1 // SPDX-License-Identifier: GPL-2.0
5 #include "block-group.h"
6 #include "space-info.h"
7 #include "disk-io.h"
8 #include "free-space-cache.h"
9 #include "free-space-tree.h"
12 #include "ref-verify.h"
14 #include "tree-log.h"
15 #include "delalloc-space.h"
28 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;  in get_restripe_target()
35 	    bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) {  in get_restripe_target()
36 		target = BTRFS_BLOCK_GROUP_DATA | bctl->data.target;  in get_restripe_target()
38 		   bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {  in get_restripe_target()
39 		target = BTRFS_BLOCK_GROUP_SYSTEM | bctl->sys.target;  in get_restripe_target()
41 		   bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) {  in get_restripe_target()
42 		target = BTRFS_BLOCK_GROUP_METADATA | bctl->meta.target;  in get_restripe_target()
57 	u64 num_devices = fs_info->fs_devices->rw_devices;  in btrfs_reduce_alloc_profile()
66 	spin_lock(&fs_info->balance_lock);  in btrfs_reduce_alloc_profile()
69 		spin_unlock(&fs_info->balance_lock);  in btrfs_reduce_alloc_profile()
72 	spin_unlock(&fs_info->balance_lock);  in btrfs_reduce_alloc_profile()
104 		seq = read_seqbegin(&fs_info->profiles_lock);  in btrfs_get_alloc_profile()
107 			flags |= fs_info->avail_data_alloc_bits;  in btrfs_get_alloc_profile()
109 			flags |= fs_info->avail_system_alloc_bits;  in btrfs_get_alloc_profile()
111 			flags |= fs_info->avail_metadata_alloc_bits;  in btrfs_get_alloc_profile()
112 	} while (read_seqretry(&fs_info->profiles_lock, seq));  in btrfs_get_alloc_profile()
117 void btrfs_get_block_group(struct btrfs_block_group *cache)  in btrfs_get_block_group()  argument
119 	refcount_inc(&cache->refs);  in btrfs_get_block_group()
122 void btrfs_put_block_group(struct btrfs_block_group *cache)  in btrfs_put_block_group()  argument
124 	if (refcount_dec_and_test(&cache->refs)) {  in btrfs_put_block_group()
125 		WARN_ON(cache->pinned > 0);  in btrfs_put_block_group()
126 		WARN_ON(cache->reserved > 0);  in btrfs_put_block_group()
133 		if (WARN_ON(!list_empty(&cache->discard_list)))  in btrfs_put_block_group()
134 			btrfs_discard_cancel_work(&cache->fs_info->discard_ctl,  in btrfs_put_block_group()
135 						  cache);  in btrfs_put_block_group()
140 		 * And it will definitely cause use-after-free when caller  in btrfs_put_block_group()
145 		WARN_ON(!RB_EMPTY_ROOT(&cache->full_stripe_locks_root.root));  in btrfs_put_block_group()
146 		kfree(cache->free_space_ctl);  in btrfs_put_block_group()
147 		kfree(cache);  in btrfs_put_block_group()
152  * This adds the block group to the fs_info rb tree for the block group cache
159 	struct btrfs_block_group *cache;  in btrfs_add_block_group_cache()  local
161 	ASSERT(block_group->length != 0);  in btrfs_add_block_group_cache()
163 	spin_lock(&info->block_group_cache_lock);  in btrfs_add_block_group_cache()
164 	p = &info->block_group_cache_tree.rb_node;  in btrfs_add_block_group_cache()
168 		cache = rb_entry(parent, struct btrfs_block_group, cache_node);  in btrfs_add_block_group_cache()
169 		if (block_group->start < cache->start) {  in btrfs_add_block_group_cache()
170 			p = &(*p)->rb_left;  in btrfs_add_block_group_cache()
171 		} else if (block_group->start > cache->start) {  in btrfs_add_block_group_cache()
172 			p = &(*p)->rb_right;  in btrfs_add_block_group_cache()
174 			spin_unlock(&info->block_group_cache_lock);  in btrfs_add_block_group_cache()
175 			return -EEXIST;  in btrfs_add_block_group_cache()
179 	rb_link_node(&block_group->cache_node, parent, p);  in btrfs_add_block_group_cache()
180 	rb_insert_color(&block_group->cache_node,  in btrfs_add_block_group_cache()
181 			&info->block_group_cache_tree);  in btrfs_add_block_group_cache()
183 	if (info->first_logical_byte > block_group->start)  in btrfs_add_block_group_cache()
184 		info->first_logical_byte = block_group->start;  in btrfs_add_block_group_cache()
186 	spin_unlock(&info->block_group_cache_lock);  in btrfs_add_block_group_cache()
192  * This will return the block group at or after bytenr if contains is 0, else
193  * it will return the block group that contains the bytenr
198 	struct btrfs_block_group *cache, *ret = NULL;  in block_group_cache_tree_search()  local
202 	spin_lock(&info->block_group_cache_lock);  in block_group_cache_tree_search()
203 	n = info->block_group_cache_tree.rb_node;  in block_group_cache_tree_search()
206 		cache = rb_entry(n, struct btrfs_block_group, cache_node);  in block_group_cache_tree_search()
207 		end = cache->start + cache->length - 1;  in block_group_cache_tree_search()
208 		start = cache->start;  in block_group_cache_tree_search()
211 			if (!contains && (!ret || start < ret->start))  in block_group_cache_tree_search()
212 				ret = cache;  in block_group_cache_tree_search()
213 			n = n->rb_left;  in block_group_cache_tree_search()
216 				ret = cache;  in block_group_cache_tree_search()
219 			n = n->rb_right;  in block_group_cache_tree_search()
221 			ret = cache;  in block_group_cache_tree_search()
227 		if (bytenr == 0 && info->first_logical_byte > ret->start)  in block_group_cache_tree_search()
228 			info->first_logical_byte = ret->start;  in block_group_cache_tree_search()
230 	spin_unlock(&info->block_group_cache_lock);  in block_group_cache_tree_search()
236  * Return the block group that starts at or after bytenr
245  * Return the block group that contains the given bytenr
254 		struct btrfs_block_group *cache)  in btrfs_next_block_group()  argument
256 	struct btrfs_fs_info *fs_info = cache->fs_info;  in btrfs_next_block_group()
259 	spin_lock(&fs_info->block_group_cache_lock);  in btrfs_next_block_group()
261 	/* If our block group was removed, we need a full search. */  in btrfs_next_block_group()
262 	if (RB_EMPTY_NODE(&cache->cache_node)) {  in btrfs_next_block_group()
263 		const u64 next_bytenr = cache->start + cache->length;  in btrfs_next_block_group()
265 		spin_unlock(&fs_info->block_group_cache_lock);  in btrfs_next_block_group()
266 		btrfs_put_block_group(cache);  in btrfs_next_block_group()
267 		cache = btrfs_lookup_first_block_group(fs_info, next_bytenr); return cache;  in btrfs_next_block_group()
269 	node = rb_next(&cache->cache_node);  in btrfs_next_block_group()
270 	btrfs_put_block_group(cache);  in btrfs_next_block_group()
272 		cache = rb_entry(node, struct btrfs_block_group, cache_node);  in btrfs_next_block_group()
273 		btrfs_get_block_group(cache);  in btrfs_next_block_group()
275 		cache = NULL;  in btrfs_next_block_group()
276 	spin_unlock(&fs_info->block_group_cache_lock);  in btrfs_next_block_group()
277 	return cache;  in btrfs_next_block_group()
289 	spin_lock(&bg->lock);  in btrfs_inc_nocow_writers()
290 	if (bg->ro)  in btrfs_inc_nocow_writers()
293 		atomic_inc(&bg->nocow_writers);  in btrfs_inc_nocow_writers()
294 	spin_unlock(&bg->lock);  in btrfs_inc_nocow_writers()
296 	/* No put on block group, done by btrfs_dec_nocow_writers */  in btrfs_inc_nocow_writers()
309 	if (atomic_dec_and_test(&bg->nocow_writers))  in btrfs_dec_nocow_writers()
310 		wake_up_var(&bg->nocow_writers);  in btrfs_dec_nocow_writers()
321 	wait_var_event(&bg->nocow_writers, !atomic_read(&bg->nocow_writers));  in btrfs_wait_nocow_writers()
331 	if (atomic_dec_and_test(&bg->reservations))  in btrfs_dec_block_group_reservations()
332 		wake_up_var(&bg->reservations);  in btrfs_dec_block_group_reservations()
338 	struct btrfs_space_info *space_info = bg->space_info;  in btrfs_wait_block_group_reservations()
340 	ASSERT(bg->ro);  in btrfs_wait_block_group_reservations()
342 	if (!(bg->flags & BTRFS_BLOCK_GROUP_DATA))  in btrfs_wait_block_group_reservations()
346 	 * Our block group is read only but before we set it to read only,  in btrfs_wait_block_group_reservations()
351 	 * block group's reservations counter is incremented while a read lock  in btrfs_wait_block_group_reservations()
355 	down_write(&space_info->groups_sem);  in btrfs_wait_block_group_reservations()
356 	up_write(&space_info->groups_sem);  in btrfs_wait_block_group_reservations()
358 	wait_var_event(&bg->reservations, !atomic_read(&bg->reservations));  in btrfs_wait_block_group_reservations()
362 		struct btrfs_block_group *cache)  in btrfs_get_caching_control()  argument
366 	spin_lock(&cache->lock);  in btrfs_get_caching_control()
367 	if (!cache->caching_ctl) {  in btrfs_get_caching_control()
368 		spin_unlock(&cache->lock);  in btrfs_get_caching_control()
372 	ctl = cache->caching_ctl;  in btrfs_get_caching_control()
373 	refcount_inc(&ctl->count);  in btrfs_get_caching_control()
374 	spin_unlock(&cache->lock);  in btrfs_get_caching_control()
380 	if (refcount_dec_and_test(&ctl->count))  in btrfs_put_caching_control()
385  * When we wait for progress in the block group caching, its because our
390  * up, and then it will check the block group free space numbers for our min
392  * a free extent of a given size, but this is a good start.
394  * Callers of this must check if cache->cached == BTRFS_CACHE_ERROR before using
395  * any of the information in this block group.
