--- /dev/null
+// SPDX-License-Identifier: GPL-2.0
+
+#include "misc.h"
+#include "ctree.h"
+#include "block-group.h"
+#include "space-info.h"
+#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"
+#include "sysfs.h"
+#include "tree-log.h"
+#include "delalloc-space.h"
+
+/*
+ * Return target flags in extended format or 0 if restripe for this chunk_type
+ * is not in progress
+ *
+ * Should be called with balance_lock held
+ */
+static u64 get_restripe_target(struct btrfs_fs_info *fs_info, u64 flags)
+{
+ struct btrfs_balance_control *bctl = fs_info->balance_ctl;
+ u64 target = 0;
+
+ if (!bctl)
+ return 0;
+
+ if (flags & BTRFS_BLOCK_GROUP_DATA &&
+ bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+ target = BTRFS_BLOCK_GROUP_DATA | bctl->data.target;
+ } else if (flags & BTRFS_BLOCK_GROUP_SYSTEM &&
+ bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+ target = BTRFS_BLOCK_GROUP_SYSTEM | bctl->sys.target;
+ } else if (flags & BTRFS_BLOCK_GROUP_METADATA &&
+ bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+ target = BTRFS_BLOCK_GROUP_METADATA | bctl->meta.target;
+ }
+
+ return target;
+}
+
+/*
+ * @flags: available profiles in extended format (see ctree.h)
+ *
+ * Return reduced profile in chunk format. If profile changing is in progress
+ * (either running or paused) picks the target profile (if it's already
+ * available), otherwise falls back to plain reducing.
+ */
+static u64 btrfs_reduce_alloc_profile(struct btrfs_fs_info *fs_info, u64 flags)
+{
+ u64 num_devices = fs_info->fs_devices->rw_devices;
+ u64 target;
+ u64 raid_type;
+ u64 allowed = 0;
+
+ /*
+ * See if restripe for this chunk_type is in progress, if so try to
+ * reduce to the target profile
+ */
+ 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);
+
+ /* First, mask out the RAID levels which aren't possible */
+ for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
+ if (num_devices >= btrfs_raid_array[raid_type].devs_min)
+ allowed |= btrfs_raid_array[raid_type].bg_flag;
+ }
+ allowed &= flags;
+
+ if (allowed & BTRFS_BLOCK_GROUP_RAID6)
+ allowed = BTRFS_BLOCK_GROUP_RAID6;
+ else if (allowed & BTRFS_BLOCK_GROUP_RAID5)
+ allowed = BTRFS_BLOCK_GROUP_RAID5;
+ else if (allowed & BTRFS_BLOCK_GROUP_RAID10)
+ allowed = BTRFS_BLOCK_GROUP_RAID10;
+ else if (allowed & BTRFS_BLOCK_GROUP_RAID1)
+ allowed = BTRFS_BLOCK_GROUP_RAID1;
+ else if (allowed & BTRFS_BLOCK_GROUP_RAID0)
+ allowed = BTRFS_BLOCK_GROUP_RAID0;
+
+ flags &= ~BTRFS_BLOCK_GROUP_PROFILE_MASK;
+
+ return extended_to_chunk(flags | allowed);
+}
+
+static u64 get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags)
+{
+ unsigned seq;
+ u64 flags;
+
+ do {
+ flags = orig_flags;
+ seq = read_seqbegin(&fs_info->profiles_lock);
+
+ if (flags & BTRFS_BLOCK_GROUP_DATA)
+ flags |= fs_info->avail_data_alloc_bits;
+ else if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+ flags |= fs_info->avail_system_alloc_bits;
+ else if (flags & BTRFS_BLOCK_GROUP_METADATA)
+ flags |= fs_info->avail_metadata_alloc_bits;
+ } while (read_seqretry(&fs_info->profiles_lock, seq));
+
+ return btrfs_reduce_alloc_profile(fs_info, flags);
+}
+
+u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags)
+{
+ return get_alloc_profile(fs_info, orig_flags);
+}
+
+void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
+{
+ atomic_inc(&cache->count);
+}
+
+void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
+{
+ if (atomic_dec_and_test(&cache->count)) {
+ WARN_ON(cache->pinned > 0);
+ WARN_ON(cache->reserved > 0);
+
+ /*
+ * If not empty, someone is still holding mutex of
+ * full_stripe_lock, which can only be released by caller.
