondisk->vol_name, sizeof(ondisk->vol_name), "%s", vol_name);
655 ondisk->vol_fstype = hmp->rootvol->ondisk->vol_fstype;
656 ondisk->vol_signature = HAMMER_FSBUF_VOLUME;
657 ondisk->vol_fsid = hmp->fsid;
658 ondisk->vol_rootvol = hmp->rootvol->vol_no;
659 ondisk->vol_no = vol_no;
660 ondisk->vol_count = vol_count;
661 ondisk->vol_version = hmp->version;
664 * Reserve space for (future) header junk, setup our poor-man's
665 * big-block allocator.
667 vol_alloc = HAMMER_BUFSIZE * 16;
668 ondisk->vol_bot_beg = vol_alloc;
669 vol_alloc += boot_area_size;
670 ondisk->vol_mem_beg = vol_alloc;
671 vol_alloc += mem_area_size;
674 * The remaining area is the zone 2 buffer allocation area. These
677 ondisk->vol_buf_beg = vol_alloc;
678 ondisk->vol_buf_end = vol_size & ~(int64_t)HAMMER_BUFMASK;
680 if (ondisk->vol_buf_end < ondisk->vol_buf_beg) {
681 kprintf("volume %d %s is too small to hold the volume header\n",
682 ondisk->vol_no, ondisk->vol_name);
686 ondisk->vol_nblocks = (ondisk->vol_buf_end - ondisk->vol_buf_beg) /
688 ondisk->vol_blocksize = HAMMER_BUFSIZE;
693 hammer_update_volumes_header(hammer_transaction_t trans,
694 struct bigblock_stat *stat)
696 struct hammer_mount *hmp = trans->hmp;
697 struct mount *mp = hmp->mp;
698 hammer_volume_t volume;
703 * Set each volume's new value of the vol_count field.
705 HAMMER_VOLUME_NUMBER_FOREACH(hmp, vol_no) {
706 volume = hammer_get_volume(hmp, vol_no, &error);
707 KKASSERT(volume != NULL && error == 0);
708 hammer_modify_volume_field(trans, volume, vol_count);
709 volume->ondisk->vol_count = hmp->nvolumes;
710 hammer_modify_volume_done(volume);
713 * Only changes to the header of the root volume
714 * are automatically flushed to disk. For all
715 * other volumes that we modify we do it here.
717 * No interlock is needed, volume buffers are not
718 * messed with by bioops.
720 if (volume != trans->rootvol && volume->io.modified) {
721 hammer_crc_set_volume(volume->ondisk);
722 hammer_io_flush(&volume->io, 0);
725 hammer_rel_volume(volume, 0);
729 * Update the total number of big-blocks.
731 hammer_modify_volume_field(trans, trans->rootvol, vol0_stat_bigblocks);
732 trans->rootvol->ondisk->vol0_stat_bigblocks += stat->total_bigblocks;
733 hammer_modify_volume_done(trans->rootvol);
736 * Big-block count changed so recompute the total number of blocks.
738 mp->mnt_stat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
739 HAMMER_BUFFERS_PER_BIGBLOCK;
740 mp->mnt_vstat.f_blocks = trans->rootvol->ondisk->vol0_stat_bigblocks *
741 HAMMER_BUFFERS_PER_BIGBLOCK;
744 * Update the total number of free big-blocks.
746 hammer_modify_volume_field(trans, trans->rootvol,
747 vol0_stat_freebigblocks);
748 trans->rootvol->ondisk->vol0_stat_freebigblocks +=
749 stat->total_free_bigblocks;
750 hammer_modify_volume_done(trans->rootvol);
753 * Update the copy in hmp.
755 hmp->copy_stat_freebigblocks =
756 trans->rootvol->ondisk->vol0_stat_freebigblocks;