397 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,  in btrfs_wait_block_group_cache_progress()  argument
402 	caching_ctl = btrfs_get_caching_control(cache);  in btrfs_wait_block_group_cache_progress()
406 	wait_event(caching_ctl->wait, btrfs_block_group_done(cache) ||  in btrfs_wait_block_group_cache_progress()
407 		   (cache->free_space_ctl->free_space >= num_bytes));  in btrfs_wait_block_group_cache_progress()
412 int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache)  in btrfs_wait_block_group_cache_done()  argument
417 	caching_ctl = btrfs_get_caching_control(cache);  in btrfs_wait_block_group_cache_done()
419 		return (cache->cached == BTRFS_CACHE_ERROR) ? -EIO : 0;  in btrfs_wait_block_group_cache_done()
421 	wait_event(caching_ctl->wait, btrfs_block_group_done(cache));  in btrfs_wait_block_group_cache_done()
422 	if (cache->cached == BTRFS_CACHE_ERROR)  in btrfs_wait_block_group_cache_done()
423 		ret = -EIO;  in btrfs_wait_block_group_cache_done()
428 static bool space_cache_v1_done(struct btrfs_block_group *cache)  in space_cache_v1_done()  argument
432 	spin_lock(&cache->lock);  in space_cache_v1_done()
433 	ret = cache->cached != BTRFS_CACHE_FAST;  in space_cache_v1_done()
434 	spin_unlock(&cache->lock);  in space_cache_v1_done()
439 void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache,  in btrfs_wait_space_cache_v1_finished()  argument
442 	wait_event(caching_ctl->wait, space_cache_v1_done(cache));  in btrfs_wait_space_cache_v1_finished()
448 	struct btrfs_fs_info *fs_info = block_group->fs_info;  in fragment_free_space()
449 	u64 start = block_group->start;  in fragment_free_space()
450 	u64 len = block_group->length;  in fragment_free_space()
451 	u64 chunk = block_group->flags & BTRFS_BLOCK_GROUP_METADATA ?  in fragment_free_space()
452 		fs_info->nodesize : fs_info->sectorsize;  in fragment_free_space()
461 			len -= step;  in fragment_free_space()
474 	struct btrfs_fs_info *info = block_group->fs_info;  in add_new_free_space()
475 	u64 extent_start, extent_end, size, total_added = 0;  in add_new_free_space()  local
479 		ret = find_first_extent_bit(&info->excluded_extents, start,  in add_new_free_space()
489 			size = extent_start - start;  in add_new_free_space()
490 			total_added += size;  in add_new_free_space()
492 								 start, size);  in add_new_free_space()
493 			BUG_ON(ret); /* -ENOMEM or logic error */  in add_new_free_space()
501 		size = end - start;  in add_new_free_space()
502 		total_added += size;  in add_new_free_space()
504 							 size);  in add_new_free_space()
505 		BUG_ON(ret); /* -ENOMEM or logic error */  in add_new_free_space()
513 	struct btrfs_block_group *block_group = caching_ctl->block_group;  in load_extent_tree_free()
514 	struct btrfs_fs_info *fs_info = block_group->fs_info;  in load_extent_tree_free()
515 	struct btrfs_root *extent_root = fs_info->extent_root;  in load_extent_tree_free()
527 		return -ENOMEM;  in load_extent_tree_free()
529 	last = max_t(u64, block_group->start, BTRFS_SUPER_INFO_OFFSET);  in load_extent_tree_free()
534 	 * allocate from this block group until we've had a chance to fragment  in load_extent_tree_free()
544 	 * root, since its read-only  in load_extent_tree_free()
546 	path->skip_locking = 1;  in load_extent_tree_free()
547 	path->search_commit_root = 1;  in load_extent_tree_free()
548 	path->reada = READA_FORWARD;  in load_extent_tree_free()
559 	leaf = path->nodes[0];  in load_extent_tree_free()
564 			last = (u64)-1;  in load_extent_tree_free()
568 		if (path->slots[0] < nritems) {  in load_extent_tree_free()
569 			btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);  in load_extent_tree_free()
576 			    rwsem_is_contended(&fs_info->commit_root_sem)) {  in load_extent_tree_free()
578 					caching_ctl->progress = last;  in load_extent_tree_free()
580 				up_read(&fs_info->commit_root_sem);  in load_extent_tree_free()
581 				mutex_unlock(&caching_ctl->mutex);  in load_extent_tree_free()
583 				mutex_lock(&caching_ctl->mutex);  in load_extent_tree_free()
584 				down_read(&fs_info->commit_root_sem);  in load_extent_tree_free()
593 			leaf = path->nodes[0];  in load_extent_tree_free()
604 				caching_ctl->progress = last;  in load_extent_tree_free()
609 		if (key.objectid < block_group->start) {  in load_extent_tree_free()
610 			path->slots[0]++;  in load_extent_tree_free()
614 		if (key.objectid >= block_group->start + block_group->length)  in load_extent_tree_free()
623 					fs_info->nodesize;  in load_extent_tree_free()
630 					wake_up(&caching_ctl->wait);  in load_extent_tree_free()
633 		path->slots[0]++;  in load_extent_tree_free()
638 				block_group->start + block_group->length);  in load_extent_tree_free()
639 	caching_ctl->progress = (u64)-1;  in load_extent_tree_free()
654 	block_group = caching_ctl->block_group;  in caching_thread()
655 	fs_info = block_group->fs_info;  in caching_thread()
657 	mutex_lock(&caching_ctl->mutex);  in caching_thread()
658 	down_read(&fs_info->commit_root_sem);  in caching_thread()
668 		 * We failed to load the space cache, set ourselves to  in caching_thread()
671 		spin_lock(&block_group->lock);  in caching_thread()
672 		block_group->cached = BTRFS_CACHE_STARTED;  in caching_thread()
673 		spin_unlock(&block_group->lock);  in caching_thread()
674 		wake_up(&caching_ctl->wait);  in caching_thread()
679 	 * can't actually cache from the free space tree as our commit root and  in caching_thread()
685 	    !(test_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags)))  in caching_thread()
690 	spin_lock(&block_group->lock);  in caching_thread()
691 	block_group->caching_ctl = NULL;  in caching_thread()
692 	block_group->cached = ret ? BTRFS_CACHE_ERROR : BTRFS_CACHE_FINISHED;  in caching_thread()
693 	spin_unlock(&block_group->lock);  in caching_thread()
699 		spin_lock(&block_group->space_info->lock);  in caching_thread()
700 		spin_lock(&block_group->lock);  in caching_thread()
701 		bytes_used = block_group->length - block_group->used;  in caching_thread()
702 		block_group->space_info->bytes_used += bytes_used >> 1;  in caching_thread()
703 		spin_unlock(&block_group->lock);  in caching_thread()
704 		spin_unlock(&block_group->space_info->lock);  in caching_thread()
709 	caching_ctl->progress = (u64)-1;  in caching_thread()
711 	up_read(&fs_info->commit_root_sem);  in caching_thread()
713 	mutex_unlock(&caching_ctl->mutex);  in caching_thread()
715 	wake_up(&caching_ctl->wait);  in caching_thread()
721 int btrfs_cache_block_group(struct btrfs_block_group *cache, int load_cache_only)  in btrfs_cache_block_group()  argument
724 	struct btrfs_fs_info *fs_info = cache->fs_info;  in btrfs_cache_block_group()
728 	/* Allocator for zoned filesystems does not use the cache at all */  in btrfs_cache_block_group()
734 		return -ENOMEM;  in btrfs_cache_block_group()
736 	INIT_LIST_HEAD(&caching_ctl->list);  in btrfs_cache_block_group()
737 	mutex_init(&caching_ctl->mutex);  in btrfs_cache_block_group()
738 	init_waitqueue_head(&caching_ctl->wait);  in btrfs_cache_block_group()
739 	caching_ctl->block_group = cache;  in btrfs_cache_block_group()
740 	caching_ctl->progress = cache->start;  in btrfs_cache_block_group()
741 	refcount_set(&caching_ctl->count, 2);  in btrfs_cache_block_group()
742 	btrfs_init_work(&caching_ctl->work, caching_thread, NULL, NULL);  in btrfs_cache_block_group()
744 	spin_lock(&cache->lock);  in btrfs_cache_block_group()
745 	if (cache->cached != BTRFS_CACHE_NO) {  in btrfs_cache_block_group()
748 		caching_ctl = cache->caching_ctl;  in btrfs_cache_block_group()
750 			refcount_inc(&caching_ctl->count);  in btrfs_cache_block_group()
751 		spin_unlock(&cache->lock);  in btrfs_cache_block_group()
754 	WARN_ON(cache->caching_ctl);  in btrfs_cache_block_group()
755 	cache->caching_ctl = caching_ctl;  in btrfs_cache_block_group()
757 		cache->cached = BTRFS_CACHE_FAST;  in btrfs_cache_block_group()
759 		cache->cached = BTRFS_CACHE_STARTED;  in btrfs_cache_block_group()
760 	cache->has_caching_ctl = 1;  in btrfs_cache_block_group()
761 	spin_unlock(&cache->lock);  in btrfs_cache_block_group()
763 	spin_lock(&fs_info->block_group_cache_lock);  in btrfs_cache_block_group()
764 	refcount_inc(&caching_ctl->count);  in btrfs_cache_block_group()
765 	list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);  in btrfs_cache_block_group()
766 	spin_unlock(&fs_info->block_group_cache_lock);  in btrfs_cache_block_group()
768 	btrfs_get_block_group(cache);  in btrfs_cache_block_group()
770 	btrfs_queue_work(fs_info->caching_workers, &caching_ctl->work);  in btrfs_cache_block_group()
773 		btrfs_wait_space_cache_v1_finished(cache, caching_ctl);  in btrfs_cache_block_group()
785 	write_seqlock(&fs_info->profiles_lock);  in clear_avail_alloc_bits()
787 		fs_info->avail_data_alloc_bits &= ~extra_flags;  in clear_avail_alloc_bits()
789 		fs_info->avail_metadata_alloc_bits &= ~extra_flags;  in clear_avail_alloc_bits()
791 		fs_info->avail_system_alloc_bits &= ~extra_flags;  in clear_avail_alloc_bits()
792 	write_sequnlock(&fs_info->profiles_lock);  in clear_avail_alloc_bits()
798  * - RAID56 - in case there's neither RAID5 nor RAID6 profile block group
801  * - RAID1C34 - same as above for RAID1C3 and RAID1C4 block groups
811 		struct list_head *head = &fs_info->space_info;  in clear_incompat_bg_bits()
815 			down_read(&sinfo->groups_sem);  in clear_incompat_bg_bits()
816 			if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID5]))  in clear_incompat_bg_bits()
818 			if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID6]))  in clear_incompat_bg_bits()
820 			if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID1C3]))  in clear_incompat_bg_bits()
822 			if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID1C4]))  in clear_incompat_bg_bits()
824 			up_read(&sinfo->groups_sem);  in clear_incompat_bg_bits()
837 	struct btrfs_fs_info *fs_info = trans->fs_info;  in remove_block_group_item()
842 	root = fs_info->extent_root;  in remove_block_group_item()
843 	key.objectid = block_group->start;  in remove_block_group_item()
845 	key.offset = block_group->length;  in remove_block_group_item()
847 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);  in remove_block_group_item()
849 		ret = -ENOENT;  in remove_block_group_item()
860 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_remove_block_group()
875 	BUG_ON(!block_group->ro);  in btrfs_remove_block_group()
879 	 * Free the reserved super bytes from this block group before  in btrfs_remove_block_group()
883 	btrfs_free_ref_tree_range(fs_info, block_group->start,  in btrfs_remove_block_group()
884 				  block_group->length);  in btrfs_remove_block_group()
886 	index = btrfs_bg_flags_to_raid_index(block_group->flags);  in btrfs_remove_block_group()
887 	factor = btrfs_bg_type_to_factor(block_group->flags);  in btrfs_remove_block_group()
889 	/* make sure this block group isn't part of an allocation cluster */  in btrfs_remove_block_group()
890 	cluster = &fs_info->data_alloc_cluster;  in btrfs_remove_block_group()
891 	spin_lock(&cluster->refill_lock);  in btrfs_remove_block_group()
893 	spin_unlock(&cluster->refill_lock);  in btrfs_remove_block_group()
896 	 * make sure this block group isn't part of a metadata  in btrfs_remove_block_group()
899 	cluster = &fs_info->meta_alloc_cluster;  in btrfs_remove_block_group()
900 	spin_lock(&cluster->refill_lock);  in btrfs_remove_block_group()
902 	spin_unlock(&cluster->refill_lock);  in btrfs_remove_block_group()
908 		ret = -ENOMEM;  in btrfs_remove_block_group()
918 	mutex_lock(&trans->transaction->cache_write_mutex);  in btrfs_remove_block_group()
920 	 * Make sure our free space cache IO is done before removing the  in btrfs_remove_block_group()
923 	spin_lock(&trans->transaction->dirty_bgs_lock);  in btrfs_remove_block_group()
924 	if (!list_empty(&block_group->io_list)) {  in btrfs_remove_block_group()
925 		list_del_init(&block_group->io_list);  in btrfs_remove_block_group()
927 		WARN_ON(!IS_ERR(inode) && inode != block_group->io_ctl.inode);  in btrfs_remove_block_group()
929 		spin_unlock(&trans->transaction->dirty_bgs_lock);  in btrfs_remove_block_group()
932 		spin_lock(&trans->transaction->dirty_bgs_lock);  in btrfs_remove_block_group()
935 	if (!list_empty(&block_group->dirty_list)) {  in btrfs_remove_block_group()
936 		list_del_init(&block_group->dirty_list);  in btrfs_remove_block_group()
940 	spin_unlock(&trans->transaction->dirty_bgs_lock);  in btrfs_remove_block_group()
941 	mutex_unlock(&trans->transaction->cache_write_mutex);  in btrfs_remove_block_group()
947 	spin_lock(&fs_info->block_group_cache_lock);  in btrfs_remove_block_group()
948 	rb_erase(&block_group->cache_node,  in btrfs_remove_block_group()
949 		 &fs_info->block_group_cache_tree);  in btrfs_remove_block_group()
950 	RB_CLEAR_NODE(&block_group->cache_node);  in btrfs_remove_block_group()
952 	/* Once for the block groups rbtree */  in btrfs_remove_block_group()
955 	if (fs_info->first_logical_byte == block_group->start)  in btrfs_remove_block_group()
956 		fs_info->first_logical_byte = (u64)-1;  in btrfs_remove_block_group()
957 	spin_unlock(&fs_info->block_group_cache_lock);  in btrfs_remove_block_group()
959 	down_write(&block_group->space_info->groups_sem);  in btrfs_remove_block_group()
964 	list_del_init(&block_group->list);  in btrfs_remove_block_group()
965 	if (list_empty(&block_group->space_info->block_groups[index])) {  in btrfs_remove_block_group()
966 		kobj = block_group->space_info->block_group_kobjs[index];  in btrfs_remove_block_group()
967 		block_group->space_info->block_group_kobjs[index] = NULL;  in btrfs_remove_block_group()
968 		clear_avail_alloc_bits(fs_info, block_group->flags);  in btrfs_remove_block_group()
970 	up_write(&block_group->space_info->groups_sem);  in btrfs_remove_block_group()
971 	clear_incompat_bg_bits(fs_info, block_group->flags);  in btrfs_remove_block_group()
977 	if (block_group->has_caching_ctl)  in btrfs_remove_block_group()
979 	if (block_group->cached == BTRFS_CACHE_STARTED)  in btrfs_remove_block_group()
981 	if (block_group->has_caching_ctl) {  in btrfs_remove_block_group()
982 		spin_lock(&fs_info->block_group_cache_lock);  in btrfs_remove_block_group()
987 				    &fs_info->caching_block_groups, list)  in btrfs_remove_block_group()
988 				if (ctl->block_group == block_group) {  in btrfs_remove_block_group()
990 					refcount_inc(&caching_ctl->count);  in btrfs_remove_block_group()
995 			list_del_init(&caching_ctl->list);  in btrfs_remove_block_group()
996 		spin_unlock(&fs_info->block_group_cache_lock);  in btrfs_remove_block_group()
1004 	spin_lock(&trans->transaction->dirty_bgs_lock);  in btrfs_remove_block_group()
1005 	WARN_ON(!list_empty(&block_group->dirty_list));  in btrfs_remove_block_group()
1006 	WARN_ON(!list_empty(&block_group->io_list));  in btrfs_remove_block_group()
1007 	spin_unlock(&trans->transaction->dirty_bgs_lock);  in btrfs_remove_block_group()
1011 	spin_lock(&block_group->space_info->lock);  in btrfs_remove_block_group()
1012 	list_del_init(&block_group->ro_list);  in btrfs_remove_block_group()
1015 		WARN_ON(block_group->space_info->total_bytes  in btrfs_remove_block_group()
1016 			< block_group->length);  in btrfs_remove_block_group()
1017 		WARN_ON(block_group->space_info->bytes_readonly  in btrfs_remove_block_group()
1018 			< block_group->length - block_group->zone_unusable);  in btrfs_remove_block_group()
1019 		WARN_ON(block_group->space_info->bytes_zone_unusable  in btrfs_remove_block_group()
1020 			< block_group->zone_unusable);  in btrfs_remove_block_group()
1021 		WARN_ON(block_group->space_info->disk_total  in btrfs_remove_block_group()
1022 			< block_group->length * factor);  in btrfs_remove_block_group()
1024 	block_group->space_info->total_bytes -= block_group->length;  in btrfs_remove_block_group()
1025 	block_group->space_info->bytes_readonly -=  in btrfs_remove_block_group()
1026 		(block_group->length - block_group->zone_unusable);  in btrfs_remove_block_group()
1027 	block_group->space_info->bytes_zone_unusable -=  in btrfs_remove_block_group()
1028 		block_group->zone_unusable;  in btrfs_remove_block_group()
1029 	block_group->space_info->disk_total -= block_group->length * factor;  in btrfs_remove_block_group()
1031 	spin_unlock(&block_group->space_info->lock);  in btrfs_remove_block_group()
1034 	 * Remove the free space for the block group from the free space tree  in btrfs_remove_block_group()
1035 	 * and the block group's item from the extent tree before marking the  in btrfs_remove_block_group()
1036 	 * block group as removed. This is to prevent races with tasks that  in btrfs_remove_block_group()
1037 	 * freeze and unfreeze a block group, this task and another task  in btrfs_remove_block_group()
1038 	 * allocating a new block group - the unfreeze task ends up removing  in btrfs_remove_block_group()
1039 	 * the block group's extent map before the task calling this function  in btrfs_remove_block_group()
1040 	 * deletes the block group item from the extent tree, allowing for  in btrfs_remove_block_group()
1041 	 * another task to attempt to create another block group with the same  in btrfs_remove_block_group()
1042 	 * item key (and failing with -EEXIST and a transaction abort).  in btrfs_remove_block_group()
1052 	spin_lock(&block_group->lock);  in btrfs_remove_block_group()
1053 	block_group->removed = 1;  in btrfs_remove_block_group()
1055 	 * At this point trimming or scrub can't start on this block group,  in btrfs_remove_block_group()
1056 	 * because we removed the block group from the rbtree  in btrfs_remove_block_group()
1057 	 * fs_info->block_group_cache_tree so no one can't find it anymore and  in btrfs_remove_block_group()
1058 	 * even if someone already got this block group before we removed it  in btrfs_remove_block_group()
1059 	 * from the rbtree, they have already incremented block_group->frozen -  in btrfs_remove_block_group()
1064 	 * And we must not remove the extent map from the fs_info->mapping_tree  in btrfs_remove_block_group()
1066 	 * ranges from being reused for a new block group. This is needed to  in btrfs_remove_block_group()
1072 	 * allowing for new block groups to be created that can reuse the same  in btrfs_remove_block_group()
1076 	 * is mounted with -odiscard. The same protections must remain  in btrfs_remove_block_group()
1080 	remove_em = (atomic_read(&block_group->frozen) == 0);  in btrfs_remove_block_group()
1081 	spin_unlock(&block_group->lock);  in btrfs_remove_block_group()
1086 		em_tree = &fs_info->mapping_tree;  in btrfs_remove_block_group()
1087 		write_lock(&em_tree->lock);  in btrfs_remove_block_group()
1089 		write_unlock(&em_tree->lock);  in btrfs_remove_block_group()
1106 	struct extent_map_tree *em_tree = &fs_info->mapping_tree;  in btrfs_start_trans_remove_block_group()
1111 	read_lock(&em_tree->lock);  in btrfs_start_trans_remove_block_group()
1113 	read_unlock(&em_tree->lock);  in btrfs_start_trans_remove_block_group()
1114 	ASSERT(em && em->start == chunk_offset);  in btrfs_start_trans_remove_block_group()
1118 	 * to remove a block group (done at btrfs_remove_chunk() and at  in btrfs_start_trans_remove_block_group()
1123 	 * 1 unit for deleting the block group item (located in the extent  in btrfs_start_trans_remove_block_group()
1130 	 * In order to remove a block group we also need to reserve units in the  in btrfs_start_trans_remove_block_group()
1135 	map = em->map_lookup;  in btrfs_start_trans_remove_block_group()
1136 	num_items = 3 + map->num_stripes;  in btrfs_start_trans_remove_block_group()
1139 	return btrfs_start_transaction_fallback_global_rsv(fs_info->extent_root,  in btrfs_start_trans_remove_block_group()
1144  * Mark block group @cache read-only, so later write won't happen to block
1145  * group @cache.
