diff options
Diffstat (limited to 'fs/btrfs/block-group.c')
| -rw-r--r-- | fs/btrfs/block-group.c | 2212 |
1 files changed, 1502 insertions, 710 deletions
diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c index 7f09147872dc..deebc8ddbd93 100644 --- a/fs/btrfs/block-group.c +++ b/fs/btrfs/block-group.c @@ -1,5 +1,6 @@ // SPDX-License-Identifier: GPL-2.0 +#include <linux/list_sort.h> #include "misc.h" #include "ctree.h" #include "block-group.h" @@ -7,7 +8,6 @@ #include "disk-io.h" #include "free-space-cache.h" #include "free-space-tree.h" -#include "disk-io.h" #include "volumes.h" #include "transaction.h" #include "ref-verify.h" @@ -16,6 +16,7 @@ #include "delalloc-space.h" #include "discard.h" #include "raid56.h" +#include "zoned.h" /* * Return target flags in extended format or 0 if restripe for this chunk_type @@ -66,11 +67,8 @@ static u64 btrfs_reduce_alloc_profile(struct btrfs_fs_info *fs_info, u64 flags) spin_lock(&fs_info->balance_lock); target = get_restripe_target(fs_info, flags); if (target) { - /* Pick target profile only if it's already available */ - if ((flags & target) & BTRFS_EXTENDED_PROFILE_MASK) { - spin_unlock(&fs_info->balance_lock); - return extended_to_chunk(target); - } + spin_unlock(&fs_info->balance_lock); + return extended_to_chunk(target); } spin_unlock(&fs_info->balance_lock); @@ -119,14 +117,23 @@ u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags) void btrfs_get_block_group(struct btrfs_block_group *cache) { - atomic_inc(&cache->count); + refcount_inc(&cache->refs); } void btrfs_put_block_group(struct btrfs_block_group *cache) { - if (atomic_dec_and_test(&cache->count)) { + if (refcount_dec_and_test(&cache->refs)) { WARN_ON(cache->pinned > 0); - WARN_ON(cache->reserved > 0); + /* + * If there was a failure to cleanup a log tree, very likely due + * to an IO failure on a writeback attempt of one or more of its + * extent buffers, we could not do proper (and cheap) unaccounting + * of their reserved space, so don't warn on reserved > 0 in that + * case. + */ + if (!(cache->flags & BTRFS_BLOCK_GROUP_METADATA) || + !BTRFS_FS_LOG_CLEANUP_ERROR(cache->fs_info)) + WARN_ON(cache->reserved > 0); /* * A block_group shouldn't be on the discard_list anymore. @@ -147,6 +154,7 @@ void btrfs_put_block_group(struct btrfs_block_group *cache) */ WARN_ON(!RB_EMPTY_ROOT(&cache->full_stripe_locks_root.root)); kfree(cache->free_space_ctl); + kfree(cache->physical_map); kfree(cache); } } @@ -160,9 +168,12 @@ static int btrfs_add_block_group_cache(struct btrfs_fs_info *info, struct rb_node **p; struct rb_node *parent = NULL; struct btrfs_block_group *cache; + bool leftmost = true; - spin_lock(&info->block_group_cache_lock); - p = &info->block_group_cache_tree.rb_node; + ASSERT(block_group->length != 0); + + write_lock(&info->block_group_cache_lock); + p = &info->block_group_cache_tree.rb_root.rb_node; while (*p) { parent = *p; @@ -171,20 +182,18 @@ static int btrfs_add_block_group_cache(struct btrfs_fs_info *info, p = &(*p)->rb_left; } else if (block_group->start > cache->start) { p = &(*p)->rb_right; + leftmost = false; } else { - spin_unlock(&info->block_group_cache_lock); + write_unlock(&info->block_group_cache_lock); return -EEXIST; } } rb_link_node(&block_group->cache_node, parent, p); - rb_insert_color(&block_group->cache_node, - &info->block_group_cache_tree); - - if (info->first_logical_byte > block_group->start) - info->first_logical_byte = block_group->start; + rb_insert_color_cached(&block_group->cache_node, + &info->block_group_cache_tree, leftmost); - spin_unlock(&info->block_group_cache_lock); + write_unlock(&info->block_group_cache_lock); return 0; } @@ -200,8 +209,8 @@ static struct btrfs_block_group *block_group_cache_tree_search( struct rb_node *n; u64 end, start; - spin_lock(&info->block_group_cache_lock); - n = info->block_group_cache_tree.rb_node; + read_lock(&info->block_group_cache_lock); + n = info->block_group_cache_tree.rb_root.rb_node; while (n) { cache = rb_entry(n, struct btrfs_block_group, cache_node); @@ -223,12 +232,9 @@ static struct btrfs_block_group *block_group_cache_tree_search( break; } } - if (ret) { + if (ret) btrfs_get_block_group(ret); - if (bytenr == 0 && info->first_logical_byte > ret->start) - info->first_logical_byte = ret->start; - } - spin_unlock(&info->block_group_cache_lock); + read_unlock(&info->block_group_cache_lock); return ret; } @@ -257,15 +263,15 @@ struct btrfs_block_group *btrfs_next_block_group( struct btrfs_fs_info *fs_info = cache->fs_info; struct rb_node *node; - spin_lock(&fs_info->block_group_cache_lock); + read_lock(&fs_info->block_group_cache_lock); /* If our block group was removed, we need a full search. */ if (RB_EMPTY_NODE(&cache->cache_node)) { const u64 next_bytenr = cache->start + cache->length; - spin_unlock(&fs_info->block_group_cache_lock); + read_unlock(&fs_info->block_group_cache_lock); btrfs_put_block_group(cache); - cache = btrfs_lookup_first_block_group(fs_info, next_bytenr); return cache; + return btrfs_lookup_first_block_group(fs_info, next_bytenr); } node = rb_next(&cache->cache_node); btrfs_put_block_group(cache); @@ -274,46 +280,70 @@ struct btrfs_block_group *btrfs_next_block_group( btrfs_get_block_group(cache); } else cache = NULL; - spin_unlock(&fs_info->block_group_cache_lock); + read_unlock(&fs_info->block_group_cache_lock); return cache; } -bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr) +/** + * Check if we can do a NOCOW write for a given extent. + * + * @fs_info: The filesystem information object. + * @bytenr: Logical start address of the extent. + * + * Check if we can do a NOCOW write for the given extent, and increments the + * number of NOCOW writers in the block group that contains the extent, as long + * as the block group exists and it's currently not in read-only mode. + * + * Returns: A non-NULL block group pointer if we can do a NOCOW write, the caller + * is responsible for calling btrfs_dec_nocow_writers() later. + * + * Or NULL if we can not do a NOCOW write + */ +struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, + u64 bytenr) { struct btrfs_block_group *bg; - bool ret = true; + bool can_nocow = true; bg = btrfs_lookup_block_group(fs_info, bytenr); if (!bg) - return false; + return NULL; spin_lock(&bg->lock); if (bg->ro) - ret = false; + can_nocow = false; else atomic_inc(&bg->nocow_writers); spin_unlock(&bg->lock); - /* No put on block group, done by btrfs_dec_nocow_writers */ - if (!ret) + if (!can_nocow) { btrfs_put_block_group(bg); + return NULL; + } - return ret; + /* No put on block group, done by btrfs_dec_nocow_writers(). */ + return bg; } -void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr) +/** + * Decrement the number of NOCOW writers in a block group. + * + * @bg: The block group. + * + * This is meant to be called after a previous call to btrfs_inc_nocow_writers(), + * and on the block group returned by that call. Typically this is called after + * creating an ordered extent for a NOCOW write, to prevent races with scrub and + * relocation. + * + * After this call, the caller should not use the block group anymore. It it wants + * to use it, then it should get a reference on it before calling this function. + */ +void btrfs_dec_nocow_writers(struct btrfs_block_group *bg) { - struct btrfs_block_group *bg; - - bg = btrfs_lookup_block_group(fs_info, bytenr); - ASSERT(bg); if (atomic_dec_and_test(&bg->nocow_writers)) wake_up_var(&bg->nocow_writers); - /* - * Once for our lookup and once for the lookup done by a previous call - * to btrfs_inc_nocow_writers() - */ - btrfs_put_block_group(bg); + + /* For the lookup done by a previous call to btrfs_inc_nocow_writers(). */ btrfs_put_block_group(bg); } @@ -410,18 +440,22 @@ void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache, btrfs_put_caching_control(caching_ctl); } -int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache) +static int btrfs_caching_ctl_wait_done(struct btrfs_block_group *cache, + struct btrfs_caching_control *caching_ctl) +{ + wait_event(caching_ctl->wait, btrfs_block_group_done(cache)); + return cache->cached == BTRFS_CACHE_ERROR ? -EIO : 0; +} + +static int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache) { struct btrfs_caching_control *caching_ctl; - int ret = 0; + int ret; caching_ctl = btrfs_get_caching_control(cache); if (!caching_ctl) return (cache->cached == BTRFS_CACHE_ERROR) ? -EIO : 0; - - wait_event(caching_ctl->wait, btrfs_block_group_done(cache)); - if (cache->cached == BTRFS_CACHE_ERROR) - ret = -EIO; + ret = btrfs_caching_ctl_wait_done(cache, caching_ctl); btrfs_put_caching_control(caching_ctl); return ret; } @@ -460,7 +494,7 @@ u64 add_new_free_space(struct btrfs_block_group *block_group, u64 start, u64 end int ret; while (start < end) { - ret = find_first_extent_bit(info->pinned_extents, start, + ret = find_first_extent_bit(&info->excluded_extents, start, &extent_start, &extent_end, EXTENT_DIRTY | EXTENT_UPTODATE, NULL); @@ -496,7 +530,7 @@ static int load_extent_tree_free(struct btrfs_caching_control *caching_ctl) { struct btrfs_block_group *block_group = caching_ctl->block_group; struct btrfs_fs_info *fs_info = block_group->fs_info; - struct btrfs_root *extent_root = fs_info->extent_root; + struct btrfs_root *extent_root; struct btrfs_path *path; struct extent_buffer *leaf; struct btrfs_key key; @@ -511,6 +545,7 @@ static int load_extent_tree_free(struct btrfs_caching_control *caching_ctl) return -ENOMEM; last = max_t(u64, block_group->start, BTRFS_SUPER_INFO_OFFSET); + extent_root = btrfs_extent_root(fs_info, last); #ifdef CONFIG_BTRFS_DEBUG /* @@ -558,8 +593,6 @@ next: if (need_resched() || rwsem_is_contended(&fs_info->commit_root_sem)) { - if (wakeup) - caching_ctl->progress = last; btrfs_release_path(path); up_read(&fs_info->commit_root_sem); mutex_unlock(&caching_ctl->mutex); @@ -583,9 +616,6 @@ next: key.objectid = last; key.offset = 0; key.type = BTRFS_EXTENT_ITEM_KEY; - - if (wakeup) - caching_ctl->progress = last; btrfs_release_path(path); goto next; } @@ -620,7 +650,6 @@ next: total_found += add_new_free_space(block_group, last, block_group->start + block_group->length); - caching_ctl->progress = (u64)-1; out: btrfs_free_path(path); @@ -641,11 +670,36 @@ static noinline void caching_thread(struct btrfs_work *work) mutex_lock(&caching_ctl->mutex); down_read(&fs_info->commit_root_sem); - if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) + if (btrfs_test_opt(fs_info, SPACE_CACHE)) { + ret = load_free_space_cache(block_group); + if (ret == 1) { + ret = 0; + goto done; + } + + /* + * We failed to load the space cache, set ourselves to + * CACHE_STARTED and carry on. + */ + spin_lock(&block_group->lock); + block_group->cached = BTRFS_CACHE_STARTED; + spin_unlock(&block_group->lock); + wake_up(&caching_ctl->wait); + } + + /* + * If we are in the transaction that populated the free space tree we + * can't actually cache from the free space tree as our commit root and + * real root are the same, so we could change the contents of the blocks + * while caching. Instead do the slow caching in this case, and after + * the transaction has committed we will be safe. + */ + if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && + !(test_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags))) ret = load_free_space_tree(caching_ctl); else ret = load_extent_tree_free(caching_ctl); - +done: spin_lock(&block_group->lock); block_group->caching_ctl = NULL; block_group->cached = ret ? BTRFS_CACHE_ERROR : BTRFS_CACHE_FINISHED; @@ -665,8 +719,6 @@ static noinline void caching_thread(struct btrfs_work *work) } #endif - caching_ctl->progress = (u64)-1; - up_read(&fs_info->commit_root_sem); btrfs_free_excluded_extents(block_group); mutex_unlock(&caching_ctl->mutex); @@ -677,13 +729,16 @@ static noinline void caching_thread(struct btrfs_work *work) btrfs_put_block_group(block_group); } -int btrfs_cache_block_group(struct btrfs_block_group *cache, int load_cache_only) +int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait) { - DEFINE_WAIT(wait); struct btrfs_fs_info *fs_info = cache->fs_info; - struct btrfs_caching_control *caching_ctl; + struct btrfs_caching_control *caching_ctl = NULL; int ret = 0; + /* Allocator for zoned filesystems does not use the cache at all */ + if (btrfs_is_zoned(fs_info)) + return 0; + caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS); if (!caching_ctl) return -ENOMEM; @@ -692,120 +747,37 @@ int btrfs_cache_block_group(struct btrfs_block_group *cache, int load_cache_only mutex_init(&caching_ctl->mutex); init_waitqueue_head(&caching_ctl->wait); caching_ctl->block_group = cache; - caching_ctl->progress = cache->start; - refcount_set(&caching_ctl->count, 1); + refcount_set(&caching_ctl->count, 2); btrfs_init_work(&caching_ctl->work, caching_thread, NULL, NULL); spin_lock(&cache->lock); - /* - * This should be a rare occasion, but this could happen I think in the - * case where one thread starts to load the space cache info, and then - * some other thread starts a transaction commit which tries to do an - * allocation while the other thread is still loading the space cache - * info. The previous loop should have kept us from choosing this block - * group, but if we've moved to the state where we will wait on caching - * block groups we need to first check if we're doing a fast load here, - * so we can wait