+ * And it will definitely cause use-after-free when caller
+ * tries to release full stripe lock.
+ *
+ * No better way to resolve, but only to warn.
+ */
+ WARN_ON(!RB_EMPTY_ROOT(&cache->full_stripe_locks_root.root));
+ kfree(cache->free_space_ctl);
+ kfree(cache);
+ }
+}
+
+/*
+ * This adds the block group to the fs_info rb tree for the block group cache
+ */
+static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
+ struct btrfs_block_group_cache *block_group)
+{
+ struct rb_node **p;
+ struct rb_node *parent = NULL;
+ struct btrfs_block_group_cache *cache;
+
+ spin_lock(&info->block_group_cache_lock);
+ p = &info->block_group_cache_tree.rb_node;
+
+ while (*p) {
+ parent = *p;
+ cache = rb_entry(parent, struct btrfs_block_group_cache,
+ cache_node);
+ if (block_group->key.objectid < cache->key.objectid) {
+ p = &(*p)->rb_left;
+ } else if (block_group->key.objectid > cache->key.objectid) {
+ p = &(*p)->rb_right;
+ } else {
+ spin_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->key.objectid)
+ info->first_logical_byte = block_group->key.objectid;
+
+ spin_unlock(&info->block_group_cache_lock);
+
+ return 0;
+}
+
+/*
+ * This will return the block group at or after bytenr if contains is 0, else
+ * it will return the block group that contains the bytenr
+ */
+static struct btrfs_block_group_cache *block_group_cache_tree_search(
+ struct btrfs_fs_info *info, u64 bytenr, int contains)
+{
+ struct btrfs_block_group_cache *cache, *ret = NULL;
+ struct rb_node *n;
+ u64 end, start;
+
+ spin_lock(&info->block_group_cache_lock);
+ n = info->block_group_cache_tree.rb_node;
+
+ while (n) {
+ cache = rb_entry(n, struct btrfs_block_group_cache,
+ cache_node);
+ end = cache->key.objectid + cache->key.offset - 1;
+ start = cache->key.objectid;
+
+ if (bytenr < start) {
+ if (!contains && (!ret || start < ret->key.objectid))
+ ret = cache;
+ n = n->rb_left;
+ } else if (bytenr > start) {
+ if (contains && bytenr <= end) {
+ ret = cache;
+ break;
+ }
+ n = n->rb_right;
+ } else {
+ ret = cache;
+ break;
+ }
+ }
+ if (ret) {
+ btrfs_get_block_group(ret);
+ if (bytenr == 0 && info->first_logical_byte > ret->key.objectid)
+ info->first_logical_byte = ret->key.objectid;
+ }
+ spin_unlock(&info->block_group_cache_lock);
+
+ return ret;
+}
+
+/*
+ * Return the block group that starts at or after bytenr
+ */
+struct btrfs_block_group_cache *btrfs_lookup_first_block_group(
+ struct btrfs_fs_info *info, u64 bytenr)
+{
+ return block_group_cache_tree_search(info, bytenr, 0);
+}
+
+/*
+ * Return the block group that contains the given bytenr
+ */
+struct btrfs_block_group_cache *btrfs_lookup_block_group(
+ struct btrfs_fs_info *info, u64 bytenr)
+{
+ return block_group_cache_tree_search(info, bytenr, 1);
+}
+
+struct btrfs_block_group_cache *btrfs_next_block_group(
+ struct btrfs_block_group_cache *cache)
+{
+ struct btrfs_fs_info *fs_info = cache->fs_info;
+ struct rb_node *node;
+
+ spin_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->key.objectid + cache->key.offset;
+
+ spin_unlock(&fs_info->block_group_cache_lock);
+ btrfs_put_block_group(cache);
+ cache = btrfs_lookup_first_block_group(fs_info, next_bytenr); return cache;
+ }
+ node = rb_next(&cache->cache_node);
+ btrfs_put_block_group(cache);
+ if (node) {