1147  * If @force is not set, this function will only mark the block group readonly
1148  * if we have enough free space (1M) in other metadata/system block groups.
1149  * If @force is not set, this function will mark the block group readonly
1152  * NOTE: This function doesn't care if other block groups can contain all the
1153  * data in this block group. That check should be done by relocation routine,
1156 static int inc_block_group_ro(struct btrfs_block_group *cache, int force)  in inc_block_group_ro()  argument
1158 	struct btrfs_space_info *sinfo = cache->space_info;  in inc_block_group_ro()
1160 	int ret = -ENOSPC;  in inc_block_group_ro()
1162 	spin_lock(&sinfo->lock);  in inc_block_group_ro()
1163 	spin_lock(&cache->lock);  in inc_block_group_ro()
1165 	if (cache->swap_extents) {  in inc_block_group_ro()
1166 		ret = -ETXTBSY;  in inc_block_group_ro()
1170 	if (cache->ro) {  in inc_block_group_ro()
1171 		cache->ro++;  in inc_block_group_ro()
1176 	num_bytes = cache->length - cache->reserved - cache->pinned -  in inc_block_group_ro()
1177 		    cache->bytes_super - cache->zone_unusable - cache->used;  in inc_block_group_ro()
1185 	} else if (sinfo->flags & BTRFS_BLOCK_GROUP_DATA) {  in inc_block_group_ro()
1192 		if (sinfo_used + num_bytes <= sinfo->total_bytes)  in inc_block_group_ro()
1199 		 * leeway to allow us to mark this block group as read only.  in inc_block_group_ro()
1201 		if (btrfs_can_overcommit(cache->fs_info, sinfo, num_bytes,  in inc_block_group_ro()
1207 		sinfo->bytes_readonly += num_bytes;  in inc_block_group_ro()
1208 		if (btrfs_is_zoned(cache->fs_info)) {  in inc_block_group_ro()
1210 			sinfo->bytes_readonly += cache->zone_unusable;  in inc_block_group_ro()
1211 			sinfo->bytes_zone_unusable -= cache->zone_unusable;  in inc_block_group_ro()
1212 			cache->zone_unusable = 0;  in inc_block_group_ro()
1214 		cache->ro++;  in inc_block_group_ro()
1215 		list_add_tail(&cache->ro_list, &sinfo->ro_bgs);  in inc_block_group_ro()
1218 	spin_unlock(&cache->lock);  in inc_block_group_ro()
1219 	spin_unlock(&sinfo->lock);  in inc_block_group_ro()
1220 	if (ret == -ENOSPC && btrfs_test_opt(cache->fs_info, ENOSPC_DEBUG)) {  in inc_block_group_ro()
1221 		btrfs_info(cache->fs_info,  in inc_block_group_ro()
1222 			"unable to make block group %llu ro", cache->start);  in inc_block_group_ro()
1223 		btrfs_dump_space_info(cache->fs_info, cache->space_info, 0, 0);  in inc_block_group_ro()
1231 	struct btrfs_fs_info *fs_info = bg->fs_info;  in clean_pinned_extents()
1233 	const u64 start = bg->start;  in clean_pinned_extents()
1234 	const u64 end = start + bg->length - 1;  in clean_pinned_extents()
1237 	spin_lock(&fs_info->trans_lock);  in clean_pinned_extents()
1238 	if (trans->transaction->list.prev != &fs_info->trans_list) {  in clean_pinned_extents()
1239 		prev_trans = list_last_entry(&trans->transaction->list,  in clean_pinned_extents()
1241 		refcount_inc(&prev_trans->use_count);  in clean_pinned_extents()
1243 	spin_unlock(&fs_info->trans_lock);  in clean_pinned_extents()
1249 	 * transaction N - 1, and have seen a range belonging to the block  in clean_pinned_extents()
1250 	 * group in pinned_extents before we were able to clear the whole block  in clean_pinned_extents()
1252 	 * the block group after we unpinned it from pinned_extents and removed  in clean_pinned_extents()
1255 	mutex_lock(&fs_info->unused_bg_unpin_mutex);  in clean_pinned_extents()
1257 		ret = clear_extent_bits(&prev_trans->pinned_extents, start, end,  in clean_pinned_extents()
1263 	ret = clear_extent_bits(&trans->transaction->pinned_extents, start, end,  in clean_pinned_extents()
1266 	mutex_unlock(&fs_info->unused_bg_unpin_mutex);  in clean_pinned_extents()
1285 	if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))  in btrfs_delete_unused_bgs()
1292 	if (!mutex_trylock(&fs_info->reclaim_bgs_lock))  in btrfs_delete_unused_bgs()
1295 	spin_lock(&fs_info->unused_bgs_lock);  in btrfs_delete_unused_bgs()
1296 	while (!list_empty(&fs_info->unused_bgs)) {  in btrfs_delete_unused_bgs()
1299 		block_group = list_first_entry(&fs_info->unused_bgs,  in btrfs_delete_unused_bgs()
1302 		list_del_init(&block_group->bg_list);  in btrfs_delete_unused_bgs()
1304 		space_info = block_group->space_info;  in btrfs_delete_unused_bgs()
1310 		spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_delete_unused_bgs()
1312 		btrfs_discard_cancel_work(&fs_info->discard_ctl, block_group);  in btrfs_delete_unused_bgs()
1315 		down_write(&space_info->groups_sem);  in btrfs_delete_unused_bgs()
1318 		 * Async discard moves the final block group discard to be prior  in btrfs_delete_unused_bgs()
1325 			up_write(&space_info->groups_sem);  in btrfs_delete_unused_bgs()
1327 			btrfs_discard_queue_work(&fs_info->discard_ctl,  in btrfs_delete_unused_bgs()
1332 		spin_lock(&block_group->lock);  in btrfs_delete_unused_bgs()
1333 		if (block_group->reserved || block_group->pinned ||  in btrfs_delete_unused_bgs()
1334 		    block_group->used || block_group->ro ||  in btrfs_delete_unused_bgs()
1335 		    list_is_singular(&block_group->list)) {  in btrfs_delete_unused_bgs()
1338 			 * outstanding allocations in this block group.  We do  in btrfs_delete_unused_bgs()
1340 			 * this block group.  in btrfs_delete_unused_bgs()
1343 			spin_unlock(&block_group->lock);  in btrfs_delete_unused_bgs()
1344 			up_write(&space_info->groups_sem);  in btrfs_delete_unused_bgs()
1347 		spin_unlock(&block_group->lock);  in btrfs_delete_unused_bgs()
1351 		up_write(&space_info->groups_sem);  in btrfs_delete_unused_bgs()
1362 						     block_group->start);  in btrfs_delete_unused_bgs()
1370 		 * We could have pending pinned extents for this block group,  in btrfs_delete_unused_bgs()
1385 		spin_lock(&fs_info->discard_ctl.lock);  in btrfs_delete_unused_bgs()
1386 		if (!list_empty(&block_group->discard_list)) {  in btrfs_delete_unused_bgs()
1387 			spin_unlock(&fs_info->discard_ctl.lock);  in btrfs_delete_unused_bgs()
1389 			btrfs_discard_queue_work(&fs_info->discard_ctl,  in btrfs_delete_unused_bgs()
1393 		spin_unlock(&fs_info->discard_ctl.lock);  in btrfs_delete_unused_bgs()
1396 		spin_lock(&space_info->lock);  in btrfs_delete_unused_bgs()
1397 		spin_lock(&block_group->lock);  in btrfs_delete_unused_bgs()
1400 						     -block_group->pinned);  in btrfs_delete_unused_bgs()
1401 		space_info->bytes_readonly += block_group->pinned;  in btrfs_delete_unused_bgs()
1402 		block_group->pinned = 0;  in btrfs_delete_unused_bgs()
1404 		spin_unlock(&block_group->lock);  in btrfs_delete_unused_bgs()
1405 		spin_unlock(&space_info->lock);  in btrfs_delete_unused_bgs()
1408 		 * The normal path here is an unused block group is passed here,  in btrfs_delete_unused_bgs()
1411 		 * before coming down the unused block group path as trimming  in btrfs_delete_unused_bgs()
1419 		 * need to reset sequential-required zones.  in btrfs_delete_unused_bgs()
1432 		ret = btrfs_remove_chunk(trans, block_group->start);  in btrfs_delete_unused_bgs()
1441 		 * If we're not mounted with -odiscard, we can just forget  in btrfs_delete_unused_bgs()
1442 		 * about this block group. Otherwise we'll need to wait  in btrfs_delete_unused_bgs()
1446 			spin_lock(&fs_info->unused_bgs_lock);  in btrfs_delete_unused_bgs()
1449 			 * fs_info->unused_bgs, so use a list_move operation  in btrfs_delete_unused_bgs()
1450 			 * to add the block group to the deleted_bgs list.  in btrfs_delete_unused_bgs()
1452 			list_move(&block_group->bg_list,  in btrfs_delete_unused_bgs()
1453 				  &trans->transaction->deleted_bgs);  in btrfs_delete_unused_bgs()
1454 			spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_delete_unused_bgs()
1461 		spin_lock(&fs_info->unused_bgs_lock);  in btrfs_delete_unused_bgs()
1463 	spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_delete_unused_bgs()
1464 	mutex_unlock(&fs_info->reclaim_bgs_lock);  in btrfs_delete_unused_bgs()
1469 	mutex_unlock(&fs_info->reclaim_bgs_lock);  in btrfs_delete_unused_bgs()
1476 	struct btrfs_fs_info *fs_info = bg->fs_info;  in btrfs_mark_bg_unused()
1478 	spin_lock(&fs_info->unused_bgs_lock);  in btrfs_mark_bg_unused()
1479 	if (list_empty(&bg->bg_list)) {  in btrfs_mark_bg_unused()
1482 		list_add_tail(&bg->bg_list, &fs_info->unused_bgs);  in btrfs_mark_bg_unused()
1484 	spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_mark_bg_unused()
1495 	if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))  in btrfs_reclaim_bgs_work()
1505 	if (!mutex_trylock(&fs_info->reclaim_bgs_lock)) {  in btrfs_reclaim_bgs_work()
1510 	spin_lock(&fs_info->unused_bgs_lock);  in btrfs_reclaim_bgs_work()
1511 	while (!list_empty(&fs_info->reclaim_bgs)) {  in btrfs_reclaim_bgs_work()
1515 		bg = list_first_entry(&fs_info->reclaim_bgs,  in btrfs_reclaim_bgs_work()
1518 		list_del_init(&bg->bg_list);  in btrfs_reclaim_bgs_work()
1520 		space_info = bg->space_info;  in btrfs_reclaim_bgs_work()
1521 		spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_reclaim_bgs_work()
1524 		down_write(&space_info->groups_sem);  in btrfs_reclaim_bgs_work()
1526 		spin_lock(&bg->lock);  in btrfs_reclaim_bgs_work()
1527 		if (bg->reserved || bg->pinned || bg->ro) {  in btrfs_reclaim_bgs_work()
1530 			 * outstanding allocations in this block group.  We do  in btrfs_reclaim_bgs_work()
1532 			 * this block group.  in btrfs_reclaim_bgs_work()
1534 			spin_unlock(&bg->lock);  in btrfs_reclaim_bgs_work()
1535 			up_write(&space_info->groups_sem);  in btrfs_reclaim_bgs_work()
1538 		spin_unlock(&bg->lock);  in btrfs_reclaim_bgs_work()
1542 			up_write(&space_info->groups_sem);  in btrfs_reclaim_bgs_work()
1547 		 * Cache the zone_unusable value before turning the block group  in btrfs_reclaim_bgs_work()
1549 		 * zone_unusable value gets moved to the block group's read-only  in btrfs_reclaim_bgs_work()
1552 		zone_unusable = bg->zone_unusable;  in btrfs_reclaim_bgs_work()
1554 		up_write(&space_info->groups_sem);  in btrfs_reclaim_bgs_work()
1560 				bg->start, div_u64(bg->used * 100, bg->length),  in btrfs_reclaim_bgs_work()
1561 				div64_u64(zone_unusable * 100, bg->length));  in btrfs_reclaim_bgs_work()
1563 		ret = btrfs_relocate_chunk(fs_info, bg->start);  in btrfs_reclaim_bgs_work()
1564 		if (ret && ret != -EAGAIN)  in btrfs_reclaim_bgs_work()
1566 				  bg->start);  in btrfs_reclaim_bgs_work()
1569 		spin_lock(&fs_info->unused_bgs_lock);  in btrfs_reclaim_bgs_work()
1570 		if (ret == -EAGAIN && list_empty(&bg->bg_list))  in btrfs_reclaim_bgs_work()
1571 			list_add_tail(&bg->bg_list, &again_list);  in btrfs_reclaim_bgs_work()
1575 	list_splice_tail(&again_list, &fs_info->reclaim_bgs);  in btrfs_reclaim_bgs_work()
1576 	spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_reclaim_bgs_work()
1577 	mutex_unlock(&fs_info->reclaim_bgs_lock);  in btrfs_reclaim_bgs_work()
1583 	spin_lock(&fs_info->unused_bgs_lock);  in btrfs_reclaim_bgs()