for it to finish, otherwise we could end up allocating - * from a block group who's cache gets evicted for one reason or - * another. - */ - while (cache->cached == BTRFS_CACHE_FAST) { - struct btrfs_caching_control *ctl; - - ctl = cache->caching_ctl; - refcount_inc(&ctl->count); - prepare_to_wait(&ctl->wait, &wait, TASK_UNINTERRUPTIBLE); - spin_unlock(&cache->lock); - - schedule(); - - finish_wait(&ctl->wait, &wait); - btrfs_put_caching_control(ctl); - spin_lock(&cache->lock); - } - if (cache->cached != BTRFS_CACHE_NO) { - spin_unlock(&cache->lock); kfree(caching_ctl); - return 0; + + caching_ctl = cache->caching_ctl; + if (caching_ctl) + refcount_inc(&caching_ctl->count); + spin_unlock(&cache->lock); + goto out; } WARN_ON(cache->caching_ctl); cache->caching_ctl = caching_ctl; - cache->cached = BTRFS_CACHE_FAST; + cache->cached = BTRFS_CACHE_STARTED; spin_unlock(&cache->lock); - if (btrfs_test_opt(fs_info, SPACE_CACHE)) { - mutex_lock(&caching_ctl->mutex); - ret = load_free_space_cache(cache); - - spin_lock(&cache->lock); - if (ret == 1) { - cache->caching_ctl = NULL; - cache->cached = BTRFS_CACHE_FINISHED; - cache->last_byte_to_unpin = (u64)-1; - caching_ctl->progress = (u64)-1; - } else { - if (load_cache_only) { - cache->caching_ctl = NULL; - cache->cached = BTRFS_CACHE_NO; - } else { - cache->cached = BTRFS_CACHE_STARTED; - cache->has_caching_ctl = 1; - } - } - spin_unlock(&cache->lock); -#ifdef CONFIG_BTRFS_DEBUG - if (ret == 1 && - btrfs_should_fragment_free_space(cache)) { - u64 bytes_used; - - spin_lock(&cache->space_info->lock); - spin_lock(&cache->lock); - bytes_used = cache->length - cache->used; - cache->space_info->bytes_used += bytes_used >> 1; - spin_unlock(&cache->lock); - spin_unlock(&cache->space_info->lock); - fragment_free_space(cache); - } -#endif - mutex_unlock(&caching_ctl->mutex); - - wake_up(&caching_ctl->wait); - if (ret == 1) { - btrfs_put_caching_control(caching_ctl); - btrfs_free_excluded_extents(cache); - return 0; - } - } else { - /* - * We're either using the free space tree or no caching at all. - * Set cached to the appropriate value and wakeup any waiters. - */ - spin_lock(&cache->lock); - if (load_cache_only) { - cache->caching_ctl = NULL; - cache->cached = BTRFS_CACHE_NO; - } else { - cache->cached = BTRFS_CACHE_STARTED; - cache->has_caching_ctl = 1; - } - spin_unlock(&cache->lock); - wake_up(&caching_ctl->wait); - } - - if (load_cache_only) { - btrfs_put_caching_control(caching_ctl); - return 0; - } - - down_write(&fs_info->commit_root_sem); + write_lock(&fs_info->block_group_cache_lock); refcount_inc(&caching_ctl->count); list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups); - up_write(&fs_info->commit_root_sem); + write_unlock(&fs_info->block_group_cache_lock); btrfs_get_block_group(cache); btrfs_queue_work(fs_info->caching_workers, &caching_ctl->work); +out: + if (wait && caching_ctl) + ret = btrfs_caching_ctl_wait_done(cache, caching_ctl); + if (caching_ctl) + btrfs_put_caching_control(caching_ctl); return ret; } @@ -863,16 +835,37 @@ static void clear_incompat_bg_bits(struct btrfs_fs_info *fs_info, u64 flags) } } +static int remove_block_group_item(struct btrfs_trans_handle *trans, + struct btrfs_path *path, + struct btrfs_block_group *block_group) +{ + struct btrfs_fs_info *fs_info = trans->fs_info; + struct btrfs_root *root; + struct btrfs_key key; + int ret; + + root = btrfs_block_group_root(fs_info); + key.objectid = block_group->start; + key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; + key.offset = block_group->length; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret > 0) + ret = -ENOENT; + if (ret < 0) + return ret; + + ret = btrfs_del_item(trans, root, path); + return ret; +} + int btrfs_remove_block_group(struct btrfs_trans_handle *trans, u64 group_start, struct extent_map *em) { struct btrfs_fs_info *fs_info = trans->fs_info; - struct btrfs_root *root = fs_info->extent_root; struct btrfs_path *path; struct btrfs_block_group *block_group; struct btrfs_free_cluster *cluster; - struct btrfs_root *tree_root = fs_info->tree_root; - struct btrfs_key key; struct inode *inode; struct kobject *kobj = NULL; int ret; @@ -913,6 +906,9 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, btrfs_return_cluster_to_free_space(block_group, cluster); spin_unlock(&cluster->refill_lock); + btrfs_clear_treelog_bg(block_group); + btrfs_clear_data_reloc_bg(block_group); + path = btrfs_alloc_path(); if (!path) { ret = -ENOMEM; @@ -950,51 +946,19 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, spin_unlock(&trans->transaction->dirty_bgs_lock); mutex_unlock(&trans->transaction->cache_write_mutex); - if (!IS_ERR(inode)) { - ret = btrfs_orphan_add(trans, BTRFS_I(inode)); - if (ret) { - btrfs_add_delayed_iput(inode); - goto out; - } - clear_nlink(inode); - /* One for the block groups ref */ - spin_lock(&block_group->lock); - if (block_group->iref) { - block_group->iref = 0; - block_group->inode = NULL; - spin_unlock(&block_group->lock); - iput(inode); - } else { - spin_unlock(&block_group->lock); - } - /* One for our lookup ref */ - btrfs_add_delayed_iput(inode); - } - - key.objectid = BTRFS_FREE_SPACE_OBJECTID; - key.type = 0; - key.offset = block_group->start; - - ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1); - if (ret < 0) + ret = btrfs_remove_free_space_inode(trans, inode, block_group); + if (ret) goto out; - if (ret > 0) - btrfs_release_path(path); - if (ret == 0) { - ret = btrfs_del_item(trans, tree_root, path); - if (ret) - goto out; - btrfs_release_path(path); - } - spin_lock(&fs_info->block_group_cache_lock); - rb_erase(&block_group->cache_node, - &fs_info->block_group_cache_tree); + write_lock(&fs_info->block_group_cache_lock); + rb_erase_cached(&block_group->cache_node, + &fs_info->block_group_cache_tree); RB_CLEAR_NODE(&block_group->cache_node); - if (fs_info->first_logical_byte == block_group->start) - fs_info->first_logical_byte = (u64)-1; - spin_unlock(&fs_info->block_group_cache_lock); + /* Once for the block groups rbtree */ + btrfs_put_block_group(block_group); + + write_unlock(&fs_info->block_group_cache_lock); down_write(&block_group->space_info->groups_sem); /* @@ -1014,32 +978,31 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, kobject_put(kobj); } - if (block_group->has_caching_ctl) - caching_ctl = btrfs_get_caching_control(block_group); if (block_group->cached == BTRFS_CACHE_STARTED) btrfs_wait_block_group_cache_done(block_group); - if (block_group->has_caching_ctl) { - down_write(&fs_info->commit_root_sem); - if (!caching_ctl) { - struct btrfs_caching_control *ctl; - - list_for_each_entry(ctl, - &fs_info->caching_block_groups, list) - if (ctl->block_group == block_group) { - caching_ctl = ctl; - refcount_inc(&caching_ctl->count); - break; - } - } - if (caching_ctl) - list_del_init(&caching_ctl->list); - up_write(&fs_info->commit_root_sem); - if (caching_ctl) { - /* Once for the caching bgs list and once for us. */ - btrfs_put_caching_control(caching_ctl); - btrfs_put_caching_control(caching_ctl); + + write_lock(&fs_info->block_group_cache_lock); + caching_ctl = btrfs_get_caching_control(block_group); + if (!caching_ctl) { + struct btrfs_caching_control *ctl; + + list_for_each_entry(ctl, &fs_info->caching_block_groups, list) { + if (ctl->block_group == block_group) { + caching_ctl = ctl; + refcount_inc(&caching_ctl->count); + break; + } } } + if (caching_ctl) + list_del_init(&caching_ctl->list); + write_unlock(&fs_info->block_group_cache_lock); + + if (caching_ctl) { + /* Once for the caching bgs list and once for us. */ + btrfs_put_caching_control(caching_ctl); + btrfs_put_caching_control(caching_ctl); + } spin_lock(&trans->transaction->dirty_bgs_lock); WARN_ON(!list_empty(&block_group->dirty_list)); @@ -1055,36 +1018,65 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, WARN_ON(block_group->space_info->total_bytes < block_group->length); WARN_ON(block_group->space_info->bytes_readonly - < block_group->length); + < block_group->length - block_group->zone_unusable); + WARN_ON(block_group->space_info->bytes_zone_unusable + < block_group->zone_unusable); WARN_ON(block_group->space_info->disk_total < block_group->length * factor); + WARN_ON(test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, + &block_group->runtime_flags) && + block_group->space_info->active_total_bytes + < block_group->length); } block_group->space_info->total_bytes -= block_group->length; - block_group->space_info->bytes_readonly -= block_group->length; + if (test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &block_group->runtime_flags)) + block_group->space_info->active_total_bytes -= block_group->length; + block_group->space_info->bytes_readonly -= + (block_group->length - block_group->zone_unusable); + block_group->space_info->bytes_zone_unusable -= + block_group->zone_unusable; block_group->space_info->disk_total -= block_group->length * factor; spin_unlock(&block_group->space_info->lock); - key.objectid = block_group->start; - key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; - key.offset = block_group->length; + /* + * Remove the free space for the block group from the free space tree + * and the block group's item from the extent tree before marking the + * block group as removed. This is to prevent races with tasks that + * freeze and unfreeze a block group, this task and another task + * allocating a new block group - the unfreeze task ends up removing + * the block group's extent map before the task calling this function + * deletes the block group item from the extent tree, allowing for + * another task to attempt to create another block group with the same + * item key (and failing with -EEXIST and a transaction abort). + */ + ret = remove_block_group_free_space(trans, block_group); + if (ret) + goto out; + + ret = remove_block_group_item(trans, path, block_group); + if (ret < 0) + goto out; - mutex_lock(&fs_info->chunk_mutex); spin_lock(&block_group->lock); - block_group->removed = 1; + set_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags); + /* - * At this point trimming can't start on this block group, because we - * removed the block group from the tree fs_info->block_group_cache_tree - * so no one can't find it anymore and even if someone already got this - * block group before we removed it from the rbtree, they have already - * incremented block_group->trimming - if they didn't, they won't find - * any free space entries because we already removed them all when we - * called btrfs_remove_free_space_cache(). + * At this point trimming or scrub can't start on this block group, + * because we removed the block group from the rbtree + * fs_info->block_group_cache_tree so no one can't find it anymore and + * even if someone already got this block group before we removed it + * from the rbtree, they have already incremented block_group->frozen - + * if they didn't, for the trimming case they won't find any free space + * entries because we already removed them all when we called + * btrfs_remove_free_space_cache(). * * And we must not remove the extent map from the fs_info->mapping_tree * to prevent the same logical address range and physical device space - * ranges from being reused for a new block group. This is because our - * fs trim operation (btrfs_trim_fs() / btrfs_ioctl_fitrim()) is + * ranges from being reused for a new block group. This is needed to + * avoid races with trimming and scrub. + * + * An fs trim operation (btrfs_trim_fs() / btrfs_ioctl_fitrim()) is * completely transactionless, so while it is trimming a range the * currently running transaction might finish and a new one start, * allowing for new block groups to be created that can reuse the same @@ -1095,28 +1087,9 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, * in place until the extents have been discarded completely when * the transaction commit has completed. */ - remove_em = (atomic_read(&block_group->trimming) == 0); + remove_em = (atomic_read(&block_group->frozen) == 0); spin_unlock(&block_group->lock); - mutex_unlock(&fs_info->chunk_mutex); - - ret = remove_block_group_free_space(trans, block_group); - if (ret) - goto out; - - btrfs_put_block_group(block_group); - btrfs_put_block_group(block_group); - - ret = btrfs_search_slot(trans, root, &key, path, -1, 1); - if (ret > 0) - ret = -EIO; - if (ret < 0) - goto out; - - ret = btrfs_del_item(trans, root, path); - if (ret) - goto out; - if (remove_em) { struct extent_map_tree *em_tree; @@ -1127,7 +1100,10 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, /* once for the tree */ free_extent_map(em); } + out: + /* Once for the lookup reference */ + btrfs_put_block_group(block_group); if (remove_rsv) btrfs_delayed_refs_rsv_release(fs_info, 1); btrfs_free_path(path); @@ -1137,6 +1113,7 @@ out: struct btrfs_trans_handle *btrfs_start_trans_remove_block_group( struct btrfs_fs_info *fs_info, const u64 chunk_offset) { + struct btrfs_root *root = btrfs_block_group_root(fs_info); struct extent_map_tree *em_tree = &fs_info->mapping_tree; struct extent_map *em; struct map_lookup *map; @@ -1170,8 +1147,7 @@ struct btrfs_trans_handle *btrfs_start_trans_remove_block_group( num_items = 3 + map->num_stripes; free_extent_map(em); - return btrfs_start_transaction_fallback_global_rsv(fs_info->extent_root, - num_items, 1); + return btrfs_start_transaction_fallback_global_rsv(root, num_items); } /* @@ -1196,6 +1172,11 @@ static int inc_block_group_ro(struct btrfs_block_group *cache, int force) spin_lock(&sinfo->lock); spin_lock(&cache->lock); + if (cache->swap_extents) { + ret = -ETXTBSY; + goto out; + } + if (cache->ro) { cache->ro++; ret = 0; @@ -1203,7 +1184,7 @@ static int inc_block_group_ro(struct btrfs_block_group *cache, int force) } num_bytes = cache->length - cache->reserved - cache->pinned - - cache->bytes_super - cache->used; + cache->bytes_super - cache->zone_unusable - cache->used; /* * Data never overcommits, even in mixed mode, so do just the straight @@ -1234,6 +1215,12 @@ static int inc_block_group_ro(struct btrfs_block_group *cache, int force) if (!ret) { sinfo->bytes_readonly += num_bytes; + if (btrfs_is_zoned(cache->fs_info)) { + /* Migrate zone_unusable bytes to readonly */ + sinfo->bytes_readonly += cache->zone_unusable; + sinfo->bytes_zone_unusable -= cache->zone_unusable; + cache->zone_unusable = 0; + } cache->ro++; list_add_tail(&cache->ro_list, &sinfo->ro_bgs); } @@ -1248,6 +1235,51 @@ out: return ret; } +static bool clean_pinned_extents(struct btrfs_trans_handle *trans, + struct btrfs_block_group *bg) +{ + struct btrfs_fs_info *fs_info = bg->fs_info; + struct btrfs_transaction *prev_trans = NULL; + const u64 start = bg->start; + const u64 end = start + bg->length - 1; + int ret; + + spin_lock(&fs_info->trans_lock); + if (trans->transaction->list.prev != &fs_info->trans_list) { + prev_trans = list_last_entry(&trans->transaction->list, + struct btrfs_transaction, list); + refcount_inc(&prev_trans->use_count); + } + spin_unlock(&fs_info->trans_lock); + + /* + * Hold the unused_bg_unpin_mutex lock to avoid racing with + * btrfs_finish_extent_commit(). If we are at transaction N, another + * task might be running finish_extent_commit() for the previous + * transaction N - 1, and have seen a range belonging to the block + * group in pinned_extents before we were able to clear the whole block + * group range from pinned_extents. This means that task can lookup for + * the block group after we unpinned it from pinned_extents and removed + * it, leading to a BUG_ON() at unpin_extent_range(). + */ + mutex_lock(&fs_info->unused_bg_unpin_mutex); + if (prev_trans) { + ret = clear_extent_bits(&prev_trans->pinned_extents, start, end, + EXTENT_DIRTY); + if (ret) + goto out; + } + + ret = clear_extent_bits(&trans->transaction->pinned_extents, start, end, + EXTENT_DIRTY); +out: + mutex_unlock(&fs_info->unused_bg_unpin_mutex); + if (prev_trans) + btrfs_put_transaction(prev_trans); + + return ret == 0; +} + /* * Process the unused_bgs list and remove any that don't have any allocated * space inside of them. @@ -1263,9 +1295,18 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) return; + if (btrfs_fs_closing(fs_info)) + return; + + /* + * Long running balances can keep us blocked here for eternity, so + * simply skip deletion if we're unable to get the mutex. + */ + if (!mutex_trylock(&fs_info->reclaim_bgs_lock)) + return; + spin_lock(&fs_info->unused_bgs_lock); while (!list_empty(&fs_info->unused_bgs)) { - u64 start, end; int trimming; block_group = list_first_entry(&fs_info->unused_bgs, @@ -1283,8 +1324,6 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) btrfs_discard_cancel_work(&fs_info->discard_ctl, block_group); - mutex_lock(&fs_info->delete_unused_bgs_mutex); - /* Don't want to race with allocators so take the groups_sem */ down_write(&space_info->groups_sem); @@ -1328,6 +1367,14 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) goto next; } + ret = btrfs_zone_finish(block_group); + if (ret < 0) { + btrfs_dec_block_group_ro(block_group); + if (ret == -EAGAIN) + ret = 0; + goto next; + } + /* * Want to do this before we do anything else so we can recover * properly if we fail to join the transaction. @@ -1344,35 +1391,10 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) * We could have pending pinned extents for this block group, * just delete them, we don't care about them anymore. */ - start = block_group->start; - end = start + block_group->length - 1; - /* - * Hold the unused_bg_unpin_mutex lock to avoid racing with - * btrfs_finish_extent_commit(). If we are at transaction N, - * another task might be running finish_extent_commit() for the - * previous transaction N - 1, and have seen a range belonging - * to the block group in freed_extents[] before we were able to - * clear the whole block group range from freed_extents[]. This - * means that task can lookup for the block group after we - * unpinned it from freed_extents[] and removed it, leading to - * a BUG_ON() at btrfs_unpin_extent_range(). - */ - mutex_lock(&fs_info->unused_bg_unpin_mutex); - ret = clear_extent_bits(&fs_info->freed_extents[0], start, end, - EXTENT_DIRTY); - if (ret) { - mutex_unlock(&fs_info->unused_bg_unpin_mutex); + if (!clean_pinned_extents(trans, block_group)) { btrfs_dec_block_group_ro(block_group); goto end_trans; } - ret = clear_extent_bits(&fs_info->freed_extents[1], start, end, - EXTENT_DIRTY); - if (ret) { - mutex_unlock(&fs_info->unused_bg_unpin_mutex); - btrfs_dec_block_group_ro(block_group); - goto end_trans; - } - mutex_unlock(&fs_info->unused_bg_unpin_mutex); /* * At this point, the block_group is read only and should fail @@ -1398,9 +1420,6 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) btrfs_space_info_update_bytes_pinned(fs_info, space_info, -block_group->pinned); space_info->bytes_readonly += block_group->pinned; - percpu_counter_add_batch(&space_info->total_bytes_pinned, - -block_group->pinned, - BTRFS_TOTAL_BYTES_PINNED_BATCH); block_group->pinned = 0; spin_unlock(&block_group->lock); @@ -1416,12 +1435,16 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) if (!async_trim_enabled && btrfs_test_opt(fs_info, DISCARD_ASYNC)) goto flip_async; - /* DISCARD can flip during remount */ - trimming = btrfs_test_opt(fs_info, DISCARD_SYNC); + /* + * DISCARD can flip during remount. On zoned filesystems, we + * need to reset sequential-required zones. + */ + trimming = btrfs_test_opt(fs_info, DISCARD_SYNC) || + btrfs_is_zoned(fs_info); /* Implicit trim during transaction commit. */ if (trimming) - btrfs_get_block_group_trimming(block_group); + btrfs_freeze_block_group(block_group); /* * Btrfs_remove_chunk will abort the transaction if things go @@ -1431,7 +1454,7 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) if (ret) { if (trimming) - btrfs_put_block_group_trimming(block_group); + btrfs_unfreeze_block_group(block_group); goto end_trans; } @@ -1455,16 +1478,16 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) end_trans: btrfs_end_transaction(trans); next: - mutex_unlock(&fs_info->delete_unused_bgs_mutex); btrfs_put_block_group(block_group); spin_lock(&fs_info->unused_bgs_lock); } spin_unlock(&fs_info->unused_bgs_lock); + mutex_unlock(&fs_info->reclaim_bgs_lock); return; flip_async: btrfs_end_transaction(trans); - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + mutex_unlock(&fs_info->reclaim_bgs_lock); btrfs_put_block_group(block_group); btrfs_discard_punt_unused_bgs_list(fs_info); } @@ -1482,83 +1505,223 @@ void btrfs_mark_bg_unused(struct btrfs_block_group *bg) spin_unlock(&fs_info->unused_bgs_lock); } -static int find_first_block_group(struct btrfs_fs_info *fs_info, - struct btrfs_path *path, - struct btrfs_key *key) +/* + * We want block groups with a low number of used bytes to be in the beginning + * of the list, so they will get reclaimed first. + */ +static int reclaim_bgs_cmp(void *unused, const struct list_head *a, + const struct list_head *b) { - struct btrfs_root *root = fs_info->extent_root; - int ret = 0; - struct btrfs_key found_key; - struct extent_buffer *leaf; + const struct btrfs_block_group *bg1, *bg2; + + bg1 = list_entry(a, struct btrfs_block_group, bg_list); + bg2 = list_entry(b, struct btrfs_block_group, bg_list); + + return bg1->used > bg2->used; +} + +static inline bool btrfs_should_reclaim(struct btrfs_fs_info *fs_info) +{ + if (btrfs_is_zoned(fs_info)) + return btrfs_zoned_should_reclaim(fs_info); + return true; +} + +void btrfs_reclaim_bgs_work(struct work_struct *work) +{ + struct btrfs_fs_info *fs_info = + container_of(work, struct btrfs_fs_info, reclaim_bgs_work); + struct btrfs_block_group *bg; + struct btrfs_space_info *space_info; + + if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) + return; + + if (btrfs_fs_closing(fs_info)) + return; + + if (!btrfs_should_reclaim(fs_info)) + return; + + sb_start_write(fs_info->sb); + + if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { + sb_end_write(fs_info->sb); + return; + } + + /* + * Long running balances can keep us blocked here for eternity, so + * simply skip reclaim if we're unable to get the mutex. + */ + if (!mutex_trylock(&fs_info->reclaim_bgs_lock)) { + btrfs_exclop_finish(fs_info); + sb_end_write(fs_info->sb); + return; + } + + spin_lock(&fs_info->unused_bgs_lock); + /* + * Sort happens under lock because we can't simply splice it and sort. + * The block groups might still be in use and reachable via bg_list, + * and their presence in the reclaim_bgs list must be preserved. + */ + list_sort(NULL, &fs_info->reclaim_bgs, reclaim_bgs_cmp); + while (!list_empty(&fs_info->reclaim_bgs)) { + u64 zone_unusable; + int ret = 0; + + bg = list_first_entry(&fs_info->reclaim_bgs, + struct btrfs_block_group, + bg_list); + list_del_init(&bg->bg_list); + + space_info = bg->space_info; + spin_unlock(&fs_info->unused_bgs_lock); + + /* Don't race with allocators so take the groups_sem */ + down_write(&space_info->groups_sem); + + spin_lock(&bg->lock); + if (bg->reserved || bg->pinned || bg->ro) { + /* + * We want to bail if we made new allocations or have + * outstanding allocations in this block group. We do + * the ro check in case balance is currently acting on + * this block group. + */ + spin_unlock(&bg->lock); + up_write(&space_info->groups_sem); + goto next; + } + spin_unlock(&bg->lock); + + /* Get out fast, in case we're unmounting the filesystem */ + if (btrfs_fs_closing(fs_info)) { + up_write(&space_info->groups_sem); + goto next; + } + + /* + * Cache the zone_unusable value before turning the block group + * to read only. As soon as the blog group is read only it's + * zone_unusable value gets moved to the block group's read-only + * bytes and isn't available for calculations anymore. + */ + zone_unusable = bg->zone_unusable; + ret = inc_block_group_ro(bg, 0); + up_write(&space_info->groups_sem); + if (ret < 0) + goto next; + + btrfs_info(fs_info, + "reclaiming chunk %llu with %llu%% used %llu%% unusable", + bg->start, div_u64(bg->used * 100, bg->length), + div64_u64(zone_unusable * 100, bg->length)); + trace_btrfs_reclaim_block_group(bg); + ret = btrfs_relocate_chunk(fs_info, bg->start); + if (ret) { + btrfs_dec_block_group_ro(bg); + btrfs_err(fs_info, "error relocating chunk %llu", + bg->start); + } + +next: + btrfs_put_block_group(bg); + spin_lock(&fs_info->unused_bgs_lock); + } + spin_unlock(&fs_info->unused_bgs_lock); + mutex_unlock(&fs_info->reclaim_bgs_lock); + btrfs_exclop_finish(fs_info); + sb_end_write(fs_info->sb); +} + +void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info) +{ + spin_lock(&fs_info->unused_bgs_lock); + if (!list_empty(&fs_info->reclaim_bgs)) + queue_work(system_unbound_wq, &fs_info->reclaim_bgs_work); + spin_unlock(&fs_info->unused_bgs_lock); +} + +void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg) +{ + struct btrfs_fs_info *fs_info = bg->fs_info; + + spin_lock(&fs_info->unused_bgs_lock); + if (list_empty(&bg->bg_list)) { + btrfs_get_block_group(bg); + trace_btrfs_add_reclaim_block_group(bg); + list_add_tail(&bg->bg_list, &fs_info->reclaim_bgs); + } + spin_unlock(&fs_info->unused_bgs_lock); +} + +static int read_bg_from_eb(struct btrfs_fs_info *fs_info, struct btrfs_key *key, + struct btrfs_path *path) +{ + struct extent_map_tree *em_tree; + struct extent_map *em; struct btrfs_block_group_item bg; - u64 flags; + struct extent_buffer *leaf; int slot; + u64 flags; + int ret = 0; - ret = btrfs_search_slot(NULL, root, key, path, 0, 0); - if (ret < 0) - goto out; + slot = path->slots[0]; + leaf = path->nodes[0]; - while (1) { - slot = path->slots[0]; - leaf = path->nodes[0]; - if (slot >= btrfs_header_nritems(leaf)) { - ret = btrfs_next_leaf(root, path); - if (ret == 0) - continue; - if (ret < 0) - goto out; - break; - } - btrfs_item_key_to_cpu(leaf, &found_key, slot); + em_tree = &fs_info->mapping_tree; + read_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, key->objectid, key->offset); + read_unlock(&em_tree->lock); + if (!em) { + btrfs_err(fs_info, + "logical %llu len %llu found bg but no related chunk", + key->objectid, key->offset); + return -ENOENT; + } + + if (em->start != key->objectid || em->len != key->offset) { + btrfs_err(fs_info, + "block group %llu len %llu mismatch with chunk %llu len %llu", + key->objectid, key->offset, em->start, em->len); + ret = -EUCLEAN; + goto out_free_em; + } + read_extent_buffer(leaf, &bg, btrfs_item_ptr_offset(leaf, slot), + sizeof(bg)); + flags = btrfs_stack_block_group_flags(&bg) & + BTRFS_BLOCK_GROUP_TYPE_MASK; + + if (flags != (em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK)) { + btrfs_err(fs_info, +"block group %llu len %llu type flags 0x%llx mismatch with chunk type flags 0x%llx", + key->objectid, key->offset, flags, + (BTRFS_BLOCK_GROUP_TYPE_MASK & em->map_lookup->type)); + ret = -EUCLEAN; + } + +out_free_em: + free_extent_map(em); + return ret; +} + +static int find_first_block_group(struct btrfs_fs_info *fs_info, + struct btrfs_path *path, + struct btrfs_key *key) +{ + struct btrfs_root *root = btrfs_block_group_root(fs_info); + int ret; + struct btrfs_key found_key; + + btrfs_for_each_slot(root, key, &found_key, path, ret) { if (found_key.objectid >= key->objectid && found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) { - struct extent_map_tree *em_tree; - struct extent_map *em; - - em_tree = &root->fs_info->mapping_tree; - read_lock(&em_tree->lock); - em = lookup_extent_mapping(em_tree, found_key.objectid, - found_key.offset); - read_unlock(&em_tree->lock); - if (!em) { - btrfs_err(fs_info, - "logical %llu len %llu found bg but no related chunk", - found_key.objectid, found_key.offset); - ret = -ENOENT; - } else if (em->start != found_key.objectid || - em->len != found_key.offset) { - btrfs_err(fs_info, - "block group %llu len %llu mismatch with chunk %llu len %llu", - found_key.objectid, found_key.offset, - em->start, em->len); - ret = -EUCLEAN; - } else { - read_extent_buffer(leaf, &bg, - btrfs_item_ptr_offset(leaf, slot), - sizeof(bg)); - flags = btrfs_stack_block_group_flags(&bg) & - BTRFS_BLOCK_GROUP_TYPE_MASK; - - if (flags != (em->map_lookup->type & - BTRFS_BLOCK_GROUP_TYPE_MASK)) { - btrfs_err(fs_info, -"block group %llu len %llu type flags 0x%llx mismatch with chunk type flags 0x%llx", - found_key.objectid, - found_key.offset, flags, - (BTRFS_BLOCK_GROUP_TYPE_MASK & - em->map_lookup->type)); - ret = -EUCLEAN; - } else { - ret = 0; - } - } - free_extent_map(em); - goto out; + return read_bg_from_eb(fs_info, &found_key, path); } - path->slots[0]++; } -out: return ret; } @@ -1578,8 +1741,11 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) } /** - * btrfs_rmap_block - Map a physical disk address to a list of logical addresses + * Map a physical disk address to a list of logical addresses + * + * @fs_info: the filesystem * @chunk_start: logical address of block group + * @bdev: physical device to resolve, can be NULL to indicate any device * @physical: physical address to map to logical addresses * @logical: return array of logical addresses which map to @physical * @naddrs: length of @logical @@ -1589,9 +1755,9 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) * Used primarily to exclude those portions of a block group that contain super * block copies. */ -EXPORT_FOR_TESTS int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start, - u64 physical, u64 **logical, int *naddrs, int *stripe_len) + struct block_device *bdev, u64 physical, u64 **logical, + int *naddrs, int *stripe_len) { struct extent_map *em; struct map_lookup *map; @@ -1607,19 +1773,13 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start, return -EIO; map = em->map_lookup; - data_stripe_length = em->len; + data_stripe_length = em->orig_block_len; io_stripe_size = map->stripe_len; + chunk_start = em->start; - if (map->type & BTRFS_BLOCK_GROUP_RAID10) - data_stripe_length = div_u64(data_stripe_length, - map->num_stripes / map->sub_stripes); - else if (map->type & BTRFS_BLOCK_GROUP_RAID0) - data_stripe_length = div_u64(data_stripe_length, map->num_stripes); - else if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) { - data_stripe_length = div_u64(data_stripe_length, - nr_data_stripes(map)); + /* For RAID5/6 adjust to a full IO stripe length */ + if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) io_stripe_size = map->stripe_len * nr_data_stripes(map); - } buf = kcalloc(map->num_stripes, sizeof(u64), GFP_NOFS); if (!buf) { @@ -1630,20 +1790,23 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start, for (i = 0; i < map->num_stripes; i++) { bool already_inserted = false; u64 stripe_nr; + u64 offset; int j; if (!in_range(physical, map->stripes[i].physical, data_stripe_length)) continue; + if (bdev && map->stripes[i].dev->bdev != bdev) + continue; + stripe_nr = physical - map->stripes[i].physical; - stripe_nr = div64_u64(stripe_nr, map->stripe_len); + stripe_nr = div64_u64_rem(stripe_nr, map->stripe_len, &offset); - if (map->type & BTRFS_BLOCK_GROUP_RAID10) { + if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID10)) { stripe_nr = stripe_nr * map->num_stripes + i; stripe_nr = div_u64(stripe_nr, map->sub_stripes); - } else if (map->type & BTRFS_BLOCK_GROUP_RAID0) { - stripe_nr = stripe_nr * map->num_stripes + i; } /* * The remaining case would be for RAID56, multiply by @@ -1651,7 +1814,7 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start, * instead of map->stripe_len */ - bytenr = chunk_start + stripe_nr * io_stripe_size; + bytenr = chunk_start + stripe_nr * io_stripe_size + offset; /* Ensure we don't add duplicate addresses */ for (j = 0; j < nr; j++) { @@ -1676,6 +1839,7 @@ out: static int exclude_super_stripes(struct btrfs_block_group *cache) { struct btrfs_fs_info *fs_info = cache->fs_info; + const bool zoned = btrfs_is_zoned(fs_info); u64 bytenr; u64 *logical; int stripe_len; @@ -1692,31 +1856,26 @@ static int exclude_super_stripes(struct btrfs_block_group *cache) for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) { bytenr = btrfs_sb_offset(i); - ret = btrfs_rmap_block(fs_info, cache->start, + ret = btrfs_rmap_block(fs_info, cache->start, NULL, bytenr, &logical, &nr, &stripe_len); if (ret) return ret; - while (nr--) { - u64 start, len; - - if (logical[nr] > cache->start + cache->length) - continue; - - if (logical[nr] + stripe_len <= cache->start) - continue; + /* Shouldn't have super stripes in sequential zones */ + if (zoned && nr) { + btrfs_err(fs_info, + "zoned: block group %llu must not contain super block", + cache->start); + return -EUCLEAN; + } - start = logical[nr]; - if (start < cache->start) { - start = cache->start; - len = (logical[nr] + stripe_len) - start; - } else { - len = min_t(u64, stripe_len, - cache->start + cache->length - start); - } + while (nr--) { + u64 len = min_t(u64, stripe_len, + cache->start + cache->length - logical[nr]); cache->bytes_super += len; - ret = btrfs_add_excluded_extent(fs_info, start, len); + ret = btrfs_add_excluded_extent(fs_info, logical[nr], + len); if (ret) { kfree(logical); return ret; @@ -1728,24 +1887,8 @@ static int exclude_super_stripes(struct btrfs_block_group *cache) return 0; } -static void link_block_group(struct btrfs_block_group *cache) -{ - struct btrfs_space_info *space_info = cache->space_info; - int index = btrfs_bg_flags_to_raid_index(cache->flags); - bool first = false; - - down_write(&space_info->groups_sem); - if (list_empty(&space_info->block_groups[index])) - first = true; - list_add_tail(&cache->list, &space_info->block_groups[index]); - up_write(&space_info->groups_sem); - - if (first) - btrfs_sysfs_add_block_group_type(cache); -} - static struct btrfs_block_group *btrfs_create_block_group_cache( - struct btrfs_fs_info *fs_info, u64 start, u64 size) + struct btrfs_fs_info *fs_info, u64 start) { struct btrfs_block_group *cache; @@ -1761,15 +1904,13 @@ static struct btrfs_block_group *btrfs_create_block_group_cache( } cache->start = start; - cache->length = size; cache->fs_info = fs_info; cache->full_stripe_len = btrfs_full_stripe_len(fs_info, start); - set_free_space_tree_thresholds(cache); cache->discard_index = BTRFS_DISCARD_INDEX_UNUSED; - atomic_set(&cache->count, 1); + refcount_set(&cache->refs, 1); spin_lock_init(&cache->lock); init_rwsem(&cache->data_rwsem); INIT_LIST_HEAD(&cache->list); @@ -1779,10 +1920,12 @@ static struct btrfs_block_group *btrfs_create_block_group_cache( INIT_LIST_HEAD(&cache->discard_list); INIT_LIST_HEAD(&cache->dirty_list); INIT_LIST_HEAD(&cache->io_list); - btrfs_init_free_space_ctl(cache); - atomic_set(&cache->trimming, 0); + INIT_LIST_HEAD(&cache->active_bg_list); + btrfs_init_free_space_ctl(cache, cache->free_space_ctl); + atomic_set(&cache->frozen, 0); mutex_init(&cache->free_space_lock); - btrfs_init_full_stripe_locks_tree(&cache->full_stripe_locks_root); + cache->full_stripe_locks_root.root = RB_ROOT; + mutex_init(&cache->full_stripe_locks_root.lock); return cache; } @@ -1842,24 +1985,27 @@ static int check_chunk_block_group_mappings(struct btrfs_fs_info *fs_info) } static int read_one_block_group(struct btrfs_fs_info *info, - struct btrfs_path *path, + struct btrfs_block_group_item *bgi, const struct btrfs_key *key, int need_clear) { - struct extent_buffer *leaf = path->nodes[0]; struct btrfs_block_group *cache; - struct btrfs_space_info *space_info; - struct btrfs_block_group_item bgi; const bool mixed = btrfs_fs_incompat(info, MIXED_GROUPS); - int slot = path->slots[0]; int ret; ASSERT(key->type == BTRFS_BLOCK_GROUP_ITEM_KEY); - cache = btrfs_create_block_group_cache(info, key->objectid, key->offset); + cache = btrfs_create_block_group_cache(info, key->objectid); if (!cache) return -ENOMEM; + cache->length = key->offset; + cache->used = btrfs_stack_block_group_used(bgi); + cache->flags = btrfs_stack_block_group_flags(bgi); + cache->global_root_id = btrfs_stack_block_group_chunk_objectid(bgi); + + set_free_space_tree_thresholds(cache); + if (need_clear) { /* * When we mount with old space cache, we need to @@ -1874,10 +2020,6 @@ static int read_one_block_group(struct btrfs_fs_info *info, if (btrfs_test_opt(info, SPACE_CACHE)) cache->disk_cache_state = BTRFS_DC_CLEAR; } - read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot), - sizeof(bgi)); - cache->used = btrfs_stack_block_group_used(&bgi); - cache->flags = btrfs_stack_block_group_flags(&bgi); if (!mixed && ((cache->flags & BTRFS_BLOCK_GROUP_METADATA) && (cache->flags & BTRFS_BLOCK_GROUP_DATA))) { btrfs_err(info, @@ -1887,6 +2029,13 @@ static int read_one_block_group(struct btrfs_fs_info *info, goto error; } + ret = btrfs_load_block_group_zone_info(cache, false); + if (ret) { + btrfs_err(info, "zoned: failed to load zone info of bg %llu", + cache->start); + goto error; + } + /* * We need to exclude the super stripes now so that the space info has * super bytes accounted for, otherwise we'll think we have more space @@ -1900,20 +2049,28 @@ static int read_one_block_group(struct btrfs_fs_info *info, } /* - * Check for two cases, either we are full, and therefore don't need - * to bother with the caching work since we won't find any space, or we - * are empty, and we can just add all the space in and be done with it. - * This saves us _a_lot_ of time, particularly in the full case. + * For zoned filesystem, space after the allocation offset is the only + * free space for a block group. So, we don't need any caching work. + * btrfs_calc_zone_unusable() will set the amount of free space and + * zone_unusable space. + * + * For regular filesystem, check for two cases, either we are full, and + * therefore don't need to bother with the caching work since we won't + * find any space, or we are empty, and we can just add all the space + * in and be done with it. This saves us _a_lot_ of time, particularly + * in the full case. */ - if (key->offset == cache->used) { - cache->last_byte_to_unpin = (u64)-1; + if (btrfs_is_zoned(info)) { + btrfs_calc_zone_unusable(cache); + /* Should not have any excluded extents. Just in case, though. */ + btrfs_free_excluded_extents(cache); + } else if (cache->length == cache->used) { cache->cached = BTRFS_CACHE_FINISHED; btrfs_free_excluded_extents(cache); } else if (cache->used == 0) { - cache->last_byte_to_unpin = (u64)-1; cache->cached = BTRFS_CACHE_FINISHED; - add_new_free_space(cache, key->objectid, - key->objectid + key->offset); + add_new_free_space(cache, cache->start, + cache->start + cache->length); btrfs_free_excluded_extents(cache); } @@ -1923,31 +2080,81 @@ static int read_one_block_group(struct btrfs_fs_info *info, goto error; } trace_btrfs_add_block_group(info, cache, 0); - btrfs_update_space_info(info, cache->flags, key->offset, - cache->used, cache->bytes_super, &space_info); - - cache->space_info = space_info; - - link_block_group(cache); + btrfs_add_bg_to_space_info(info, cache); set_avail_alloc_bits(info, cache->flags); - if (btrfs_chunk_readonly(info, cache->start)) { + if (btrfs_chunk_writeable(info, cache->start)) { + if (cache->used == 0) { + ASSERT(list_empty(&cache->bg_list)); + if (btrfs_test_opt(info, DISCARD_ASYNC)) + btrfs_discard_queue_work(&info->discard_ctl, cache); + else + btrfs_mark_bg_unused(cache); + } + } else { inc_block_group_ro(cache, 1); - } else if (cache->used == 0) { - ASSERT(list_empty(&cache->bg_list)); - if (btrfs_test_opt(info, DISCARD_ASYNC)) - btrfs_discard_queue_work(&info->discard_ctl, cache); - else - btrfs_mark_bg_unused(cache); } + return 0; error: btrfs_put_block_group(cache); return ret; } +static int fill_dummy_bgs(struct btrfs_fs_info *fs_info) +{ + struct extent_map_tree *em_tree = &fs_info->mapping_tree; + struct rb_node *node; + int ret = 0; + + for (node = rb_first_cached(&em_tree->map); node; node = rb_next(node)) { + struct extent_map *em; + struct map_lookup *map; + struct btrfs_block_group *bg; + + em = rb_entry(node, struct extent_map, rb_node); + map = em->map_lookup; + bg = btrfs_create_block_group_cache(fs_info, em->start); + if (!bg) { + ret = -ENOMEM; + break; + } + + /* Fill dummy cache as FULL */ + bg->length = em->len; + bg->flags = map->type; + bg->cached = BTRFS_CACHE_FINISHED; + bg->used = em->len; + bg->flags = map->type; + ret = btrfs_add_block_group_cache(fs_info, bg); + /* + * We may have some valid block group cache added already, in + * that case we skip to the next one. + */ + if (ret == -EEXIST) { + ret = 0; + btrfs_put_block_group(bg); + continue; + } + + if (ret) { + btrfs_remove_free_space_cache(bg); + btrfs_put_block_group(bg); + break; + } + + btrfs_add_bg_to_space_info(fs_info, bg); + + set_avail_alloc_bits(fs_info, bg->flags); + } + if (!ret) + btrfs_init_global_block_rsv(fs_info); + return ret; +} + int btrfs_read_block_groups(struct btrfs_fs_info *info) { + struct btrfs_root *root = btrfs_block_group_root(info); struct btrfs_path *path; int ret; struct btrfs_block_group *cache; @@ -1956,13 +2163,24 @@ int btrfs_read_block_groups(struct btrfs_fs_info *info) int need_clear = 0; u64 cache_gen; + /* + * Either no extent root (with ibadroots rescue option) or we have + * unsupported RO options. The fs can never be mounted read-write, so no + * need to waste time searching block group items. + * + * This also allows new extent tree related changes to be RO compat, + * no need for a full incompat flag. + */ + if (!root || (btrfs_super_compat_ro_flags(info->super_copy) & + ~BTRFS_FEATURE_COMPAT_RO_SUPP)) + return fill_dummy_bgs(info); + key.objectid = 0; key.offset = 0; key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; path = btrfs_alloc_path(); if (!path) return -ENOMEM; - path->reada = READA_FORWARD; cache_gen = btrfs_super_cache_generation(info->super_copy); if (btrfs_test_opt(info, SPACE_CACHE) && @@ -1972,22 +2190,44 @@ int btrfs_read_block_groups(struct btrfs_fs_info *info) need_clear = 1; while (1) { + struct btrfs_block_group_item bgi; + struct extent_buffer *leaf; + int slot; + ret = find_first_block_group(info, path, &key); if (ret > 0) break; if (ret != 0) goto error; - btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); - ret = read_one_block_group(info, path, &key, need_clear); + leaf = path->nodes[0]; + slot = path->slots[0]; + + read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot), + sizeof(bgi)); + + btrfs_item_key_to_cpu(leaf, &key, slot); + btrfs_release_path(path); + ret = read_one_block_group(info, &bgi, &key, need_clear); if (ret < 0) goto error; key.objectid += key.offset; key.offset = 0; - btrfs_release_path(path); } + btrfs_release_path(path); + + list_for_each_entry(space_info, &info->space_info, list) { + int i; + + for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) { + if (list_empty(&space_info->block_groups[i])) + continue; + cache = list_first_entry(&space_info->block_groups[i], + struct btrfs_block_group, + list); + btrfs_sysfs_add_block_group_type(cache); + } - list_for_each_entry_rcu(space_info, &info->space_info, list) { if (!(btrfs_get_alloc_profile(info, space_info->flags) & (BTRFS_BLOCK_GROUP_RAID10 | BTRFS_BLOCK_GROUP_RAID1_MASK | @@ -2012,46 +2252,184 @@ int btrfs_read_block_groups(struct btrfs_fs_info *info) ret = check_chunk_block_group_mappings(info); error: btrfs_free_path(path); + /* + * We've hit some error while reading the extent tree, and have + * rescue=ibadroots mount option. + * Try to fill the tree using dummy block groups so that the user can + * continue to mount and grab their data. + */ + if (ret && btrfs_test_opt(info, IGNOREBADROOTS)) + ret = fill_dummy_bgs(info); return ret; } -void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans) +/* + * This function, insert_block_group_item(), belongs to the phase 2 of chunk + * allocation. + * + * See the comment at btrfs_chunk_alloc() for details about the chunk allocation + * phases. + */ +static int insert_block_group_item(struct btrfs_trans_handle *trans, + struct btrfs_block_group *block_group) { struct btrfs_fs_info *fs_info = trans->fs_info; - struct btrfs_block_group *block_group; - struct btrfs_root *extent_root = fs_info->extent_root; - struct btrfs_block_group_item item; + struct btrfs_block_group_item bgi; + struct btrfs_root *root = btrfs_block_group_root(fs_info); + struct btrfs_key key; + + spin_lock(&block_group->lock); + btrfs_set_stack_block_group_used(&bgi, block_group->used); + btrfs_set_stack_block_group_chunk_objectid(&bgi, + block_group->global_root_id); + btrfs_set_stack_block_group_flags(&bgi, block_group->flags); + key.objectid = block_group->start; + key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; + key.offset = block_group->length; + spin_unlock(&block_group->lock); + + return btrfs_insert_item(trans, root, &key, &bgi, sizeof(bgi)); +} + +static int insert_dev_extent(struct btrfs_trans_handle *trans, + struct btrfs_device *device, u64 chunk_offset, + u64 start, u64 num_bytes) +{ + struct btrfs_fs_info *fs_info = device->fs_info; + struct btrfs_root *root = fs_info->dev_root; + struct btrfs_path *path; + struct btrfs_dev_extent *extent; + struct extent_buffer *leaf; struct btrfs_key key; + int ret; + + WARN_ON(!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &device->dev_state)); + WARN_ON(test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)); + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + key.objectid = device->devid; + key.type = BTRFS_DEV_EXTENT_KEY; + key.offset = start; + ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*extent)); + if (ret) + goto out; + + leaf = path->nodes[0]; + extent = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_extent); + btrfs_set_dev_extent_chunk_tree(leaf, extent, BTRFS_CHUNK_TREE_OBJECTID); + btrfs_set_dev_extent_chunk_objectid(leaf, extent, + BTRFS_FIRST_CHUNK_TREE_OBJECTID); + btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset); + + btrfs_set_dev_extent_length(leaf, extent, num_bytes); + btrfs_mark_buffer_dirty(leaf); +out: + btrfs_free_path(path); + return ret; +} + +/* + * This function belongs to phase 2. + * + * See the comment at btrfs_chunk_alloc() for details about the chunk allocation + * phases. + */ +static int insert_dev_extents(struct btrfs_trans_handle *trans, + u64 chunk_offset, u64 chunk_size) +{ + struct btrfs_fs_info *fs_info = trans->fs_info; + struct btrfs_device *device; + struct extent_map *em; + struct map_lookup *map; + u64 dev_offset; + u64 stripe_size; + int i; int ret = 0; - if (!trans->can_flush_pending_bgs) - return; + em = btrfs_get_chunk_map(fs_info, chunk_offset, chunk_size); + if (IS_ERR(em)) + return PTR_ERR(em); + + map = em->map_lookup; + stripe_size = em->orig_block_len; + + /* + * Take the device list mutex to prevent races with the final phase of + * a device replace operation that replaces the device object associated + * with the map's stripes, because the device object's id can change + * at any time during that final phase of the device replace operation + * (dev-replace.c:btrfs_dev_replace_finishing()), so we could grab the + * replaced device and then see it with an ID of BTRFS_DEV_REPLACE_DEVID, + * resulting in persisting a device extent item with such ID. + */ + mutex_lock(&fs_info->fs_devices->device_list_mutex); + for (i = 0; i < map->num_stripes; i++) { + device = map->stripes[i].dev; + dev_offset = map->stripes[i].physical; + + ret = insert_dev_extent(trans, device, chunk_offset, dev_offset, + stripe_size); + if (ret) + break; + } + mutex_unlock(&fs_info->fs_devices->device_list_mutex); + + free_extent_map(em); + return ret; +} + +/* + * This function, btrfs_create_pending_block_groups(), belongs to the phase 2 of + * chunk allocation. + * + * See the comment at btrfs_chunk_alloc() for details about the chunk allocation + * phases. + */ +void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans) +{ + struct btrfs_fs_info *fs_info = trans->fs_info; + struct btrfs_block_group *block_group; + int ret = 0; while (!list_empty(&trans->new_bgs)) { + int index; + block_group = list_first_entry(&trans->new_bgs, struct btrfs_block_group, bg_list); if (ret) goto next; - spin_lock(&block_group->lock); - btrfs_set_stack_block_group_used(&item, block_group->used); - btrfs_set_stack_block_group_chunk_objectid(&item, - BTRFS_FIRST_CHUNK_TREE_OBJECTID); - btrfs_set_stack_block_group_flags(&item, block_group->flags); - key.objectid = block_group->start; - key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; - key.offset = block_group->length; - spin_unlock(&block_group->lock); + index = btrfs_bg_flags_to_raid_index(block_group->flags); - ret = btrfs_insert_item(trans, extent_root, &key, &item, - sizeof(item)); + ret = insert_block_group_item(trans, block_group); if (ret) btrfs_abort_transaction(trans, ret); - ret = btrfs_finish_chunk_alloc(trans, key.objectid, key.offset); + if (!test_bit(BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED, + &block_group->runtime_flags)) { + mutex_lock(&fs_info->chunk_mutex); + ret = btrfs_chunk_alloc_add_chunk_item(trans, block_group); + mutex_unlock(&fs_info->chunk_mutex); + if (ret) + btrfs_abort_transaction(trans, ret); + } + ret = insert_dev_extents(trans, block_group->start, + block_group->length); if (ret) btrfs_abort_transaction(trans, ret); add_block_group_free_space(trans, block_group); + + /* + * If we restriped during balance, we may have added a new raid + * type, so now add the sysfs entries when it is safe to do so. + * We don't have to worry about locking here as it's handled in + * btrfs_sysfs_add_block_group_type. + */ + if (block_group->space_info->block_group_kobjs[index] == NULL) + btrfs_sysfs_add_block_group_type(block_group); + /* Already aborted the transaction if it failed. */ next: btrfs_delayed_refs_rsv_release(fs_info, 1); @@ -2060,8 +2438,30 @@ next: btrfs_trans_release_chunk_metadata(trans); } -int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used, - u64 type, u64 chunk_offset, u64 size) +/* + * For extent tree v2 we use the block_group_item->chunk_offset to point at our + * global root id. For v1 it's always set to BTRFS_FIRST_CHUNK_TREE_OBJECTID. + */ +static u64 calculate_global_root_id(struct btrfs_fs_info *fs_info, u64 offset) +{ + u64 div = SZ_1G; + u64 index; + + if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) + return BTRFS_FIRST_CHUNK_TREE_OBJECTID; + + /* If we have a smaller fs index based on 128MiB. */ + if (btrfs_super_total_bytes(fs_info->super_copy) <= (SZ_1G * 10ULL)) + div = SZ_128M; + + offset = div64_u64(offset, div); + div64_u64_rem(offset, fs_info->nr_global_roots, &index); + return index; +} + +struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans, + u64 bytes_used, u64 type, + u64 chunk_offset, u64 size) { struct btrfs_fs_info *fs_info = trans->fs_info; struct btrfs_block_group *cache; @@ -2069,35 +2469,38 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used, btrfs_set_log_full_commit(trans); - cache = btrfs_create_block_group_cache(fs_info, chunk_offset, size); + cache = btrfs_create_block_group_cache(fs_info, chunk_offset); if (!cache) - return -ENOMEM; + return ERR_PTR(-ENOMEM); + cache->length = size; + set_free_space_tree_thresholds(cache); cache->used = bytes_used; cache->flags = type; - cache->last_byte_to_unpin = (u64)-1; cache->cached = BTRFS_CACHE_FINISHED; - cache->needs_free_space = 1; + cache->global_root_id = calculate_global_root_id(fs_info, cache->start); + + if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) + cache->needs_free_space = 1; + + ret = btrfs_load_block_group_zone_info(cache, true); + if (ret) { + btrfs_put_block_group(cache); + return ERR_PTR(ret); + } + ret = exclude_super_stripes(cache); if (ret) { /* We may have excluded something, so call this just in case */ btrfs_free_excluded_extents(cache); btrfs_put_block_group(cache); - return ret; + return ERR_PTR(ret); } add_new_free_space(cache, chunk_offset, chunk_offset + size); btrfs_free_excluded_extents(cache); -#ifdef CONFIG_BTRFS_DEBUG - if (btrfs_should_fragment_free_space(cache)) { - u64 new_bytes_used = size - bytes_used; - - bytes_used += new_bytes_used >> 1; - fragment_free_space(cache); - } -#endif /* * Ensure the corresponding space_info object is created and * assigned to our block group. We want our bg to be added to the rbtree @@ -2110,7 +2513,7 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used, if (ret) { btrfs_remove_free_space_cache(cache); btrfs_put_block_group(cache); - return ret; + return ERR_PTR(ret); } /* @@ -2118,66 +2521,24 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used, * the rbtree, update the space info's counters. */ trace_btrfs_add_block_group(fs_info, cache, 1); - btrfs_update_space_info(fs_info, cache->flags, size, bytes_used, - cache->bytes_super, &cache->space_info); + btrfs_add_bg_to_space_info(fs_info, cache); btrfs_update_global_block_rsv(fs_info); - link_block_group(cache); +#ifdef CONFIG_BTRFS_DEBUG + if (btrfs_should_fragment_free_space(cache)) { + u64 new_bytes_used = size - bytes_used; + + cache->space_info->bytes_used += new_bytes_used >> 1; + fragment_free_space(cache); + } +#endif list_add_tail(&cache->bg_list, &trans->new_bgs); trans->delayed_ref_updates++; btrfs_update_delayed_refs_rsv(trans); set_avail_alloc_bits(fs_info, type); - return 0; -} - -static u64 update_block_group_flags(struct btrfs_fs_info *fs_info, u64 flags) -{ - u64 num_devices; - u64 stripped; - - /* - * if restripe for this chunk_type is on pick target profile and - * return, otherwise do the usual balance - */ - stripped = get_restripe_target(fs_info, flags); - if (stripped) - return extended_to_chunk(stripped); - - num_devices = fs_info->fs_devices->rw_devices; - - stripped = BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID56_MASK | - BTRFS_BLOCK_GROUP_RAID1_MASK | BTRFS_BLOCK_GROUP_RAID10; - - if (num_devices == 1) { - stripped |= BTRFS_BLOCK_GROUP_DUP; - stripped = flags & ~stripped; - - /* turn raid0 into single device chunks */ - if (flags & BTRFS_BLOCK_GROUP_RAID0) - return stripped; - - /* turn mirroring into duplication */ - if (flags & (BTRFS_BLOCK_GROUP_RAID1_MASK | - BTRFS_BLOCK_GROUP_RAID10)) - return stripped | BTRFS_BLOCK_GROUP_DUP; - } else { - /* they already had raid on here, just return */ - if (flags & stripped) - return flags; - - stripped |= BTRFS_BLOCK_GROUP_DUP; - stripped = flags & ~stripped; - - /* switch duplicated blocks with raid1 */ - if (flags & BTRFS_BLOCK_GROUP_DUP) - return stripped | BTRFS_BLOCK_GROUP_RAID1; - - /* this is drive concat, leave it alone */ - } - - return flags; + return cache; } /* @@ -2194,38 +2555,56 @@ int btrfs_inc_block_group_ro(struct btrfs_block_group *cache, { struct btrfs_fs_info *fs_info = cache->fs_info; struct btrfs_trans_handle *trans; + struct btrfs_root *root = btrfs_block_group_root(fs_info); u64 alloc_flags; int ret; - -again: - trans = btrfs_join_transaction(fs_info->extent_root); - if (IS_ERR(trans)) - return PTR_ERR(trans); + bool dirty_bg_running; /* - * we're not allowed to set block groups readonly after the dirty - * block groups cache has started writing. If it already started, - * back off and let this transaction commit + * This can only happen when we are doing read-only scrub on read-only + * mount. + * In that case we should not start a new transaction on read-only fs. + * Thus here we skip all chunk allocations. */ - mutex_lock(&fs_info->ro_block_group_mutex); - if (test_bit(BTRFS_TRANS_DIRTY_BG_RUN, &trans->transaction->flags)) { - u64 transid = trans->transid; - + if (sb_rdonly(fs_info->sb)) { + mutex_lock(&fs_info->ro_block_group_mutex); + ret = inc_block_group_ro(cache, 0); mutex_unlock(&fs_info->ro_block_group_mutex); - btrfs_end_transaction(trans); - - ret = btrfs_wait_for_commit(fs_info, transid); - if (ret) - return ret; - goto again; + return ret; } + do { + trans = btrfs_join_transaction(root); + if (IS_ERR(trans)) + return PTR_ERR(trans); + + dirty_bg_running = false; + + /* + * We're not allowed to set block groups readonly after the dirty + * block group cache has started writing. If it already started, + * back off and let this transaction commit. + */ + mutex_lock(&fs_info->ro_block_group_mutex); + if (test_bit(BTRFS_TRANS_DIRTY_BG_RUN, &trans->transaction->flags)) { + u64 transid = trans->transid; + + mutex_unlock(&fs_info->ro_block_group_mutex); + btrfs_end_transaction(trans); + + ret = btrfs_wait_for_commit(fs_info, transid); + if (ret) + return ret; + dirty_bg_running = true; + } + } while (dirty_bg_running); + if (do_chunk_alloc) { /* * If we are changing raid levels, try to allocate a * corresponding block group with the new raid level. */ - alloc_flags = update_block_group_flags(fs_info, cache->flags); + alloc_flags = btrfs_get_alloc_profile(fs_info, cache->flags); if (alloc_flags != cache->flags) { ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE); @@ -2241,7 +2620,7 @@ again: } ret = inc_block_group_ro(cache, 0); - if (!do_chunk_alloc) + if (!do_chunk_alloc || ret == -ETXTBSY) goto unlock_out; if (!ret) goto out; @@ -2249,10 +2628,20 @@ again: ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE); if (ret < 0) goto out; + /* + * We have allocated a new chunk. We also need to activate that chunk to + * grant metadata tickets for zoned filesystem. + */ + ret = btrfs_zoned_activate_one_bg(fs_info, cache->space_info, true); + if (ret < 0) + goto out; + ret = inc_block_group_ro(cache, 0); + if (ret == -ETXTBSY) + goto unlock_out; out: if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { - alloc_flags = update_block_group_flags(fs_info, cache->flags); + alloc_flags = btrfs_get_alloc_profile(fs_info, cache->flags); mutex_lock(&fs_info->chunk_mutex); check_system_chunk(trans, alloc_flags); mutex_unlock(&fs_info->chunk_mutex); @@ -2274,8 +2663,17 @@ void btrfs_dec_block_group_ro(struct btrfs_block_group *cache) spin_lock(&sinfo->lock); spin_lock(&cache->lock); if (!--cache->ro) { + if (btrfs_is_zoned(cache->fs_info)) { + /* Migrate zone_unusable bytes back */ + cache->zone_unusable = + (cache->alloc_offset - cache->used) + + (cache->length - cache->zone_capacity); + sinfo->bytes_zone_unusable += cache->zone_unusable; + sinfo->bytes_readonly -= cache->zone_unusable; + } num_bytes = cache->length - cache->reserved - - cache->pinned - cache->bytes_super - cache->used; + cache->pinned - cache->bytes_super - + cache->zone_unusable - cache->used; sinfo->bytes_readonly -= num_bytes; list_del_init(&cache->ro_list); } @@ -2283,13 +2681,13 @@ void btrfs_dec_block_group_ro(struct btrfs_block_group *cache) spin_unlock(&sinfo->lock); } -static int write_one_cache_group(struct btrfs_trans_handle *trans, - struct btrfs_path *path, - struct btrfs_block_group *cache) +static int update_block_group_item(struct btrfs_trans_handle *trans, + struct btrfs_path *path, + struct btrfs_block_group *cache) { struct btrfs_fs_info *fs_info = trans->fs_info; int ret; - struct btrfs_root *extent_root = fs_info->extent_root; + struct btrfs_root *root = btrfs_block_group_root(fs_info); unsigned long bi; struct extent_buffer *leaf; struct btrfs_block_group_item bgi; @@ -2299,7 +2697,7 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans, key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; key.offset = cache->length; - ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 1); + ret = btrfs_search_slot(trans, root, &key, path, 0, 1); if (ret) { if (ret > 0) ret = -ENOENT; @@ -2310,7 +2708,7 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans, bi = btrfs_item_ptr_offset(leaf, path->slots[0]); btrfs_set_stack_block_group_used(&bgi, cache->used); btrfs_set_stack_block_group_chunk_objectid(&bgi, - BTRFS_FIRST_CHUNK_TREE_OBJECTID); + cache->global_root_id); btrfs_set_stack_block_group_flags(&bgi, cache->flags); write_extent_buffer(leaf, &bgi, bi, sizeof(bgi)); btrfs_mark_buffer_dirty(leaf); @@ -2330,10 +2728,13 @@ static int cache_save_setup(struct btrfs_block_group *block_group, struct extent_changeset *data_reserved = NULL; u64 alloc_hint = 0; int dcs = BTRFS_DC_ERROR; - u64 num_pages = 0; + u64 cache_size = 0; int retries = 0; int ret = 0; + if (!btrfs_test_opt(fs_info, SPACE_CACHE)) + return 0; + /* * If this block group is smaller than 100 megs don't bother caching the * block group. @@ -2345,7 +2746,7 @@ static int cache_save_setup(struct btrfs_block_group *block_group, return 0; } - if (trans->aborted) + if (TRANS_ABORTED(trans)) return 0; again: inode = lookup_free_space_inode(block_group, path); @@ -2374,7 +2775,7 @@ again: * time. */ BTRFS_I(inode)->generation = 0; - ret = btrfs_update_inode(trans, root, inode); + ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); if (ret) { /* * So theoretically we could recover from this, simply set the @@ -2439,19 +2840,20 @@ again: * taking up quite a bit since it's not folded into the other space * cache. */ - num_pages = div_u64(block_group->length, SZ_256M); - if (!num_pages) - num_pages = 1; + cache_size = div_u64(block_group->length, SZ_256M); + if (!cache_size) + cache_size = 1; - num_pages *= 16; - num_pages *= PAGE_SIZE; + cache_size *= 16; + cache_size *= fs_info->sectorsize; - ret = btrfs_check_data_free_space(inode, &data_reserved, 0, num_pages); + ret = btrfs_check_data_free_space(BTRFS_I(inode), &data_reserved, 0, + cache_size, false); if (ret) goto out_put; - ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages, - num_pages, num_pages, + ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, cache_size, + cache_size, cache_size, &alloc_hint); /* * Our cache requires contiguous chunks so that we don't modify a bunch @@ -2529,7 +2931,6 @@ int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans) struct btrfs_path *path = NULL; LIST_HEAD(dirty); struct list_head *io = &cur_trans->io_bgs; - int num_started = 0; int loops = 0; spin_lock(&cur_trans->dirty_bgs_lock); @@ -2546,8 +2947,10 @@ again: if (!path) { path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; + if (!path) { + ret = -ENOMEM; + goto out; + } } /* @@ -2593,7 +2996,6 @@ again: cache->io_ctl.inode = NULL; ret = btrfs_write_out_cache(trans, cache, path); if (ret == 0 && cache->io_ctl.inode) { - num_started++; should_put = 0; /* @@ -2611,7 +3013,7 @@ again: } } if (!ret) { - ret = write_one_cache_group(trans, path, cache); + ret = update_block_group_item(trans, path, cache); /* * Our block group might still be attached to the list * of new block groups in the transaction handle of some @@ -2641,16 +3043,14 @@ again: btrfs_put_block_group(cache); if (drop_reserve) btrfs_delayed_refs_rsv_release(fs_info, 1); - - if (ret) - break; - /* * Avoid blocking other tasks for too long. It might even save * us from writing caches for block groups that are going to be * removed. */ mutex_unlock(&trans->transaction->cache_write_mutex); + if (ret) + goto out; mutex_lock(&trans->transaction->cache_write_mutex); } mutex_unlock(&trans->transaction->cache_write_mutex); @@ -2659,7 +3059,8 @@ again: * Go through delayed refs for all the stuff we've just kicked off * and then loop back (just once) */ - ret = btrfs_run_delayed_refs(trans, 0); + if (!ret) + ret = btrfs_run_delayed_refs(trans, 0); if (!ret && loops == 0) { loops++; spin_lock(&cur_trans->dirty_bgs_lock); @@ -2673,7 +3074,12 @@ again: goto again; } spin_unlock(&cur_trans->dirty_bgs_lock); - } else if (ret < 0) { + } +out: + if (ret < 0) { + spin_lock(&cur_trans->dirty_bgs_lock); + list_splice_init(&dirty, &cur_trans->dirty_bgs); + spin_unlock(&cur_trans->dirty_bgs_lock); btrfs_cleanup_dirty_bgs(cur_trans, fs_info); } @@ -2690,7 +3096,6 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans) int should_put; struct btrfs_path *path; struct list_head *io = &cur_trans->io_bgs; - int num_started = 0; path = btrfs_alloc_path(); if (!path) @@ -2748,7 +3153,6 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans) cache->io_ctl.inode = NULL; ret = btrfs_write_out_cache(trans, cache, path); if (ret == 0 && cache->io_ctl.inode) { - num_started++; should_put = 0; list_add_tail(&cache->io_list, io); } else { @@ -2760,7 +3164,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans) } } if (!ret) { - ret = write_one_cache_group(trans, path, cache); + ret = update_block_group_item(trans, path, cache); /* * One of the free space endio workers might have * created a new block group while updating a free space @@ -2771,13 +3175,13 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans) * finished yet (no block group item in the extent tree * yet, etc). If this is the case, wait for all free * space endio workers to finish and retry. This is a - * a very rare case so no need for a more efficient and + * very rare case so no need for a more efficient and * complex approach. */ if (ret == -ENOENT) { wait_event(cur_trans->writer_wait, atomic_read(&cur_trans->num_writers) == 1); - ret = write_one_cache_group(trans, path, cache); + ret = update_block_group_item(trans, path, cache); } if (ret) btrfs_abort_transaction(trans, ret); @@ -2807,8 +3211,33 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans) return ret; } +static inline bool should_reclaim_block_group(struct btrfs_block_group *bg, + u64 bytes_freed) +{ + const struct btrfs_space_info *space_info = bg->space_info; + const int reclaim_thresh = READ_ONCE(space_info->bg_reclaim_threshold); + const u64 new_val = bg->used; + const u64 old_val = new_val + bytes_freed; + u64 thresh; + + if (reclaim_thresh == 0) + return false; + + thresh = div_factor_fine(bg->length, reclaim_thresh); + + /* + * If we were below the threshold before don't reclaim, we are likely a + * brand new block group and we don't want to relocate new block groups. + */ + if (old_val < thresh) + return false; + if (new_val >= thresh) + return false; + return true; +} + int btrfs_update_block_group(struct btrfs_trans_handle *trans, - u64 bytenr, u64 num_bytes, int alloc) + u64 bytenr, u64 num_bytes, bool alloc) { struct btrfs_fs_info *info = trans->fs_info; struct btrfs_block_group *cache = NULL; @@ -2829,6 +3258,8 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans, spin_unlock(&info->delalloc_root_lock); while (total) { + bool reclaim; + cache = btrfs_lookup_block_group(info, bytenr); if (!cache) { ret = -ENOENT; @@ -2843,7 +3274,7 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans, * space back to the block group, otherwise we will leak space. */ if (!alloc && !btrfs_block_group_done(cache)) - btrfs_cache_block_group(cache, 1); + btrfs_cache_block_group(cache, true); byte_in_group = bytenr - cache->start; WARN_ON(byte_in_group > cache->length); @@ -2874,14 +3305,12 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans, cache->space_info, num_bytes); cache->space_info->bytes_used -= num_bytes; cache->space_info->disk_used -= num_bytes * factor; + + reclaim = should_reclaim_block_group(cache, num_bytes); spin_unlock(&cache->lock); spin_unlock(&cache->space_info->lock); - percpu_counter_add_batch( - &cache->space_info->total_bytes_pinned, - num_bytes, - BTRFS_TOTAL_BYTES_PINNED_BATCH); - set_extent_dirty(info->pinned_extents, + set_extent_dirty(&trans->transaction->pinned_extents, bytenr, bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL); } @@ -2904,6 +3333,8 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans, if (!alloc && old_val == 0) { if (!btrfs_test_opt(info, DISCARD_ASYNC)) btrfs_mark_bg_unused(cache); + } else if (!alloc && reclaim) { + btrfs_mark_bg_to_reclaim(cache); } btrfs_put_block_group(cache); @@ -2947,6 +3378,13 @@ int btrfs_add_reserved_bytes(struct btrfs_block_group *cache, space_info, -ram_bytes); if (delalloc) cache->delalloc_bytes += num_bytes; + + /* + * Compression can use less space than we reserved, so wake + * tickets if that happens + */ + if (num_bytes < ram_bytes) + btrfs_try_granting_tickets(cache->fs_info, space_info); } spin_unlock(&cache->lock); spin_unlock(&space_info->lock); @@ -2980,6 +3418,8 @@ void btrfs_free_reserved_bytes(struct btrfs_block_group *cache, if (delalloc) cache->delalloc_bytes -= num_bytes; spin_unlock(&cache->lock); + + btrfs_try_granting_tickets(cache->fs_info, space_info); spin_unlock(&space_info->lock); } @@ -2988,12 +3428,10 @@ static void force_metadata_allocation(struct btrfs_fs_info *info) struct list_head *head = &info->space_info; struct btrfs_space_info *found; - rcu_read_lock(); - list_for_each_entry_rcu(found, head, list) { + list_for_each_entry(found, head, list) { if (found->flags & BTRFS_BLOCK_GROUP_METADATA) found->force_alloc = CHUNK_ALLOC_FORCE; } - rcu_read_unlock(); } static int should_alloc_chunk(struct btrfs_fs_info *fs_info, @@ -3029,11 +3467,203 @@ int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type) return btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE); } +static struct btrfs_block_group *do_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags) +{ + struct btrfs_block_group *bg; + int ret; + + /* + * Check if we have enough space in the system space info because we + * will need to update device items in the chunk btree and insert a new + * chunk item in the chunk btree as well. This will allocate a new + * system block group if needed. + */ + check_system_chunk(trans, flags); + + bg = btrfs_create_chunk(trans, flags); + if (IS_ERR(bg)) { + ret = PTR_ERR(bg); + goto out; + } + + ret = btrfs_chunk_alloc_add_chunk_item(trans, bg); + /* + * Normally we are not expected to fail with -ENOSPC here, since we have + * previously reserved space in the system space_info and allocated one + * new system chunk if necessary. However there are three exceptions: + * + * 1) We may have enough free space in the system space_info but all the + * existing system block groups have a profile which can not be used + * for extent allocation. + * + * This happens when mounting in degraded mode. For example we have a + * RAID1 filesystem with 2 devices, lose one device and mount the fs + * using the other device in degraded mode. If we then allocate a chunk, + * we may have enough free space in the existing system space_info, but + * none of the block groups can be used for extent allocation since they + * have a RAID1 profile, and because we are in degraded mode with a + * single device, we are forced to allocate a new system chunk with a + * SINGLE profile. Making check_system_chunk() iterate over all system + * block groups and check if they have a usable profile and enough space + * can be slow on very large filesystems, so we tolerate the -ENOSPC and + * try again after forcing allocation of a new system chunk. Like this + * we avoid paying the cost of that search in normal circumstances, when + * we were not mounted in degraded mode; + * + * 2) We had enough free space info the system space_info, and one suitable + * block group to allocate from when we called check_system_chunk() + * above. However right after we called it, the only system block group + * with enough free space got turned into RO mode by a running scrub, + * and in this case we have to allocate a new one and retry. We only + * need do this allocate and retry once, since we have a transaction + * handle and scrub uses the commit root to search for block groups; + * + * 3) We had one system block group with enough free space when we called + * check_system_chunk(), but after that, right before we tried to + * allocate the last extent buffer we needed, a discard operation came + * in and it temporarily removed the last free space entry from the + * block group (discard removes a free space entry, discards it, and + * then adds back the entry to the block group cache). + */ + if (ret == -ENOSPC) { + const u64 sys_flags = btrfs_system_alloc_profile(trans->fs_info); + struct btrfs_block_group *sys_bg; + + sys_bg = btrfs_create_chunk(trans, sys_flags); + if (IS_ERR(sys_bg)) { + ret = PTR_ERR(sys_bg); + btrfs_abort_transaction(trans, ret); + goto out; + } + + ret = btrfs_chunk_alloc_add_chunk_item(trans, sys_bg); + if (ret) { + btrfs_abort_transaction(trans, ret); + goto out; + } + + ret = btrfs_chunk_alloc_add_chunk_item(trans, bg); + if (ret) { + btrfs_abort_transaction(trans, ret); + goto out; + } + } else if (ret) { + btrfs_abort_transaction(trans, ret); + goto out; + } +out: + btrfs_trans_release_chunk_metadata(trans); + + if (ret) + return ERR_PTR(ret); + + btrfs_get_block_group(bg); + return bg; +} + /* - * If force is CHUNK_ALLOC_FORCE: + * Chunk allocation is done in 2 phases: + * + * 1) Phase 1 - through btrfs_chunk_alloc() we allocate device extents for + * the chunk, the chunk mapping, create its block group and add the items + * that belong in the chunk btree to it - more specifically, we need to + * update device items in the chunk btree and add a new chunk item to it. + * + * 2) Phase 2 - through btrfs_create_pending_block_groups(), we add the block + * group item to the extent btree and the device extent items to the devices + * btree. + * + * This is done to prevent deadlocks. For example when COWing a node from the + * extent btree we are holding a write lock on the node's parent and if we + * trigger chunk allocation and attempted to insert the new block group item + * in the extent btree right way, we could deadlock because the path for the + * insertion can include that parent node. At first glance it seems impossible + * to trigger chunk allocation after starting a transaction since tasks should + * reserve enough transaction units (metadata space), however while that is true + * most of the time, chunk allocation may still be triggered for several reasons: + * + * 1) When reserving metadata, we check if there is enough free space in the + * metadata space_info and therefore don't trigger allocation of a new chunk. + * However later when the task actually tries to COW an extent buffer from + * the extent btree or from the device btree for example, it is forced to + * allocate a new block group (chunk) because the only one that had enough + * free space was just turned to RO mode by a running scrub for example (or + * device replace, block group reclaim thread, etc), so we can not use it + * for allocating an extent and end up being forced to allocate a new one; + * + * 2) Because we only check that the metadata space_info has enough free bytes, + * we end up not allocating a new metadata chunk in that case. However if + * the filesystem was mounted in degraded mode, none of the existing block + * groups might be suitable for extent allocation due to their incompatible + * profile (for e.g. mounting a 2 devices filesystem, where all block groups + * use a RAID1 profile, in degraded mode using a single device). In this case + * when the task attempts to COW some extent buffer of the extent btree for + * example, it will trigger allocation of a new metadata block group with a + * suitable profile (SINGLE profile in the example of the degraded mount of + * the RAID1 filesystem); + * + * 3) The task has reserved enough transaction units / metadata space, but when + * it attempts to COW an extent buffer from the extent or device btree for + * example, it does not find any free extent in any metadata block group, + * therefore forced to try to allocate a new metadata block group. + * This is because some other task allocated all available extents in the + * meanwhile - this typically happens with tasks that don't reserve space + * properly, either intentionally or as a bug. One example where this is + * done intentionally is fsync, as it does not reserve any transaction units + * and ends up allocating a variable number of metadata extents for log + * tree extent buffers; + * + * 4) The task has reserved enough transaction units / metadata space, but right + * before it tries to allocate the last extent buffer it needs, a discard + * operation comes in and, temporarily, removes the last free space entry from + * the only metadata block group that had free space (discard starts by + * removing a free space entry from a block group, then does the discard + * operation and, once it's done, it adds back the free space entry to the + * block group). + * + * We also need this 2 phases setup when adding a device to a filesystem with + * a seed device - we must create new metadata and system chunks without adding + * any of the block group items to the chunk, extent and device btrees. If we + * did not do it this way, we would get ENOSPC when attempting to update those + * btrees, since all the chunks from the seed device are read-only. + * + * Phase 1 does the updates and insertions to the chunk btree because if we had + * it done in phase 2 and have a thundering herd of tasks allocating chunks in + * parallel, we risk having too many system chunks allocated by many tasks if + * many tasks reach phase 1 without the previous ones completing phase 2. In the + * extreme case this leads to exhaustion of the system chunk array in the + * superblock. This is easier to trigger if using a btree node/leaf size of 64K + * and with RAID filesystems (so we have more device items in the chunk btree). + * This has happened before and commit eafa4fd0ad0607 ("btrfs: fix exhaustion of + * the system chunk array due to concurrent allocations") provides more details. + * + * Allocation of system chunks does not happen through this function. A task that + * needs to update the chunk btree (the only btree that uses system chunks), must + * preallocate chunk space by calling either check_system_chunk() or + * btrfs_reserve_chunk_metadata() - the former is used when allocating a data or + * metadata chunk or when removing a chunk, while the later is used before doing + * a modification to the chunk btree - use cases for the later are adding, + * removing and resizing a device as well as relocation of a system chunk. + * See the comment below for more details. + * + * The reservation of system space, done through check_system_chunk(), as well + * as all the updates and insertions into the chunk btree must be done while + * holding fs_info->chunk_mutex. This is important to guarantee that while COWing + * an extent buffer from the chunks btree we never trigger allocation of a new + * system chunk, which would result in a deadlock (trying to lock twice an + * extent buffer of the chunk btree, first time before triggering the chunk + * allocation and the second time during chunk allocation while attempting to + * update the chunks btree). The system chunk array is also updated while holding + * that mutex. The same logic applies to removing chunks - we must reserve system + * space, update the chunk btree and the system chunk array in the superblock + * while holding fs_info->chunk_mutex. + * + * This function, btrfs_chunk_alloc(), belongs to phase 1. + * + * If @force is CHUNK_ALLOC_FORCE: * - return 1 if it successfully allocates a chunk, * - return errors including -ENOSPC otherwise. - * If force is NOT CHUNK_ALLOC_FORCE: + * If @force is NOT CHUNK_ALLOC_FORCE: * - return 0 if it doesn't need to allocate a new chunk, * - return 1 if it successfully allocates a chunk, * - return errors including -ENOSPC otherwise. @@ -3043,13 +3673,43 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, { struct btrfs_fs_info *fs_info = trans->fs_info; struct btrfs_space_info *space_info; + struct btrfs_block_group *ret_bg; bool wait_for_alloc = false; bool should_alloc = false; + bool from_extent_allocation = false; int ret = 0; + if (force == CHUNK_ALLOC_FORCE_FOR_EXTENT) { + from_extent_allocation = true; + force = CHUNK_ALLOC_FORCE; + } + /* Don't re-enter if we're already allocating a chunk */ if (trans->allocating_chunk) return -ENOSPC; + /* + * Allocation of system chunks can not happen through this path, as we + * could end up in a deadlock if we are allocating a data or metadata + * chunk and there is another task modifying the chunk btree. + * + * This is because while we are holding the chunk mutex, we will attempt + * to add the new chunk item to the chunk btree or update an existing + * device item in the chunk btree, while the other task that is modifying + * the chunk btree is attempting to COW an extent buffer while holding a + * lock on it and on its parent - if the COW operation triggers a system + * chunk allocation, then we can deadlock because we are holding the + * chunk mutex and we may need to access that extent buffer or its parent + * in order to add the chunk item or update a device item. + * + * Tasks that want to modify the chunk tree should reserve system space + * before updating the chunk btree, by calling either + * btrfs_reserve_chunk_metadata() or