+ cache = rb_entry(node, struct btrfs_block_group_cache,
+ cache_node);
+ btrfs_get_block_group(cache);
+ } else
+ cache = NULL;
+ spin_unlock(&fs_info->block_group_cache_lock);
+ return cache;
+}
+
+bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr)
+{
+ struct btrfs_block_group_cache *bg;
+ bool ret = true;
+
+ bg = btrfs_lookup_block_group(fs_info, bytenr);
+ if (!bg)
+ return false;
+
+ spin_lock(&bg->lock);
+ if (bg->ro)
+ ret = false;
+ else
+ atomic_inc(&bg->nocow_writers);
+ spin_unlock(&bg->lock);
+
+ /* No put on block group, done by btrfs_dec_nocow_writers */
+ if (!ret)
+ btrfs_put_block_group(bg);
+
+ return ret;
+}
+
+void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr)
+{
+ struct btrfs_block_group_cache *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);
+ btrfs_put_block_group(bg);
+}
+
+void btrfs_wait_nocow_writers(struct btrfs_block_group_cache *bg)
+{
+ wait_var_event(&bg->nocow_writers, !atomic_read(&bg->nocow_writers));
+}
+
+void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
+ const u64 start)
+{
+ struct btrfs_block_group_cache *bg;
+
+ bg = btrfs_lookup_block_group(fs_info, start);
+ ASSERT(bg);
+ if (atomic_dec_and_test(&bg->reservations))
+ wake_up_var(&bg->reservations);
+ btrfs_put_block_group(bg);
+}
+
+void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
+{
+ struct btrfs_space_info *space_info = bg->space_info;
+
+ ASSERT(bg->ro);
+
+ if (!(bg->flags & BTRFS_BLOCK_GROUP_DATA))
+ return;
+
+ /*
+ * Our block group is read only but before we set it to read only,
+ * some task might have had allocated an extent from it already, but it
+ * has not yet created a respective ordered extent (and added it to a
+ * root's list of ordered extents).
+ * Therefore wait for any task currently allocating extents, since the
+ * block group's reservations counter is incremented while a read lock
+ * on the groups' semaphore is held and decremented after releasing
+ * the read access on that semaphore and creating the ordered extent.
+ */
+ down_write(&space_info->groups_sem);
+ up_write(&space_info->groups_sem);
+
+ wait_var_event(&bg->reservations, !atomic_read(&bg->reservations));
+}
+
+struct btrfs_caching_control *btrfs_get_caching_control(
+ struct btrfs_block_group_cache *cache)
+{
+ struct btrfs_caching_control *ctl;
+
+ spin_lock(&cache->lock);
+ if (!cache->caching_ctl) {
+ spin_unlock(&cache->lock);
+ return NULL;
+ }
+
+ ctl = cache->caching_ctl;
+ refcount_inc(&ctl->count);
+ spin_unlock(&cache->lock);
+ return ctl;
+}
+
+void btrfs_put_caching_control(struct btrfs_caching_control *ctl)
+{
+ if (refcount_dec_and_test(&ctl->count))
+ kfree(ctl);
+}
+
+/*
+ * When we wait for progress in the block group caching, its because our
+ * allocation attempt failed at least once. So, we must sleep and let some
+ * progress happen before we try again.
+ *
+ * This function will sleep at least once waiting for new free space to show
+ * up, and then it will check the block group free space numbers for our min
+ * num_bytes. Another option is to have it go ahead and look in the rbtree for
+ * a free extent of a given size, but this is a good start.
+ *
+ * Callers of this must check if cache->cached == BTRFS_CACHE_ERROR before using
+ * any of the information in this block group.