1584 	if (!list_empty(&fs_info->reclaim_bgs))  in btrfs_reclaim_bgs()
1585 		queue_work(system_unbound_wq, &fs_info->reclaim_bgs_work);  in btrfs_reclaim_bgs()
1586 	spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_reclaim_bgs()
1591 	struct btrfs_fs_info *fs_info = bg->fs_info;  in btrfs_mark_bg_to_reclaim()
1593 	spin_lock(&fs_info->unused_bgs_lock);  in btrfs_mark_bg_to_reclaim()
1594 	if (list_empty(&bg->bg_list)) {  in btrfs_mark_bg_to_reclaim()
1597 		list_add_tail(&bg->bg_list, &fs_info->reclaim_bgs);  in btrfs_mark_bg_to_reclaim()
1599 	spin_unlock(&fs_info->unused_bgs_lock);  in btrfs_mark_bg_to_reclaim()
1613 	slot = path->slots[0];  in read_bg_from_eb()
1614 	leaf = path->nodes[0];  in read_bg_from_eb()
1616 	em_tree = &fs_info->mapping_tree;  in read_bg_from_eb()
1617 	read_lock(&em_tree->lock);  in read_bg_from_eb()
1618 	em = lookup_extent_mapping(em_tree, key->objectid, key->offset);  in read_bg_from_eb()
1619 	read_unlock(&em_tree->lock);  in read_bg_from_eb()
1623 			  key->objectid, key->offset);  in read_bg_from_eb()
1624 		return -ENOENT;  in read_bg_from_eb()
1627 	if (em->start != key->objectid || em->len != key->offset) {  in read_bg_from_eb()
1629 			"block group %llu len %llu mismatch with chunk %llu len %llu",  in read_bg_from_eb()
1630 			key->objectid, key->offset, em->start, em->len);  in read_bg_from_eb()
1631 		ret = -EUCLEAN;  in read_bg_from_eb()
1640 	if (flags != (em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK)) {  in read_bg_from_eb()
1642 "block group %llu len %llu type flags 0x%llx mismatch with chunk type flags 0x%llx",  in read_bg_from_eb()
1643 			  key->objectid, key->offset, flags,  in read_bg_from_eb()
1644 			  (BTRFS_BLOCK_GROUP_TYPE_MASK & em->map_lookup->type));  in read_bg_from_eb()
1645 		ret = -EUCLEAN;  in read_bg_from_eb()
1657 	struct btrfs_root *root = fs_info->extent_root;  in find_first_block_group()
1668 		slot = path->slots[0];  in find_first_block_group()
1669 		leaf = path->nodes[0];  in find_first_block_group()
1680 		if (found_key.objectid >= key->objectid &&  in find_first_block_group()
1686 		path->slots[0]++;  in find_first_block_group()
1697 	write_seqlock(&fs_info->profiles_lock);  in set_avail_alloc_bits()
1699 		fs_info->avail_data_alloc_bits |= extra_flags;  in set_avail_alloc_bits()
1701 		fs_info->avail_metadata_alloc_bits |= extra_flags;  in set_avail_alloc_bits()
1703 		fs_info->avail_system_alloc_bits |= extra_flags;  in set_avail_alloc_bits()
1704 	write_sequnlock(&fs_info->profiles_lock);  in set_avail_alloc_bits()
1711  * @chunk_start:   logical address of block group
1716  * @stripe_len:    size of IO stripe for the given block group
1719  * Used primarily to exclude those portions of a block group that contain super
1720  * block copies.
1737 		return -EIO;  in btrfs_rmap_block()
1739 	map = em->map_lookup;  in btrfs_rmap_block()
1740 	data_stripe_length = em->orig_block_len;  in btrfs_rmap_block()
1741 	io_stripe_size = map->stripe_len;  in btrfs_rmap_block()
1742 	chunk_start = em->start;  in btrfs_rmap_block()
1745 	if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK)  in btrfs_rmap_block()
1746 		io_stripe_size = map->stripe_len * nr_data_stripes(map);  in btrfs_rmap_block()
1748 	buf = kcalloc(map->num_stripes, sizeof(u64), GFP_NOFS);  in btrfs_rmap_block()
1750 		ret = -ENOMEM;  in btrfs_rmap_block()
1754 	for (i = 0; i < map->num_stripes; i++) {  in btrfs_rmap_block()
1760 		if (!in_range(physical, map->stripes[i].physical,  in btrfs_rmap_block()
1764 		if (bdev && map->stripes[i].dev->bdev != bdev)  in btrfs_rmap_block()
1767 		stripe_nr = physical - map->stripes[i].physical;  in btrfs_rmap_block()
1768 		stripe_nr = div64_u64_rem(stripe_nr, map->stripe_len, &offset);  in btrfs_rmap_block()
1770 		if (map->type & BTRFS_BLOCK_GROUP_RAID10) {  in btrfs_rmap_block()
1771 			stripe_nr = stripe_nr * map->num_stripes + i;  in btrfs_rmap_block()
1772 			stripe_nr = div_u64(stripe_nr, map->sub_stripes);  in btrfs_rmap_block()
1773 		} else if (map->type & BTRFS_BLOCK_GROUP_RAID0) {  in btrfs_rmap_block()
1774 			stripe_nr = stripe_nr * map->num_stripes + i;  in btrfs_rmap_block()
1779 		 * instead of map->stripe_len  in btrfs_rmap_block()
1804 static int exclude_super_stripes(struct btrfs_block_group *cache)  in exclude_super_stripes()  argument
1806 	struct btrfs_fs_info *fs_info = cache->fs_info;  in exclude_super_stripes()
1813 	if (cache->start < BTRFS_SUPER_INFO_OFFSET) {  in exclude_super_stripes()
1814 		stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->start;  in exclude_super_stripes()
1815 		cache->bytes_super += stripe_len;  in exclude_super_stripes()
1816 		ret = btrfs_add_excluded_extent(fs_info, cache->start,  in exclude_super_stripes()
1824 		ret = btrfs_rmap_block(fs_info, cache->start, NULL,  in exclude_super_stripes()
1832 			"zoned: block group %llu must not contain super block",  in exclude_super_stripes()
1833 				  cache->start);  in exclude_super_stripes()
1834 			return -EUCLEAN;  in exclude_super_stripes()
1837 		while (nr--) {  in exclude_super_stripes()
1839 				cache->start + cache->length - logical[nr]);  in exclude_super_stripes()
1841 			cache->bytes_super += len;  in exclude_super_stripes()
1855 static void link_block_group(struct btrfs_block_group *cache)  in link_block_group()  argument
1857 	struct btrfs_space_info *space_info = cache->space_info;  in link_block_group()
1858 	int index = btrfs_bg_flags_to_raid_index(cache->flags);  in link_block_group()
1860 	down_write(&space_info->groups_sem);  in link_block_group()
1861 	list_add_tail(&cache->list, &space_info->block_groups[index]);  in link_block_group()
1862 	up_write(&space_info->groups_sem);  in link_block_group()
1868 	struct btrfs_block_group *cache;  in btrfs_create_block_group_cache()  local
1870 	cache = kzalloc(sizeof(*cache), GFP_NOFS);  in btrfs_create_block_group_cache()
1871 	if (!cache)  in btrfs_create_block_group_cache()
1874 	cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),  in btrfs_create_block_group_cache()
1876 	if (!cache->free_space_ctl) {  in btrfs_create_block_group_cache()
1877 		kfree(cache);  in btrfs_create_block_group_cache()
1881 	cache->start = start;  in btrfs_create_block_group_cache()
1883 	cache->fs_info = fs_info;  in btrfs_create_block_group_cache()
1884 	cache->full_stripe_len = btrfs_full_stripe_len(fs_info, start);  in btrfs_create_block_group_cache()
1886 	cache->discard_index = BTRFS_DISCARD_INDEX_UNUSED;  in btrfs_create_block_group_cache()
1888 	refcount_set(&cache->refs, 1);  in btrfs_create_block_group_cache()
1889 	spin_lock_init(&cache->lock);  in btrfs_create_block_group_cache()
1890 	init_rwsem(&cache->data_rwsem);  in btrfs_create_block_group_cache()
1891 	INIT_LIST_HEAD(&cache->list);  in btrfs_create_block_group_cache()
1892 	INIT_LIST_HEAD(&cache->cluster_list);  in btrfs_create_block_group_cache()
1893 	INIT_LIST_HEAD(&cache->bg_list);  in btrfs_create_block_group_cache()
1894 	INIT_LIST_HEAD(&cache->ro_list);  in btrfs_create_block_group_cache()
1895 	INIT_LIST_HEAD(&cache->discard_list);  in btrfs_create_block_group_cache()
1896 	INIT_LIST_HEAD(&cache->dirty_list);  in btrfs_create_block_group_cache()
1897 	INIT_LIST_HEAD(&cache->io_list);  in btrfs_create_block_group_cache()
1898 	btrfs_init_free_space_ctl(cache, cache->free_space_ctl);  in btrfs_create_block_group_cache()
1899 	atomic_set(&cache->frozen, 0);  in btrfs_create_block_group_cache()
1900 	mutex_init(&cache->free_space_lock);  in btrfs_create_block_group_cache()
1901 	btrfs_init_full_stripe_locks_tree(&cache->full_stripe_locks_root);  in btrfs_create_block_group_cache()
1903 	return cache;  in btrfs_create_block_group_cache()
1907  * Iterate all chunks and verify that each of them has the corresponding block
1912 	struct extent_map_tree *map_tree = &fs_info->mapping_tree;  in check_chunk_block_group_mappings()
1919 		read_lock(&map_tree->lock);  in check_chunk_block_group_mappings()
1926 		read_unlock(&map_tree->lock);  in check_chunk_block_group_mappings()
1930 		bg = btrfs_lookup_block_group(fs_info, em->start);  in check_chunk_block_group_mappings()
1933 	"chunk start=%llu len=%llu doesn't have corresponding block group",  in check_chunk_block_group_mappings()
1934 				     em->start, em->len);  in check_chunk_block_group_mappings()
1935 			ret = -EUCLEAN;  in check_chunk_block_group_mappings()
1939 		if (bg->start != em->start || bg->length != em->len ||  in check_chunk_block_group_mappings()
1940 		    (bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK) !=  in check_chunk_block_group_mappings()
1941 		    (em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK)) {  in check_chunk_block_group_mappings()
1943 "chunk start=%llu len=%llu flags=0x%llx doesn't match block group start=%llu len=%llu flags=0x%llx",  in check_chunk_block_group_mappings()
1944 				em->start, em->len,  in check_chunk_block_group_mappings()
1945 				em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK,  in check_chunk_block_group_mappings()
1946 				bg->start, bg->length,  in check_chunk_block_group_mappings()
1947 				bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK);  in check_chunk_block_group_mappings()
1948 			ret = -EUCLEAN;  in check_chunk_block_group_mappings()
1953 		start = em->start + em->len;  in check_chunk_block_group_mappings()
1965 	struct btrfs_block_group *cache;  in read_one_block_group()  local
1970 	ASSERT(key->type == BTRFS_BLOCK_GROUP_ITEM_KEY);  in read_one_block_group()
1972 	cache = btrfs_create_block_group_cache(info, key->objectid);  in read_one_block_group()
1973 	if (!cache)  in read_one_block_group()
1974 		return -ENOMEM;  in read_one_block_group()
1976 	cache->length = key->offset;  in read_one_block_group()
1977 	cache->used = btrfs_stack_block_group_used(bgi);  in read_one_block_group()
1978 	cache->flags = btrfs_stack_block_group_flags(bgi);  in read_one_block_group()
1980 	set_free_space_tree_thresholds(cache);  in read_one_block_group()
1984 		 * When we mount with old space cache, we need to  in read_one_block_group()
1988 		 *    truncate the old free space cache inode and  in read_one_block_group()
1991 		 *    the new space cache info onto disk.  in read_one_block_group()
1994 			cache->disk_cache_state = BTRFS_DC_CLEAR;  in read_one_block_group()
1996 	if (!mixed && ((cache->flags & BTRFS_BLOCK_GROUP_METADATA) &&  in read_one_block_group()
1997 	    (cache->flags & BTRFS_BLOCK_GROUP_DATA))) {  in read_one_block_group()
1999 "bg %llu is a mixed block group but filesystem hasn't enabled mixed block groups",  in read_one_block_group()
2000 				  cache->start);  in read_one_block_group()
2001 			ret = -EINVAL;  in read_one_block_group()
2005 	ret = btrfs_load_block_group_zone_info(cache, false);  in read_one_block_group()
2008 			  cache->start);  in read_one_block_group()
2017 	ret = exclude_super_stripes(cache);  in read_one_block_group()