check_system_chunk(). + * It's possible that after a task reserves the space, it still ends up + * here - this happens in the cases described above at do_chunk_alloc(). + * The task will have to either retry or fail. + */ + if (flags & BTRFS_BLOCK_GROUP_SYSTEM) + return -ENOSPC; space_info = btrfs_find_space_info(fs_info, flags); ASSERT(space_info); @@ -3078,6 +3738,7 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, * attempt. */ wait_for_alloc = true; + force = CHUNK_ALLOC_NO_FORCE; spin_unlock(&space_info->lock); mutex_lock(&fs_info->chunk_mutex); mutex_unlock(&fs_info->chunk_mutex); @@ -3113,15 +3774,22 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, force_metadata_allocation(fs_info); } - /* - * Check if we have enough space in SYSTEM chunk because we may need - * to update devices. - */ - check_system_chunk(trans, flags); - - ret = btrfs_alloc_chunk(trans, flags); + ret_bg = do_chunk_alloc(trans, flags); trans->allocating_chunk = false; + if (IS_ERR(ret_bg)) { + ret = PTR_ERR(ret_bg); + } else if (from_extent_allocation) { + /* + * New block group is likely to be used soon. Try to activate + * it now. Failure is OK for now. + */ + btrfs_zone_activate(ret_bg); + } + + if (!ret) + btrfs_put_block_group(ret_bg); + spin_lock(&space_info->lock); if (ret < 0) { if (ret == -ENOSPC) @@ -3138,22 +3806,6 @@ out: space_info->chunk_alloc = 0; spin_unlock(&space_info->lock); mutex_unlock(&fs_info->chunk_mutex); - /* - * When we allocate a new chunk we reserve space in the chunk block - * reserve to make sure we can COW nodes/leafs in the chunk tree or - * add new nodes/leafs to it if we end up needing to do it when - * inserting the chunk item and updating device items as part of the - * second phase of chunk allocation, performed by - * btrfs_finish_chunk_alloc(). So make sure we don't accumulate a - * large number of new block groups to create in our transaction - * handle's new_bgs list to avoid exhausting the chunk block reserve - * in extreme cases - like having a single transaction create many new - * block groups when starting to write out the free space caches of all - * the block groups that were made dirty during the lifetime of the - * transaction. - */ - if (trans->chunk_bytes_reserved >= (u64)SZ_2M) - btrfs_create_pending_block_groups(trans); return ret; } @@ -3169,17 +3821,14 @@ static u64 get_profile_num_devs(struct btrfs_fs_info *fs_info, u64 type) return num_dev; } -/* - * Reserve space in the system space for allocating or removing a chunk - */ -void check_system_chunk(struct btrfs_trans_handle *trans, u64 type) +static void reserve_chunk_space(struct btrfs_trans_handle *trans, + u64 bytes, + u64 type) { struct btrfs_fs_info *fs_info = trans->fs_info; struct btrfs_space_info *info; u64 left; - u64 thresh; int ret = 0; - u64 num_devs; /* * Needed because we can end up allocating a system chunk and for an @@ -3192,20 +3841,15 @@ void check_system_chunk(struct btrfs_trans_handle *trans, u64 type) left = info->total_bytes - btrfs_space_info_used(info, true); spin_unlock(&info->lock); - num_devs = get_profile_num_devs(fs_info, type); - - /* num_devs device items to update and 1 chunk item to add or remove */ - thresh = btrfs_calc_metadata_size(fs_info, num_devs) + - btrfs_calc_insert_metadata_size(fs_info, 1); - - if (left < thresh && btrfs_test_opt(fs_info, ENOSPC_DEBUG)) { + if (left < bytes && btrfs_test_opt(fs_info, ENOSPC_DEBUG)) { btrfs_info(fs_info, "left=%llu, need=%llu, flags=%llu", - left, thresh, type); + left, bytes, type); btrfs_dump_space_info(fs_info, info, 0, 0); } - if (left < thresh) { + if (left < bytes) { u64 flags = btrfs_system_alloc_profile(fs_info); + struct btrfs_block_group *bg; /* * Ignore failure to create system chunk. We might end up not @@ -3213,50 +3857,111 @@ void check_system_chunk(struct btrfs_trans_handle *trans, u64 type) * the paths we visit in the chunk tree (they were already COWed * or created in the current transaction for example). */ - ret = btrfs_alloc_chunk(trans, flags); + bg = btrfs_create_chunk(trans, flags); + if (IS_ERR(bg)) { + ret = PTR_ERR(bg); + } else { + /* + * We have a new chunk. We also need to activate it for + * zoned filesystem. + */ + ret = btrfs_zoned_activate_one_bg(fs_info, info, true); + if (ret < 0) + return; + + /* + * If we fail to add the chunk item here, we end up + * trying again at phase 2 of chunk allocation, at + * btrfs_create_pending_block_groups(). So ignore + * any error here. An ENOSPC here could happen, due to + * the cases described at do_chunk_alloc() - the system + * block group we just created was just turned into RO + * mode by a scrub for example, or a running discard + * temporarily removed its free space entries, etc. + */ + btrfs_chunk_alloc_add_chunk_item(trans, bg); + } } if (!ret) { - ret = btrfs_block_rsv_add(fs_info->chunk_root, + ret = btrfs_block_rsv_add(fs_info, &fs_info->chunk_block_rsv, - thresh, BTRFS_RESERVE_NO_FLUSH); + bytes, BTRFS_RESERVE_NO_FLUSH); if (!ret) - trans->chunk_bytes_reserved += thresh; + trans->chunk_bytes_reserved += bytes; } } +/* + * Reserve space in the system space for allocating or removing a chunk. + * The caller must be holding fs_info->chunk_mutex. + */ +void check_system_chunk(struct btrfs_trans_handle *trans, u64 type) +{ + struct btrfs_fs_info *fs_info = trans->fs_info; + const u64 num_devs = get_profile_num_devs(fs_info, type); + u64 bytes; + + /* num_devs device items to update and 1 chunk item to add or remove. */ + bytes = btrfs_calc_metadata_size(fs_info, num_devs) + + btrfs_calc_insert_metadata_size(fs_info, 1); + + reserve_chunk_space(trans, bytes, type); +} + +/* + * Reserve space in the system space, if needed, for doing a modification to the + * chunk btree. + * + * @trans: A transaction handle. + * @is_item_insertion: Indicate if the modification is for inserting a new item + * in the chunk btree or if it's for the deletion or update + * of an existing item. + * + * This is used in a context where we need to update the chunk btree outside + * block group allocation and removal, to avoid a deadlock with a concurrent + * task that is allocating a metadata or data block group and therefore needs to + * update the chunk btree while holding the chunk mutex. After the update to the + * chunk btree is done, btrfs_trans_release_chunk_metadata() should be called. + * + */ +void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans, + bool is_item_insertion) +{ + struct btrfs_fs_info *fs_info = trans->fs_info; + u64 bytes; + + if (is_item_insertion) + bytes = btrfs_calc_insert_metadata_size(fs_info, 1); + else + bytes = btrfs_calc_metadata_size(fs_info, 1); + + mutex_lock(&fs_info->chunk_mutex); + reserve_chunk_space(trans, bytes, BTRFS_BLOCK_GROUP_SYSTEM); + mutex_unlock(&fs_info->chunk_mutex); +} + void btrfs_put_block_group_cache(struct btrfs_fs_info *info) { struct btrfs_block_group *block_group; - u64 last = 0; - while (1) { - struct inode *inode; + block_group = btrfs_lookup_first_block_group(info, 0); + while (block_group) { + btrfs_wait_block_group_cache_done(block_group); + spin_lock(&block_group->lock); + if (test_and_clear_bit(BLOCK_GROUP_FLAG_IREF, + &block_group->runtime_flags)) { + struct inode *inode = block_group->inode; - block_group = btrfs_lookup_first_block_group(info, last); - while (block_group) { - btrfs_wait_block_group_cache_done(block_group); - spin_lock(&block_group->lock); - if (block_group->iref) - break; + block_group->inode = NULL; spin_unlock(&block_group->lock); - block_group = btrfs_next_block_group(block_group); - } - if (!block_group) { - if (last == 0) - break; - last = 0; - continue; - } - inode = block_group->inode; - block_group->iref = 0; - block_group->inode = NULL; - spin_unlock(&block_group->lock); - ASSERT(block_group->io_ctl.inode == NULL); - iput(inode); - last = block_group->start + block_group->length; - btrfs_put_block_group(block_group); + ASSERT(block_group->io_ctl.inode == NULL); + iput(inode); + } else { + spin_unlock(&block_group->lock); + } + block_group = btrfs_next_block_group(block_group); } } @@ -3272,14 +3977,14 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) struct btrfs_caching_control *caching_ctl; struct rb_node *n; - down_write(&info->commit_root_sem); + write_lock(&info->block_group_cache_lock); while (!list_empty(&info->caching_block_groups)) { caching_ctl = list_entry(info->caching_block_groups.next, struct btrfs_caching_control, list); list_del(&caching_ctl->list); btrfs_put_caching_control(caching_ctl); } - up_write(&info->commit_root_sem); + write_unlock(&info->block_group_cache_lock); spin_lock(&info->unused_bgs_lock); while (!list_empty(&info->unused_bgs)) { @@ -3289,16 +3994,34 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) list_del_init(&block_group->bg_list); btrfs_put_block_group(block_group); } + + while (!list_empty(&info->reclaim_bgs)) { + block_group = list_first_entry(&info->reclaim_bgs, + struct btrfs_block_group, + bg_list); + list_del_init(&block_group->bg_list); + btrfs_put_block_group(block_group); + } spin_unlock(&info->unused_bgs_lock); - spin_lock(&info->block_group_cache_lock); - while ((n = rb_last(&info->block_group_cache_tree)) != NULL) { + spin_lock(&info->zone_active_bgs_lock); + while (!list_empty(&info->zone_active_bgs)) { + block_group = list_first_entry(&info->zone_active_bgs, + struct btrfs_block_group, + active_bg_list); + list_del_init(&block_group->active_bg_list); + btrfs_put_block_group(block_group); + } + spin_unlock(&info->zone_active_bgs_lock); + + write_lock(&info->block_group_cache_lock); + while ((n = rb_last(&info->block_group_cache_tree.rb_root)) != NULL) { block_group = rb_entry(n, struct btrfs_block_group, cache_node); - rb_erase(&block_group->cache_node, - &info->block_group_cache_tree); + rb_erase_cached(&block_group->cache_node, + &info->block_group_cache_tree); RB_CLEAR_NODE(&block_group->cache_node); - spin_unlock(&info->block_group_cache_lock); + write_unlock(&info->block_group_cache_lock); down_write(&block_group->space_info->groups_sem); list_del(&block_group->list); @@ -3317,20 +4040,13 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) ASSERT(list_empty(&block_group->dirty_list)); ASSERT(list_empty(&block_group->io_list)); ASSERT(list_empty(&block_group->bg_list)); - ASSERT(atomic_read(&block_group->count) == 1); + ASSERT(refcount_read(&block_group->refs) == 1); + ASSERT(block_group->swap_extents == 0); btrfs_put_block_group(block_group); - spin_lock(&info->block_group_cache_lock); + write_lock(&info->block_group_cache_lock); } - spin_unlock(&info->block_group_cache_lock); - - /* - * Now that all the block groups are freed, go through and free all the - * space_info structs. This is only called during the final stages of - * unmount, and so we know nobody is using them. We call - * synchronize_rcu() once before we start, just to be on the safe side. - */ - synchronize_rcu(); + write_unlock(&info->block_group_cache_lock); btrfs_release_global_block_rsv(info); @@ -3344,11 +4060,87 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) * important and indicates a real bug if this happens. */ if (WARN_ON(space_info->bytes_pinned > 0 || - space_info->bytes_reserved > 0 || space_info->bytes_may_use > 0)) btrfs_dump_space_info(info, space_info, 0, 0); + + /* + * If there was a failure to cleanup a log tree, very likely due + * to an IO failure on a writeback attempt of one or more of its + * extent buffers, we could not do proper (and cheap) unaccounting + * of their reserved space, so don't warn on bytes_reserved > 0 in + * that case. + */ + if (!(space_info->flags & BTRFS_BLOCK_GROUP_METADATA) || + !BTRFS_FS_LOG_CLEANUP_ERROR(info)) { + if (WARN_ON(space_info->bytes_reserved > 0)) + btrfs_dump_space_info(info, space_info, 0, 0); + } + + WARN_ON(space_info->reclaim_size > 0); list_del(&space_info->list); btrfs_sysfs_remove_space_info(space_info); } return 0; } + +void btrfs_freeze_block_group(struct btrfs_block_group *cache) +{ + atomic_inc(&cache->frozen); +} + +void btrfs_unfreeze_block_group(struct btrfs_block_group *block_group) +{ + struct btrfs_fs_info *fs_info = block_group->fs_info; + struct extent_map_tree *em_tree; + struct extent_map *em; + bool cleanup; + + spin_lock(&block_group->lock); + cleanup = (atomic_dec_and_test(&block_group->frozen) && + test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)); + spin_unlock(&block_group->lock); + + if (cleanup) { + em_tree = &fs_info->mapping_tree; + write_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, block_group->start, + 1); + BUG_ON(!em); /* logic error, can't happen */ + remove_extent_mapping(em_tree, em); + write_unlock(&em_tree->lock); + + /* once for us and once for the tree */ + free_extent_map(em); + free_extent_map(em); + + /* + * We may have left one free space entry and other possible + * tasks trimming this block group have left 1 entry each one. + * Free them if any. + */ + btrfs_remove_free_space_cache(block_group); + } +} + +bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg) +{ + bool ret = true; + + spin_lock(&bg->lock); + if (bg->ro) + ret = false; + else + bg->swap_extents++; + spin_unlock(&bg->lock); + + return ret; +} + +void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount) +{ + spin_lock(&bg->lock); + ASSERT(!bg->ro); + ASSERT(bg->swap_extents >= amount); + bg->swap_extents -= amount; + spin_unlock(&bg->lock); +} |