+ */
+void btrfs_wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
+ u64 num_bytes)
+{
+ struct btrfs_caching_control *caching_ctl;
+
+ caching_ctl = btrfs_get_caching_control(cache);
+ if (!caching_ctl)
+ return;
+
+ wait_event(caching_ctl->wait, btrfs_block_group_cache_done(cache) ||
+ (cache->free_space_ctl->free_space >= num_bytes));
+
+ btrfs_put_caching_control(caching_ctl);
+}
+
+int btrfs_wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
+{
+ struct btrfs_caching_control *caching_ctl;
+ int ret = 0;
+
+ 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_cache_done(cache));
+ if (cache->cached == BTRFS_CACHE_ERROR)
+ ret = -EIO;
+ btrfs_put_caching_control(caching_ctl);
+ return ret;
+}
+
+#ifdef CONFIG_BTRFS_DEBUG
+static void fragment_free_space(struct btrfs_block_group_cache *block_group)
+{
+ struct btrfs_fs_info *fs_info = block_group->fs_info;
+ u64 start = block_group->key.objectid;
+ u64 len = block_group->key.offset;
+ u64 chunk = block_group->flags & BTRFS_BLOCK_GROUP_METADATA ?
+ fs_info->nodesize : fs_info->sectorsize;
+ u64 step = chunk << 1;
+
+ while (len > chunk) {
+ btrfs_remove_free_space(block_group, start, chunk);
+ start += step;
+ if (len < step)
+ len = 0;
+ else
+ len -= step;
+ }
+}
+#endif
+
+/*
+ * This is only called by btrfs_cache_block_group, since we could have freed
+ * extents we need to check the pinned_extents for any extents that can't be
+ * used yet since their free space will be released as soon as the transaction
+ * commits.
+ */
+u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
+ u64 start, u64 end)
+{
+ struct btrfs_fs_info *info = block_group->fs_info;
+ u64 extent_start, extent_end, size, total_added = 0;
+ int ret;
+
+ while (start < end) {
+ ret = find_first_extent_bit(info->pinned_extents, start,
+ &extent_start, &extent_end,
+ EXTENT_DIRTY | EXTENT_UPTODATE,
+ NULL);
+ if (ret)
+ break;
+
+ if (extent_start <= start) {
+ start = extent_end + 1;
+ } else if (extent_start > start && extent_start < end) {
+ size = extent_start - start;
+ total_added += size;
+ ret = btrfs_add_free_space(block_group, start,
+ size);
+ BUG_ON(ret); /* -ENOMEM or logic error */
+ start = extent_end + 1;
+ } else {
+ break;
+ }
+ }
+
+ if (start < end) {
+ size = end - start;
+ total_added += size;
+ ret = btrfs_add_free_space(block_group, start, size);
+ BUG_ON(ret); /* -ENOMEM or logic error */
+ }
+
+ return total_added;
+}
+
+static int load_extent_tree_free(struct btrfs_caching_control *caching_ctl)
+{
+ struct btrfs_block_group_cache *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_path *path;
+ struct extent_buffer *leaf;
+ struct btrfs_key key;
+ u64 total_found = 0;
+ u64 last = 0;
+ u32 nritems;
+ int ret;
+ bool wakeup = true;
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
+
+#ifdef CONFIG_BTRFS_DEBUG
+ /*
+ * If we're fragmenting we don't want to make anybody think we can
+ * allocate from this block group until we've had a chance to fragment
+ * the free space.