2020 		btrfs_free_excluded_extents(cache);  in read_one_block_group()
2026 	 * free space for a block group. So, we don't need any caching work.  in read_one_block_group()
2037 		btrfs_calc_zone_unusable(cache);  in read_one_block_group()
2038 	} else if (cache->length == cache->used) {  in read_one_block_group()
2039 		cache->last_byte_to_unpin = (u64)-1;  in read_one_block_group()
2040 		cache->cached = BTRFS_CACHE_FINISHED;  in read_one_block_group()
2041 		btrfs_free_excluded_extents(cache);  in read_one_block_group()
2042 	} else if (cache->used == 0) {  in read_one_block_group()
2043 		cache->last_byte_to_unpin = (u64)-1;  in read_one_block_group()
2044 		cache->cached = BTRFS_CACHE_FINISHED;  in read_one_block_group()
2045 		add_new_free_space(cache, cache->start,  in read_one_block_group()
2046 				   cache->start + cache->length);  in read_one_block_group()
2047 		btrfs_free_excluded_extents(cache);  in read_one_block_group()
2050 	ret = btrfs_add_block_group_cache(info, cache);  in read_one_block_group()
2052 		btrfs_remove_free_space_cache(cache);  in read_one_block_group()
2055 	trace_btrfs_add_block_group(info, cache, 0);  in read_one_block_group()
2056 	btrfs_update_space_info(info, cache->flags, cache->length,  in read_one_block_group()
2057 				cache->used, cache->bytes_super,  in read_one_block_group()
2058 				cache->zone_unusable, &space_info);  in read_one_block_group()
2060 	cache->space_info = space_info;  in read_one_block_group()
2062 	link_block_group(cache);  in read_one_block_group()
2064 	set_avail_alloc_bits(info, cache->flags);  in read_one_block_group()
2065 	if (btrfs_chunk_readonly(info, cache->start)) {  in read_one_block_group()
2066 		inc_block_group_ro(cache, 1);  in read_one_block_group()
2067 	} else if (cache->used == 0) {  in read_one_block_group()
2068 		ASSERT(list_empty(&cache->bg_list));  in read_one_block_group()
2070 			btrfs_discard_queue_work(&info->discard_ctl, cache);  in read_one_block_group()
2072 			btrfs_mark_bg_unused(cache);  in read_one_block_group()
2076 	btrfs_put_block_group(cache);  in read_one_block_group()
2082 	struct extent_map_tree *em_tree = &fs_info->mapping_tree;  in fill_dummy_bgs()
2087 	for (node = rb_first_cached(&em_tree->map); node; node = rb_next(node)) {  in fill_dummy_bgs()
2093 		map = em->map_lookup;  in fill_dummy_bgs()
2094 		bg = btrfs_create_block_group_cache(fs_info, em->start);  in fill_dummy_bgs()
2096 			ret = -ENOMEM;  in fill_dummy_bgs()
2100 		/* Fill dummy cache as FULL */  in fill_dummy_bgs()
2101 		bg->length = em->len;  in fill_dummy_bgs()
2102 		bg->flags = map->type;  in fill_dummy_bgs()
2103 		bg->last_byte_to_unpin = (u64)-1;  in fill_dummy_bgs()
2104 		bg->cached = BTRFS_CACHE_FINISHED;  in fill_dummy_bgs()
2105 		bg->used = em->len;  in fill_dummy_bgs()
2106 		bg->flags = map->type;  in fill_dummy_bgs()
2109 		 * We may have some valid block group cache added already, in  in fill_dummy_bgs()
2112 		if (ret == -EEXIST) {  in fill_dummy_bgs()
2124 		btrfs_update_space_info(fs_info, bg->flags, em->len, em->len,  in fill_dummy_bgs()
2126 		bg->space_info = space_info;  in fill_dummy_bgs()
2129 		set_avail_alloc_bits(fs_info, bg->flags);  in fill_dummy_bgs()
2140 	struct btrfs_block_group *cache;  in btrfs_read_block_groups()  local
2146 	if (!info->extent_root)  in btrfs_read_block_groups()
2154 		return -ENOMEM;  in btrfs_read_block_groups()
2156 	cache_gen = btrfs_super_cache_generation(info->super_copy);  in btrfs_read_block_groups()
2158 	    btrfs_super_generation(info->super_copy) != cache_gen)  in btrfs_read_block_groups()
2174 		leaf = path->nodes[0];  in btrfs_read_block_groups()
2175 		slot = path->slots[0];  in btrfs_read_block_groups()
2190 	list_for_each_entry(space_info, &info->space_info, list) {  in btrfs_read_block_groups()
2194 			if (list_empty(&space_info->block_groups[i]))  in btrfs_read_block_groups()
2196 			cache = list_first_entry(&space_info->block_groups[i],  in btrfs_read_block_groups()
2199 			btrfs_sysfs_add_block_group_type(cache);  in btrfs_read_block_groups()
2202 		if (!(btrfs_get_alloc_profile(info, space_info->flags) &  in btrfs_read_block_groups()
2209 		 * Avoid allocating from un-mirrored block group if there are  in btrfs_read_block_groups()
2210 		 * mirrored block groups.  in btrfs_read_block_groups()
2212 		list_for_each_entry(cache,  in btrfs_read_block_groups()
2213 				&space_info->block_groups[BTRFS_RAID_RAID0],  in btrfs_read_block_groups()
2215 			inc_block_group_ro(cache, 1);  in btrfs_read_block_groups()
2216 		list_for_each_entry(cache,  in btrfs_read_block_groups()
2217 				&space_info->block_groups[BTRFS_RAID_SINGLE],  in btrfs_read_block_groups()
2219 			inc_block_group_ro(cache, 1);  in btrfs_read_block_groups()
2229 	 * Try to fill the tree using dummy block groups so that the user can  in btrfs_read_block_groups()
2247 	struct btrfs_fs_info *fs_info = trans->fs_info;  in insert_block_group_item()
2252 	spin_lock(&block_group->lock);  in insert_block_group_item()
2253 	btrfs_set_stack_block_group_used(&bgi, block_group->used);  in insert_block_group_item()
2256 	btrfs_set_stack_block_group_flags(&bgi, block_group->flags);  in insert_block_group_item()
2257 	key.objectid = block_group->start;  in insert_block_group_item()
2259 	key.offset = block_group->length;  in insert_block_group_item()
2260 	spin_unlock(&block_group->lock);  in insert_block_group_item()
2262 	root = fs_info->extent_root;  in insert_block_group_item()
2270 	struct btrfs_fs_info *fs_info = device->fs_info;  in insert_dev_extent()
2271 	struct btrfs_root *root = fs_info->dev_root;  in insert_dev_extent()
2278 	WARN_ON(!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &device->dev_state));  in insert_dev_extent()
2279 	WARN_ON(test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state));  in insert_dev_extent()
2282 		return -ENOMEM;  in insert_dev_extent()
2284 	key.objectid = device->devid;  in insert_dev_extent()
2291 	leaf = path->nodes[0];  in insert_dev_extent()
2292 	extent = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_extent);  in insert_dev_extent()
2314 	struct btrfs_fs_info *fs_info = trans->fs_info;  in insert_dev_extents()
2327 	map = em->map_lookup;  in insert_dev_extents()
2328 	stripe_size = em->orig_block_len;  in insert_dev_extents()
2335 	 * (dev-replace.c:btrfs_dev_replace_finishing()), so we could grab the  in insert_dev_extents()
2339 	mutex_lock(&fs_info->fs_devices->device_list_mutex);  in insert_dev_extents()
2340 	for (i = 0; i < map->num_stripes; i++) {  in insert_dev_extents()
2341 		device = map->stripes[i].dev;  in insert_dev_extents()
2342 		dev_offset = map->stripes[i].physical;  in insert_dev_extents()
2349 	mutex_unlock(&fs_info->fs_devices->device_list_mutex);  in insert_dev_extents()
2364 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_create_pending_block_groups()
2368 	while (!list_empty(&trans->new_bgs)) {  in btrfs_create_pending_block_groups()
2371 		block_group = list_first_entry(&trans->new_bgs,  in btrfs_create_pending_block_groups()
2377 		index = btrfs_bg_flags_to_raid_index(block_group->flags);  in btrfs_create_pending_block_groups()
2382 		if (!block_group->chunk_item_inserted) {  in btrfs_create_pending_block_groups()
2383 			mutex_lock(&fs_info->chunk_mutex);  in btrfs_create_pending_block_groups()
2385 			mutex_unlock(&fs_info->chunk_mutex);  in btrfs_create_pending_block_groups()
2389 		ret = insert_dev_extents(trans, block_group->start,  in btrfs_create_pending_block_groups()
2390 					 block_group->length);  in btrfs_create_pending_block_groups()
2401 		if (block_group->space_info->block_group_kobjs[index] == NULL)  in btrfs_create_pending_block_groups()
2407 		list_del_init(&block_group->bg_list);  in btrfs_create_pending_block_groups()
2414 						 u64 chunk_offset, u64 size)  in btrfs_make_block_group()  argument
2416 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_make_block_group()
2417 	struct btrfs_block_group *cache;  in btrfs_make_block_group()  local
2422 	cache = btrfs_create_block_group_cache(fs_info, chunk_offset);  in btrfs_make_block_group()
2423 	if (!cache)  in btrfs_make_block_group()
2424 		return ERR_PTR(-ENOMEM);  in btrfs_make_block_group()
2426 	cache->length = size;  in btrfs_make_block_group()
2427 	set_free_space_tree_thresholds(cache);  in btrfs_make_block_group()
2428 	cache->used = bytes_used;  in btrfs_make_block_group()
2429 	cache->flags = type;  in btrfs_make_block_group()
2430 	cache->last_byte_to_unpin = (u64)-1;  in btrfs_make_block_group()
2431 	cache->cached = BTRFS_CACHE_FINISHED;  in btrfs_make_block_group()
2433 		cache->needs_free_space = 1;  in btrfs_make_block_group()
2435 	ret = btrfs_load_block_group_zone_info(cache, true);  in btrfs_make_block_group()
2437 		btrfs_put_block_group(cache);  in btrfs_make_block_group()
2441 	ret = exclude_super_stripes(cache);  in btrfs_make_block_group()
2444 		btrfs_free_excluded_extents(cache);  in btrfs_make_block_group()
2445 		btrfs_put_block_group(cache);  in btrfs_make_block_group()
2449 	add_new_free_space(cache, chunk_offset, chunk_offset + size);  in btrfs_make_block_group()
2451 	btrfs_free_excluded_extents(cache);  in btrfs_make_block_group()
2454 	if (btrfs_should_fragment_free_space(cache)) {  in btrfs_make_block_group()
2455 		u64 new_bytes_used = size - bytes_used;  in btrfs_make_block_group()
2458 		fragment_free_space(cache);  in btrfs_make_block_group()
2463 	 * assigned to our block group. We want our bg to be added to the rbtree  in btrfs_make_block_group()
2464 	 * with its ->space_info set.  in btrfs_make_block_group()
2466 	cache->space_info = btrfs_find_space_info(fs_info, cache->flags);  in btrfs_make_block_group()
2467 	ASSERT(cache->space_info);  in btrfs_make_block_group()
2469 	ret = btrfs_add_block_group_cache(fs_info, cache);  in btrfs_make_block_group()
2471 		btrfs_remove_free_space_cache(cache);  in btrfs_make_block_group()
2472 		btrfs_put_block_group(cache);  in btrfs_make_block_group()
2477 	 * Now that our block group has its ->space_info set and is inserted in  in btrfs_make_block_group()
2480 	trace_btrfs_add_block_group(fs_info, cache, 1);  in btrfs_make_block_group()
2481 	btrfs_update_space_info(fs_info, cache->flags, size, bytes_used,  in btrfs_make_block_group()
2482 				cache->bytes_super, 0, &cache->space_info);  in btrfs_make_block_group()
2485 	link_block_group(cache);  in btrfs_make_block_group()
2487 	list_add_tail(&cache->bg_list, &trans->new_bgs);  in btrfs_make_block_group()
2488 	trans->delayed_ref_updates++;  in btrfs_make_block_group()
2492 	return cache;  in btrfs_make_block_group()
2496  * Mark one block group RO, can be called several times for the same block
2499  * @cache:		the destination block group
2500  * @do_chunk_alloc:	whether need to do chunk pre-allocation, this is to
2502  * 			block group RO.
2504 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,  in btrfs_inc_block_group_ro()  argument
2507 	struct btrfs_fs_info *fs_info = cache->fs_info;  in btrfs_inc_block_group_ro()
2514 		trans = btrfs_join_transaction(fs_info->extent_root);  in btrfs_inc_block_group_ro()
2521 		 * We're not allowed to set block groups readonly after the dirty  in btrfs_inc_block_group_ro()
2522 		 * block group cache has started writing.  If it already started,  in btrfs_inc_block_group_ro()
2525 		mutex_lock(&fs_info->ro_block_group_mutex);  in btrfs_inc_block_group_ro()
2526 		if (test_bit(BTRFS_TRANS_DIRTY_BG_RUN, &trans->transaction->flags)) {  in btrfs_inc_block_group_ro()
2527 			u64 transid = trans->transid;  in btrfs_inc_block_group_ro()
2529 			mutex_unlock(&fs_info->ro_block_group_mutex);  in btrfs_inc_block_group_ro()
2542 		 * corresponding block group with the new raid level.  in btrfs_inc_block_group_ro()
2544 		alloc_flags = btrfs_get_alloc_profile(fs_info, cache->flags);  in btrfs_inc_block_group_ro()
2545 		if (alloc_flags != cache->flags) {  in btrfs_inc_block_group_ro()
2552 			if (ret == -ENOSPC)  in btrfs_inc_block_group_ro()
2559 	ret = inc_block_group_ro(cache, 0);  in btrfs_inc_block_group_ro()
2560 	if (!do_chunk_alloc || ret == -ETXTBSY)  in btrfs_inc_block_group_ro()
2564 	alloc_flags = btrfs_get_alloc_profile(fs_info, cache->space_info->flags);  in btrfs_inc_block_group_ro()
2568 	ret = inc_block_group_ro(cache, 0);  in btrfs_inc_block_group_ro()
2569 	if (ret == -ETXTBSY)  in btrfs_inc_block_group_ro()
2572 	if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {  in btrfs_inc_block_group_ro()
2573 		alloc_flags = btrfs_get_alloc_profile(fs_info, cache->flags);  in btrfs_inc_block_group_ro()
2574 		mutex_lock(&fs_info->chunk_mutex);  in btrfs_inc_block_group_ro()
2576 		mutex_unlock(&fs_info->chunk_mutex);  in btrfs_inc_block_group_ro()
2579 	mutex_unlock(&fs_info->ro_block_group_mutex);  in btrfs_inc_block_group_ro()
2585 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache)  in btrfs_dec_block_group_ro()  argument
2587 	struct btrfs_space_info *sinfo = cache->space_info;  in btrfs_dec_block_group_ro()
2590 	BUG_ON(!cache->ro);  in btrfs_dec_block_group_ro()
2592 	spin_lock(&sinfo->lock);  in btrfs_dec_block_group_ro()
2593 	spin_lock(&cache->lock);  in btrfs_dec_block_group_ro()
2594 	if (!--cache->ro) {  in btrfs_dec_block_group_ro()
2595 		if (btrfs_is_zoned(cache->fs_info)) {  in btrfs_dec_block_group_ro()
2597 			cache->zone_unusable = cache->alloc_offset - cache->used;  in btrfs_dec_block_group_ro()
2598 			sinfo->bytes_zone_unusable += cache->zone_unusable;  in btrfs_dec_block_group_ro()
2599 			sinfo->bytes_readonly -= cache->zone_unusable;  in btrfs_dec_block_group_ro()
2601 		num_bytes = cache->length - cache->reserved -  in btrfs_dec_block_group_ro()
2602 			    cache->pinned - cache->bytes_super -  in btrfs_dec_block_group_ro()
2603 			    cache->zone_unusable - cache->used;  in btrfs_dec_block_group_ro()
2604 		sinfo->bytes_readonly -= num_bytes;  in btrfs_dec_block_group_ro()
2605 		list_del_init(&cache->ro_list);  in btrfs_dec_block_group_ro()
2607 	spin_unlock(&cache->lock);  in btrfs_dec_block_group_ro()
2608 	spin_unlock(&sinfo->lock);  in btrfs_dec_block_group_ro()
2613 				   struct btrfs_block_group *cache)  in update_block_group_item()  argument
2615 	struct btrfs_fs_info *fs_info = trans->fs_info;  in update_block_group_item()
2617 	struct btrfs_root *root = fs_info->extent_root;  in update_block_group_item()
2623 	key.objectid = cache->start;  in update_block_group_item()
2625 	key.offset = cache->length;  in update_block_group_item()
2630 			ret = -ENOENT;  in update_block_group_item()
2634 	leaf = path->nodes[0];  in update_block_group_item()
2635 	bi = btrfs_item_ptr_offset(leaf, path->slots[0]);  in update_block_group_item()
2636 	btrfs_set_stack_block_group_used(&bgi, cache->used);  in update_block_group_item()
2639 	btrfs_set_stack_block_group_flags(&bgi, cache->flags);  in update_block_group_item()
2652 	struct btrfs_fs_info *fs_info = block_group->fs_info;  in cache_save_setup()
2653 	struct btrfs_root *root = fs_info->tree_root;  in cache_save_setup()
2666 	 * If this block group is smaller than 100 megs don't bother caching the  in cache_save_setup()
2667 	 * block group.  in cache_save_setup()
2669 	if (block_group->length < (100 * SZ_1M)) {  in cache_save_setup()
2670 		spin_lock(&block_group->lock);  in cache_save_setup()
2671 		block_group->disk_cache_state = BTRFS_DC_WRITTEN;  in cache_save_setup()
2672 		spin_unlock(&block_group->lock);  in cache_save_setup()
2680 	if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {  in cache_save_setup()
2690 		if (block_group->ro)  in cache_save_setup()
2701 	 * from here on out we know not to trust this cache when we load up next  in cache_save_setup()
2704 	BTRFS_I(inode)->generation = 0;  in cache_save_setup()
2709 		 * super cache generation to 0 so we know to invalidate the  in cache_save_setup()
2710 		 * cache, but then we'd have to keep track of the block groups  in cache_save_setup()
2711 		 * that fail this way so we know we _have_ to reset this cache  in cache_save_setup()
2712 		 * before the next commit or risk reading stale cache.  So to  in cache_save_setup()
2723 	if (block_group->cache_generation == trans->transid &&  in cache_save_setup()
2731 					&fs_info->global_block_rsv);  in cache_save_setup()
2740 	spin_lock(&block_group->lock);  in cache_save_setup()
2741 	if (block_group->cached != BTRFS_CACHE_FINISHED ||  in cache_save_setup()
2750 		spin_unlock(&block_group->lock);  in cache_save_setup()
2753 	spin_unlock(&block_group->lock);  in cache_save_setup()
2756 	 * We hit an ENOSPC when setting up the cache in this transaction, just  in cache_save_setup()
2757 	 * skip doing the setup, we've already cleared the cache so we're safe.  in cache_save_setup()
2759 	if (test_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags)) {  in cache_save_setup()
2760 		ret = -ENOSPC;  in cache_save_setup()
2765 	 * Try to preallocate enough space based on how big the block group is.  in cache_save_setup()
2768 	 * cache.  in cache_save_setup()
2770 	cache_size = div_u64(block_group->length, SZ_256M);  in cache_save_setup()
2775 	cache_size *= fs_info->sectorsize;  in cache_save_setup()
2786 	 * Our cache requires contiguous chunks so that we don't modify a bunch  in cache_save_setup()
2787 	 * of metadata or split extents when writing the cache out, which means  in cache_save_setup()
2790 	 * other block groups for this transaction, maybe we'll unpin enough  in cache_save_setup()
2795 	else if (ret == -ENOSPC)  in cache_save_setup()
2796 		set_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags);  in cache_save_setup()
2803 	spin_lock(&block_group->lock);  in cache_save_setup()
2805 		block_group->cache_generation = trans->transid;  in cache_save_setup()
2806 	block_group->disk_cache_state = dcs;  in cache_save_setup()
2807 	spin_unlock(&block_group->lock);  in cache_save_setup()
2815 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_setup_space_cache()
2816 	struct btrfs_block_group *cache, *tmp;  in btrfs_setup_space_cache()  local
2817 	struct btrfs_transaction *cur_trans = trans->transaction;  in btrfs_setup_space_cache()
2820 	if (list_empty(&cur_trans->dirty_bgs) ||  in btrfs_setup_space_cache()
2826 		return -ENOMEM;  in btrfs_setup_space_cache()
2828 	/* Could add new block groups, use _safe just in case */  in btrfs_setup_space_cache()
2829 	list_for_each_entry_safe(cache, tmp, &cur_trans->dirty_bgs,  in btrfs_setup_space_cache()
2831 		if (cache->disk_cache_state == BTRFS_DC_CLEAR)  in btrfs_setup_space_cache()
2832 			cache_save_setup(cache, trans, path);  in btrfs_setup_space_cache()
2840  * Transaction commit does final block group cache writeback during a critical
2842  * order for the cache to actually match the block group, but can introduce a
2845  * So, btrfs_start_dirty_block_groups is here to kick off block group cache IO.
2846  * There's a chance we'll have to redo some of it if the block group changes
2848  * getting rid of the easy block groups while we're still allowing others to
2853 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_start_dirty_block_groups()
2854 	struct btrfs_block_group *cache;  in btrfs_start_dirty_block_groups()  local
2855 	struct btrfs_transaction *cur_trans = trans->transaction;  in btrfs_start_dirty_block_groups()
2860 	struct list_head *io = &cur_trans->io_bgs;  in btrfs_start_dirty_block_groups()
2864 	spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2865 	if (list_empty(&cur_trans->dirty_bgs)) {  in btrfs_start_dirty_block_groups()
2866 		spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2869 	list_splice_init(&cur_trans->dirty_bgs, &dirty);  in btrfs_start_dirty_block_groups()
2870 	spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2873 	/* Make sure all the block groups on our dirty list actually exist */  in btrfs_start_dirty_block_groups()
2879 			ret = -ENOMEM;  in btrfs_start_dirty_block_groups()
2886 	 * removal of empty block groups deleting this block group while we are  in btrfs_start_dirty_block_groups()
2887 	 * writing out the cache  in btrfs_start_dirty_block_groups()
2889 	mutex_lock(&trans->transaction->cache_write_mutex);  in btrfs_start_dirty_block_groups()
2893 		cache = list_first_entry(&dirty, struct btrfs_block_group,  in btrfs_start_dirty_block_groups()
2896 		 * This can happen if something re-dirties a block group that  in btrfs_start_dirty_block_groups()
2900 		if (!list_empty(&cache->io_list)) {  in btrfs_start_dirty_block_groups()
2901 			list_del_init(&cache->io_list);  in btrfs_start_dirty_block_groups()
2902 			btrfs_wait_cache_io(trans, cache, path);  in btrfs_start_dirty_block_groups()
2903 			btrfs_put_block_group(cache);  in btrfs_start_dirty_block_groups()
2908 		 * btrfs_wait_cache_io uses the cache->dirty_list to decide if  in btrfs_start_dirty_block_groups()
2915 		spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2916 		list_del_init(&cache->dirty_list);  in btrfs_start_dirty_block_groups()
2917 		spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2921 		cache_save_setup(cache, trans, path);  in btrfs_start_dirty_block_groups()
2923 		if (cache->disk_cache_state == BTRFS_DC_SETUP) {  in btrfs_start_dirty_block_groups()
2924 			cache->io_ctl.inode = NULL;  in btrfs_start_dirty_block_groups()
2925 			ret = btrfs_write_out_cache(trans, cache, path);  in btrfs_start_dirty_block_groups()
2926 			if (ret == 0 && cache->io_ctl.inode) {  in btrfs_start_dirty_block_groups()
2935 				list_add_tail(&cache->io_list, io);  in btrfs_start_dirty_block_groups()
2938 				 * If we failed to write the cache, the  in btrfs_start_dirty_block_groups()
2945 			ret = update_block_group_item(trans, path, cache);  in btrfs_start_dirty_block_groups()
2947 			 * Our block group might still be attached to the list  in btrfs_start_dirty_block_groups()
2948 			 * of new block groups in the transaction handle of some  in btrfs_start_dirty_block_groups()
2949 			 * other task (struct btrfs_trans_handle->new_bgs). This  in btrfs_start_dirty_block_groups()
2950 			 * means its block group item isn't yet in the extent  in btrfs_start_dirty_block_groups()
2955 			if (ret == -ENOENT) {  in btrfs_start_dirty_block_groups()
2957 				spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2958 				if (list_empty(&cache->dirty_list)) {  in btrfs_start_dirty_block_groups()
2959 					list_add_tail(&cache->dirty_list,  in btrfs_start_dirty_block_groups()
2960 						      &cur_trans->dirty_bgs);  in btrfs_start_dirty_block_groups()
2961 					btrfs_get_block_group(cache);  in btrfs_start_dirty_block_groups()
2964 				spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2970 		/* If it's not on the io list, we need to put the block group */  in btrfs_start_dirty_block_groups()
2972 			btrfs_put_block_group(cache);  in btrfs_start_dirty_block_groups()
2977 		 * us from writing caches for block groups that are going to be  in btrfs_start_dirty_block_groups()
2980 		mutex_unlock(&trans->transaction->cache_write_mutex);  in btrfs_start_dirty_block_groups()
2983 		mutex_lock(&trans->transaction->cache_write_mutex);  in btrfs_start_dirty_block_groups()
2985 	mutex_unlock(&trans->transaction->cache_write_mutex);  in btrfs_start_dirty_block_groups()
2995 		spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
2996 		list_splice_init(&cur_trans->dirty_bgs, &dirty);  in btrfs_start_dirty_block_groups()
2998 		 * dirty_bgs_lock protects us from concurrent block group  in btrfs_start_dirty_block_groups()
3002 			spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
3005 		spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
3009 		spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
3010 		list_splice_init(&dirty, &cur_trans->dirty_bgs);  in btrfs_start_dirty_block_groups()
3011 		spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_start_dirty_block_groups()
3021 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_write_dirty_block_groups()
3022 	struct btrfs_block_group *cache;  in btrfs_write_dirty_block_groups()  local
3023 	struct btrfs_transaction *cur_trans = trans->transaction;  in btrfs_write_dirty_block_groups()
3027 	struct list_head *io = &cur_trans->io_bgs;  in btrfs_write_dirty_block_groups()
3032 		return -ENOMEM;  in btrfs_write_dirty_block_groups()
3037 	 * transaction's list of dirty block groups. These tasks correspond to  in btrfs_write_dirty_block_groups()
3039 	 * space cache, which run inode.c:btrfs_finish_ordered_io(), and can  in btrfs_write_dirty_block_groups()
3040 	 * allocate new block groups as a result of COWing nodes of the root  in btrfs_write_dirty_block_groups()
3049 	spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_write_dirty_block_groups()
3050 	while (!list_empty(&cur_trans->dirty_bgs)) {  in btrfs_write_dirty_block_groups()
3051 		cache = list_first_entry(&cur_trans->dirty_bgs,  in btrfs_write_dirty_block_groups()
3056 		 * This can happen if cache_save_setup re-dirties a block group  in btrfs_write_dirty_block_groups()
3060 		if (!list_empty(&cache->io_list)) {  in btrfs_write_dirty_block_groups()
3061 			spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_write_dirty_block_groups()
3062 			list_del_init(&cache->io_list);  in btrfs_write_dirty_block_groups()
3063 			btrfs_wait_cache_io(trans, cache, path);  in btrfs_write_dirty_block_groups()
3064 			btrfs_put_block_group(cache);  in btrfs_write_dirty_block_groups()
3065 			spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_write_dirty_block_groups()
3072 		list_del_init(&cache->dirty_list);  in btrfs_write_dirty_block_groups()
3073 		spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_write_dirty_block_groups()
3076 		cache_save_setup(cache, trans, path);  in btrfs_write_dirty_block_groups()
3080 						     (unsigned long) -1);  in btrfs_write_dirty_block_groups()
3082 		if (!ret && cache->disk_cache_state == BTRFS_DC_SETUP) {  in btrfs_write_dirty_block_groups()
3083 			cache->io_ctl.inode = NULL;  in btrfs_write_dirty_block_groups()
3084 			ret = btrfs_write_out_cache(trans, cache, path);  in btrfs_write_dirty_block_groups()
3085 			if (ret == 0 && cache->io_ctl.inode) {  in btrfs_write_dirty_block_groups()
3088 				list_add_tail(&cache->io_list, io);  in btrfs_write_dirty_block_groups()
3091 				 * If we failed to write the cache, the  in btrfs_write_dirty_block_groups()
3098 			ret = update_block_group_item(trans, path, cache);  in btrfs_write_dirty_block_groups()
3101 			 * created a new block group while updating a free space  in btrfs_write_dirty_block_groups()
3102 			 * cache's inode (at inode.c:btrfs_finish_ordered_io())  in btrfs_write_dirty_block_groups()
3104 			 * which case the new block group is still attached to  in btrfs_write_dirty_block_groups()
3106 			 * finished yet (no block group item in the extent tree  in btrfs_write_dirty_block_groups()
3112 			if (ret == -ENOENT) {  in btrfs_write_dirty_block_groups()
3113 				wait_event(cur_trans->writer_wait,  in btrfs_write_dirty_block_groups()
3114 				   atomic_read(&cur_trans->num_writers) == 1);  in btrfs_write_dirty_block_groups()
3115 				ret = update_block_group_item(trans, path, cache);  in btrfs_write_dirty_block_groups()
3121 		/* If its not on the io list, we need to put the block group */  in btrfs_write_dirty_block_groups()
3123 			btrfs_put_block_group(cache);  in btrfs_write_dirty_block_groups()
3125 		spin_lock(&cur_trans->dirty_bgs_lock);  in btrfs_write_dirty_block_groups()
3127 	spin_unlock(&cur_trans->dirty_bgs_lock);  in btrfs_write_dirty_block_groups()
3134 		cache = list_first_entry(io, struct btrfs_block_group,  in btrfs_write_dirty_block_groups()
3136 		list_del_init(&cache->io_list);  in btrfs_write_dirty_block_groups()
3137 		btrfs_wait_cache_io(trans, cache, path);  in btrfs_write_dirty_block_groups()
3138 		btrfs_put_block_group(cache);  in btrfs_write_dirty_block_groups()
3148 	struct btrfs_fs_info *info = trans->fs_info;  in btrfs_update_block_group()
3149 	struct btrfs_block_group *cache = NULL;  in btrfs_update_block_group()  local
3156 	/* Block accounting for super block */  in btrfs_update_block_group()
3157 	spin_lock(&info->delalloc_root_lock);  in btrfs_update_block_group()
3158 	old_val = btrfs_super_bytes_used(info->super_copy);  in btrfs_update_block_group()
3162 		old_val -= num_bytes;  in btrfs_update_block_group()
3163 	btrfs_set_super_bytes_used(info->super_copy, old_val);  in btrfs_update_block_group()
3164 	spin_unlock(&info->delalloc_root_lock);  in btrfs_update_block_group()
3167 		cache = btrfs_lookup_block_group(info, bytenr);  in btrfs_update_block_group()
3168 		if (!cache) {  in btrfs_update_block_group()
3169 			ret = -ENOENT;  in btrfs_update_block_group()
3172 		factor = btrfs_bg_type_to_factor(cache->flags);  in btrfs_update_block_group()
3175 		 * If this block group has free space cache written out, we  in btrfs_update_block_group()
3178 		 * space back to the block group, otherwise we will leak space.  in btrfs_update_block_group()
3180 		if (!alloc && !btrfs_block_group_done(cache))  in btrfs_update_block_group()
3181 			btrfs_cache_block_group(cache, 1);  in btrfs_update_block_group()
3183 		byte_in_group = bytenr - cache->start;  in btrfs_update_block_group()
3184 		WARN_ON(byte_in_group > cache->length);  in btrfs_update_block_group()
3186 		spin_lock(&cache->space_info->lock);  in btrfs_update_block_group()
3187 		spin_lock(&cache->lock);  in btrfs_update_block_group()
3190 		    cache->disk_cache_state < BTRFS_DC_CLEAR)  in btrfs_update_block_group()
3191 			cache->disk_cache_state = BTRFS_DC_CLEAR;  in btrfs_update_block_group()
3193 		old_val = cache->used;  in btrfs_update_block_group()
3194 		num_bytes = min(total, cache->length - byte_in_group);  in btrfs_update_block_group()
3197 			cache->used = old_val;  in btrfs_update_block_group()
3198 			cache->reserved -= num_bytes;  in btrfs_update_block_group()
3199 			cache->space_info->bytes_reserved -= num_bytes;  in btrfs_update_block_group()
3200 			cache->space_info->bytes_used += num_bytes;  in btrfs_update_block_group()
3201 			cache->space_info->disk_used += num_bytes * factor;  in btrfs_update_block_group()
3202 			spin_unlock(&cache->lock);  in btrfs_update_block_group()
3203 			spin_unlock(&cache->space_info->lock);  in btrfs_update_block_group()
3205 			old_val -= num_bytes;  in btrfs_update_block_group()
3206 			cache->used = old_val;  in btrfs_update_block_group()
3207 			cache->pinned += num_bytes;  in btrfs_update_block_group()
3209 					cache->space_info, num_bytes);  in btrfs_update_block_group()
3210 			cache->space_info->bytes_used -= num_bytes;  in btrfs_update_block_group()
3211 			cache->space_info->disk_used -= num_bytes * factor;  in btrfs_update_block_group()
3212 			spin_unlock(&cache->lock);  in btrfs_update_block_group()
3213 			spin_unlock(&cache->space_info->lock);  in btrfs_update_block_group()
3215 			set_extent_dirty(&trans->transaction->pinned_extents,  in btrfs_update_block_group()
3216 					 bytenr, bytenr + num_bytes - 1,  in btrfs_update_block_group()
3220 		spin_lock(&trans->transaction->dirty_bgs_lock);  in btrfs_update_block_group()
3221 		if (list_empty(&cache->dirty_list)) {  in btrfs_update_block_group()
3222 			list_add_tail(&cache->dirty_list,  in btrfs_update_block_group()
3223 				      &trans->transaction->dirty_bgs);  in btrfs_update_block_group()
3224 			trans->delayed_ref_updates++;  in btrfs_update_block_group()
3225 			btrfs_get_block_group(cache);  in btrfs_update_block_group()
3227 		spin_unlock(&trans->transaction->dirty_bgs_lock);  in btrfs_update_block_group()
3230 		 * No longer have used bytes in this block group, queue it for  in btrfs_update_block_group()
3231 		 * deletion. We do this after adding the block group to the  in btrfs_update_block_group()
3233 		 * cache writeout.  in btrfs_update_block_group()
3237 				btrfs_mark_bg_unused(cache);  in btrfs_update_block_group()
3240 		btrfs_put_block_group(cache);  in btrfs_update_block_group()
3241 		total -= num_bytes;  in btrfs_update_block_group()
3245 	/* Modified block groups are accounted for in the delayed_refs_rsv. */  in btrfs_update_block_group()
3251  * btrfs_add_reserved_bytes - update the block_group and space info counters
3252  * @cache:	The cache we are manipulating
3259  * reservation and the block group has become read only we cannot make the
3260  * reservation and return -EAGAIN, otherwise this function always succeeds.
3262 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,  in btrfs_add_reserved_bytes()  argument
3265 	struct btrfs_space_info *space_info = cache->space_info;  in btrfs_add_reserved_bytes()
3268 	spin_lock(&space_info->lock);  in btrfs_add_reserved_bytes()
3269 	spin_lock(&cache->lock);  in btrfs_add_reserved_bytes()
3270 	if (cache->ro) {  in btrfs_add_reserved_bytes()
3271 		ret = -EAGAIN;  in btrfs_add_reserved_bytes()
3273 		cache->reserved += num_bytes;  in btrfs_add_reserved_bytes()
3274 		space_info->bytes_reserved += num_bytes;  in btrfs_add_reserved_bytes()
3275 		trace_btrfs_space_reservation(cache->fs_info, "space_info",  in btrfs_add_reserved_bytes()
3276 					      space_info->flags, num_bytes, 1);  in btrfs_add_reserved_bytes()
3277 		btrfs_space_info_update_bytes_may_use(cache->fs_info,  in btrfs_add_reserved_bytes()
3278 						      space_info, -ram_bytes);  in btrfs_add_reserved_bytes()
3280 			cache->delalloc_bytes += num_bytes;  in btrfs_add_reserved_bytes()
3287 			btrfs_try_granting_tickets(cache->fs_info, space_info);  in btrfs_add_reserved_bytes()
3289 	spin_unlock(&cache->lock);  in btrfs_add_reserved_bytes()
3290 	spin_unlock(&space_info->lock);  in btrfs_add_reserved_bytes()
3295  * btrfs_free_reserved_bytes - update the block_group and space info counters
3296  * @cache:      The cache we are manipulating
3305 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,  in btrfs_free_reserved_bytes()  argument
3308 	struct btrfs_space_info *space_info = cache->space_info;  in btrfs_free_reserved_bytes()
3310 	spin_lock(&space_info->lock);  in btrfs_free_reserved_bytes()
3311 	spin_lock(&cache->lock);  in btrfs_free_reserved_bytes()
3312 	if (cache->ro)  in btrfs_free_reserved_bytes()
3313 		space_info->bytes_readonly += num_bytes;  in btrfs_free_reserved_bytes()
3314 	cache->reserved -= num_bytes;  in btrfs_free_reserved_bytes()
3315 	space_info->bytes_reserved -= num_bytes;  in btrfs_free_reserved_bytes()
3316 	space_info->max_extent_size = 0;  in btrfs_free_reserved_bytes()
3319 		cache->delalloc_bytes -= num_bytes;  in btrfs_free_reserved_bytes()
3320 	spin_unlock(&cache->lock);  in btrfs_free_reserved_bytes()
3322 	btrfs_try_granting_tickets(cache->fs_info, space_info);  in btrfs_free_reserved_bytes()
3323 	spin_unlock(&space_info->lock);  in btrfs_free_reserved_bytes()
3328 	struct list_head *head = &info->space_info;  in force_metadata_allocation()
3332 		if (found->flags & BTRFS_BLOCK_GROUP_METADATA)  in force_metadata_allocation()
3333 			found->force_alloc = CHUNK_ALLOC_FORCE;  in force_metadata_allocation()
3348 	 * about 1% of the FS size.  in should_alloc_chunk()
3351 		thresh = btrfs_super_total_bytes(fs_info->super_copy);  in should_alloc_chunk()
3354 		if (sinfo->total_bytes - bytes_used < thresh)  in should_alloc_chunk()
3358 	if (bytes_used + SZ_2M < div_factor(sinfo->total_bytes, 8))  in should_alloc_chunk()
3365 	u64 alloc_flags = btrfs_get_alloc_profile(trans->fs_info, type);  in btrfs_force_chunk_alloc()
3379 	 * system block group if needed.  in do_chunk_alloc()
3410 	 * Normally we are not expected to fail with -ENOSPC here, since we have  in do_chunk_alloc()
3415 	 *    existing system block groups have a profile which can not be used  in do_chunk_alloc()
3422 	 *    none of the block groups can be used for extent allocation since they  in do_chunk_alloc()
3426 	 *    block groups and check if they have a usable profile and enough space  in do_chunk_alloc()
3427 	 *    can be slow on very large filesystems, so we tolerate the -ENOSPC and  in do_chunk_alloc()
3433 	 *    block group to allocate from when we called check_system_chunk()  in do_chunk_alloc()
3434 	 *    above. However right after we called it, the only system block group  in do_chunk_alloc()
3438 	 *    handle and scrub uses the commit root to search for block groups.  in do_chunk_alloc()
3440 	if (ret == -ENOSPC) {  in do_chunk_alloc()
3441 		const u64 sys_flags = btrfs_system_alloc_profile(trans->fs_info);  in do_chunk_alloc()
3475  * 1) Phase 1 - through btrfs_chunk_alloc() we allocate device extents for
3476  *    the chunk, the chunk mapping, create its block group and add the items
3477  *    that belong in the chunk btree to it - more specifically, we need to
3480  * 2) Phase 2 - through btrfs_create_pending_block_groups(), we add the block
3486  * trigger chunk allocation and attempted to insert the new block group item
3497  *    allocate a new block group (chunk) because the only one that had enough
3499  *    device replace, block group reclaim thread, etc), so we can not use it
3504  *    the filesystem was mounted in degraded mode, none of the existing block
3506  *    profile (for e.g. mounting a 2 devices filesystem, where all block groups
3509  *    example, it will trigger allocation of a new metadata block group with a
3515  *    example, it does not find any free extent in any metadata block group,
3516  *    therefore forced to try to allocate a new metadata block group.
3518  *    meanwhile - this typically happens with tasks that don't reserve space
3525  * a seed device - we must create new metadata and system chunks without adding
3526  * any of the block group items to the chunk, extent and device btrees. If we
3528  * btrees, since all the chunks from the seed device are read-only.
3535  * superblock. This is easier to trigger if using a btree node/leaf size of 64K
3551  * holding fs_info->chunk_mutex. This is important to guarantee that while COWing
3557  * that mutex. The same logic applies to removing chunks - we must reserve system
3559  * while holding fs_info->chunk_mutex.
3564  *    - return 1 if it successfully allocates a chunk,
3565  *    - return errors including -ENOSPC otherwise.
3567  *    - return 0 if it doesn't need to allocate a new chunk,
3568  *    - return 1 if it successfully allocates a chunk,
3569  *    - return errors including -ENOSPC otherwise.
3574 	struct btrfs_fs_info *fs_info = trans->fs_info;  in btrfs_chunk_alloc()
3580 	/* Don't re-enter if we're already allocating a chunk */  in btrfs_chunk_alloc()
3581 	if (trans->allocating_chunk)  in btrfs_chunk_alloc()
3582 		return -ENOSPC;  in btrfs_chunk_alloc()
3584 	 * If we are removing a chunk, don't re-enter or we would deadlock.  in btrfs_chunk_alloc()
3588 	if (trans->removing_chunk)  in btrfs_chunk_alloc()
3589 		return -ENOSPC;  in btrfs_chunk_alloc()
3595 		spin_lock(&space_info->lock);  in btrfs_chunk_alloc()
3596 		if (force < space_info->force_alloc)  in btrfs_chunk_alloc()
3597 			force = space_info->force_alloc;  in btrfs_chunk_alloc()
3599 		if (space_info->full) {  in btrfs_chunk_alloc()
3602 				ret = -ENOSPC;  in btrfs_chunk_alloc()
3605 			spin_unlock(&space_info->lock);  in btrfs_chunk_alloc()
3608 			spin_unlock(&space_info->lock);  in btrfs_chunk_alloc()
3610 		} else if (space_info->chunk_alloc) {  in btrfs_chunk_alloc()
3612 			 * Someone is already allocating, so we need to block  in btrfs_chunk_alloc()
3618 			spin_unlock(&space_info->lock);  in btrfs_chunk_alloc()
3619 			mutex_lock(&fs_info->chunk_mutex);  in btrfs_chunk_alloc()
3620 			mutex_unlock(&fs_info->chunk_mutex);  in btrfs_chunk_alloc()
3623 			space_info->chunk_alloc = 1;  in btrfs_chunk_alloc()
3625 			spin_unlock(&space_info->lock);  in btrfs_chunk_alloc()
3631 	mutex_lock(&fs_info->chunk_mutex);  in btrfs_chunk_alloc()
3632 	trans->allocating_chunk = true;  in btrfs_chunk_alloc()
3646 	if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {  in btrfs_chunk_alloc()
3647 		fs_info->data_chunk_allocations++;  in btrfs_chunk_alloc()
3648 		if (!(fs_info->data_chunk_allocations %  in btrfs_chunk_alloc()
3649 		      fs_info->metadata_ratio))  in btrfs_chunk_alloc()
3654 	trans->allocating_chunk = false;  in btrfs_chunk_alloc()
3656 	spin_lock(&space_info->lock);  in btrfs_chunk_alloc()
3658 		if (ret == -ENOSPC)  in btrfs_chunk_alloc()
3659 			space_info->full = 1;  in btrfs_chunk_alloc()
3664 		space_info->max_extent_size = 0;  in btrfs_chunk_alloc()
3667 	space_info->force_alloc = CHUNK_ALLOC_NO_FORCE;  in btrfs_chunk_alloc()
3669 	space_info->chunk_alloc = 0;  in btrfs_chunk_alloc()
3670 	spin_unlock(&space_info->lock);  in btrfs_chunk_alloc()
3671 	mutex_unlock(&fs_info->chunk_mutex);  in btrfs_chunk_alloc()
3682 		num_dev = fs_info->fs_devices->rw_devices;  in get_profile_num_devs()
3692 	struct btrfs_fs_info *fs_info = trans->fs_info;  in check_system_chunk()
3701 	 * atomic and race free space reservation in the chunk block reserve.  in check_system_chunk()
3703 	lockdep_assert_held(&fs_info->chunk_mutex);  in check_system_chunk()
3706 	spin_lock(&info->lock);  in check_system_chunk()
3707 	left = info->total_bytes - btrfs_space_info_used(info, true);  in check_system_chunk()
3708 	spin_unlock(&info->lock);  in check_system_chunk()
3752 		ret = btrfs_block_rsv_add(fs_info->chunk_root,  in check_system_chunk()
3753 					  &fs_info->chunk_block_rsv,  in check_system_chunk()
3756 			trans->chunk_bytes_reserved += thresh;  in check_system_chunk()
3771 			spin_lock(&block_group->lock);  in btrfs_put_block_group_cache()
3772 			if (block_group->iref)  in btrfs_put_block_group_cache()
3774 			spin_unlock(&block_group->lock);  in btrfs_put_block_group_cache()
3784 		inode = block_group->inode;  in btrfs_put_block_group_cache()
3785 		block_group->iref = 0;  in btrfs_put_block_group_cache()
3786 		block_group->inode = NULL;  in btrfs_put_block_group_cache()
3787 		spin_unlock(&block_group->lock);  in btrfs_put_block_group_cache()
3788 		ASSERT(block_group->io_ctl.inode == NULL);  in btrfs_put_block_group_cache()
3790 		last = block_group->start + block_group->length;  in btrfs_put_block_group_cache()
3796  * Must be called only after stopping all workers, since we could have block
3798  * freed the block groups before stopping them.
3807 	spin_lock(&info->block_group_cache_lock);  in btrfs_free_block_groups()
3808 	while (!list_empty(&info->caching_block_groups)) {  in btrfs_free_block_groups()
3809 		caching_ctl = list_entry(info->caching_block_groups.next,  in btrfs_free_block_groups()
3811 		list_del(&caching_ctl->list);  in btrfs_free_block_groups()
3814 	spin_unlock(&info->block_group_cache_lock);  in btrfs_free_block_groups()
3816 	spin_lock(&info->unused_bgs_lock);  in btrfs_free_block_groups()
3817 	while (!list_empty(&info->unused_bgs)) {  in btrfs_free_block_groups()
3818 		block_group = list_first_entry(&info->unused_bgs,  in btrfs_free_block_groups()
3821 		list_del_init(&block_group->bg_list);  in btrfs_free_block_groups()
3824 	spin_unlock(&info->unused_bgs_lock);  in btrfs_free_block_groups()
3826 	spin_lock(&info->unused_bgs_lock);  in btrfs_free_block_groups()
3827 	while (!list_empty(&info->reclaim_bgs)) {  in btrfs_free_block_groups()
3828 		block_group = list_first_entry(&info->reclaim_bgs,  in btrfs_free_block_groups()
3831 		list_del_init(&block_group->bg_list);  in btrfs_free_block_groups()
3834 	spin_unlock(&info->unused_bgs_lock);  in btrfs_free_block_groups()
3836 	spin_lock(&info->block_group_cache_lock);  in btrfs_free_block_groups()
3837 	while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {  in btrfs_free_block_groups()
3840 		rb_erase(&block_group->cache_node,  in btrfs_free_block_groups()
3841 			 &info->block_group_cache_tree);  in btrfs_free_block_groups()
3842 		RB_CLEAR_NODE(&block_group->cache_node);  in btrfs_free_block_groups()
3843 		spin_unlock(&info->block_group_cache_lock);  in btrfs_free_block_groups()
3845 		down_write(&block_group->space_info->groups_sem);  in btrfs_free_block_groups()
3846 		list_del(&block_group->list);  in btrfs_free_block_groups()
3847 		up_write(&block_group->space_info->groups_sem);  in btrfs_free_block_groups()
3850 		 * We haven't cached this block group, which means we could  in btrfs_free_block_groups()
3851 		 * possibly have excluded extents on this block group.  in btrfs_free_block_groups()
3853 		if (block_group->cached == BTRFS_CACHE_NO ||  in btrfs_free_block_groups()
3854 		    block_group->cached == BTRFS_CACHE_ERROR)  in btrfs_free_block_groups()
3858 		ASSERT(block_group->cached != BTRFS_CACHE_STARTED);  in btrfs_free_block_groups()
3859 		ASSERT(list_empty(&block_group->dirty_list));  in btrfs_free_block_groups()
3860 		ASSERT(list_empty(&block_group->io_list));  in btrfs_free_block_groups()
3861 		ASSERT(list_empty(&block_group->bg_list));  in btrfs_free_block_groups()
3862 		ASSERT(refcount_read(&block_group->refs) == 1);  in btrfs_free_block_groups()
3863 		ASSERT(block_group->swap_extents == 0);  in btrfs_free_block_groups()
3866 		spin_lock(&info->block_group_cache_lock);  in btrfs_free_block_groups()
3868 	spin_unlock(&info->block_group_cache_lock);  in btrfs_free_block_groups()
3872 	while (!list_empty(&info->space_info)) {  in btrfs_free_block_groups()
3873 		space_info = list_entry(info->space_info.next,  in btrfs_free_block_groups()
3881 		if (WARN_ON(space_info->bytes_pinned > 0 ||  in btrfs_free_block_groups()
3882 			    space_info->bytes_reserved > 0 ||  in btrfs_free_block_groups()
3883 			    space_info->bytes_may_use > 0))  in btrfs_free_block_groups()
3885 		WARN_ON(space_info->reclaim_size > 0);  in btrfs_free_block_groups()
3886 		list_del(&space_info->list);  in btrfs_free_block_groups()
3892 void btrfs_freeze_block_group(struct btrfs_block_group *cache)  in btrfs_freeze_block_group()  argument
3894 	atomic_inc(&cache->frozen);  in btrfs_freeze_block_group()
3899 	struct btrfs_fs_info *fs_info = block_group->fs_info;  in btrfs_unfreeze_block_group()
3904 	spin_lock(&block_group->lock);  in btrfs_unfreeze_block_group()
3905 	cleanup = (atomic_dec_and_test(&block_group->frozen) &&  in btrfs_unfreeze_block_group()
3906 		   block_group->removed);  in btrfs_unfreeze_block_group()
3907 	spin_unlock(&block_group->lock);  in btrfs_unfreeze_block_group()
3910 		em_tree = &fs_info->mapping_tree;  in btrfs_unfreeze_block_group()
3911 		write_lock(&em_tree->lock);  in btrfs_unfreeze_block_group()
3912 		em = lookup_extent_mapping(em_tree, block_group->start,  in btrfs_unfreeze_block_group()
3916 		write_unlock(&em_tree->lock);  in btrfs_unfreeze_block_group()
3924 		 * tasks trimming this block group have left 1 entry each one.  in btrfs_unfreeze_block_group()
3927 		__btrfs_remove_free_space_cache(block_group->free_space_ctl);  in btrfs_unfreeze_block_group()
3935 	spin_lock(&bg->lock);  in btrfs_inc_block_group_swap_extents()
3936 	if (bg->ro)  in btrfs_inc_block_group_swap_extents()
3939 		bg->swap_extents++;  in btrfs_inc_block_group_swap_extents()
3940 	spin_unlock(&bg->lock);  in btrfs_inc_block_group_swap_extents()
3947 	spin_lock(&bg->lock);  in btrfs_dec_block_group_swap_extents()
3948 	ASSERT(!bg->ro);  in btrfs_dec_block_group_swap_extents()
3949 	ASSERT(bg->swap_extents >= amount);  in btrfs_dec_block_group_swap_extents()
3950 	bg->swap_extents -= amount;  in btrfs_dec_block_group_swap_extents()
3951 	spin_unlock(&bg->lock);  in btrfs_dec_block_group_swap_extents()