+ */
+ if (btrfs_should_fragment_free_space(block_group))
+ wakeup = false;
+#endif
+ /*
+ * We don't want to deadlock with somebody trying to allocate a new
+ * extent for the extent root while also trying to search the extent
+ * root to add free space. So we skip locking and search the commit
+ * root, since its read-only
+ */
+ path->skip_locking = 1;
+ path->search_commit_root = 1;
+ path->reada = READA_FORWARD;
+
+ key.objectid = last;
+ key.offset = 0;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+
+next:
+ ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
+ if (ret < 0)
+ goto out;
+
+ leaf = path->nodes[0];
+ nritems = btrfs_header_nritems(leaf);
+
+ while (1) {
+ if (btrfs_fs_closing(fs_info) > 1) {
+ last = (u64)-1;
+ break;
+ }
+
+ if (path->slots[0] < nritems) {
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+ } else {
+ ret = btrfs_find_next_key(extent_root, path, &key, 0, 0);
+ if (ret)
+ break;
+
+ 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);
+ cond_resched();
+ mutex_lock(&caching_ctl->mutex);
+ down_read(&fs_info->commit_root_sem);
+ goto next;
+ }
+
+ ret = btrfs_next_leaf(extent_root, path);
+ if (ret < 0)
+ goto out;
+ if (ret)
+ break;
+ leaf = path->nodes[0];
+ nritems = btrfs_header_nritems(leaf);
+ continue;
+ }
+
+ if (key.objectid < last) {
+ key.objectid = last;
+ key.offset = 0;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+
+ if (wakeup)
+ caching_ctl->progress = last;
+ btrfs_release_path(path);
+ goto next;
+ }
+
+ if (key.objectid < block_group->key.objectid) {
+ path->slots[0]++;
+ continue;
+ }
+
+ if (key.objectid >= block_group->key.objectid +
+ block_group->key.offset)
+ break;
+
+ if (key.type == BTRFS_EXTENT_ITEM_KEY ||
+ key.type == BTRFS_METADATA_ITEM_KEY) {
+ total_found += add_new_free_space(block_group, last,
+ key.objectid);
+ if (key.type == BTRFS_METADATA_ITEM_KEY)
+ last = key.objectid +
+ fs_info->nodesize;
+ else
+ last = key.objectid + key.offset;
+
+ if (total_found > CACHING_CTL_WAKE_UP) {
+ total_found = 0;
+ if (wakeup)
+ wake_up(&caching_ctl->wait);
+ }
+ }
+ path->slots[0]++;
+ }
+ ret = 0;
+
+ total_found += add_new_free_space(block_group, last,
+ block_group->key.objectid +
+ block_group->key.offset);
+ caching_ctl->progress = (u64)-1;
+
+out:
+ btrfs_free_path(path);
+ return ret;
+}
+
+static noinline void caching_thread(struct btrfs_work *work)
+{
+ struct btrfs_block_group_cache *block_group;
+ struct btrfs_fs_info *fs_info;
+ struct btrfs_caching_control *caching_ctl;
+ int ret;
+
+ caching_ctl = container_of(work, struct btrfs_caching_control, work);
+ block_group = caching_ctl->block_group;
+ fs_info = block_group->fs_info;
+
+ mutex_lock(&caching_ctl->mutex);
+ down_read(&fs_info->commit_root_sem);
+
+ if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
+ ret = load_free_space_tree(caching_ctl);
+ else
+ ret = load_extent_tree_free(caching_ctl);
+
+ spin_lock(&block_group->lock);
+ block_group->caching_ctl = NULL;
+ block_group->cached = ret ? BTRFS_CACHE_ERROR : BTRFS_CACHE_FINISHED;
+ spin_unlock(&block_group->lock);
+
+#ifdef CONFIG_BTRFS_DEBUG
+ if (btrfs_should_fragment_free_space(block_group)) {
+ u64 bytes_used;
+
+ spin_lock(&block_group->space_info->lock);
+ spin_lock(&block_group->lock);
+ bytes_used = block_group->key.offset -
+ btrfs_block_group_used(&block_group->item);
+ block_group->space_info->bytes_used += bytes_used >> 1;
+ spin_unlock(&block_group->lock);
+ spin_unlock(&block_group->space_info->lock);
+ fragment_free_space(block_group);
+ }
+#endif
+
+ caching_ctl->progress = (u64)-1;
+
+ up_read(&fs_info->commit_root_sem);
+ btrfs_free_excluded_extents(block_group);
+ mutex_unlock(&caching_ctl->mutex);
+
+ wake_up(&caching_ctl->wait);
+
+ btrfs_put_caching_control(caching_ctl);
+ btrfs_put_block_group(block_group);
+}
+
+int btrfs_cache_block_group(struct btrfs_block_group_cache *cache,
+ int load_cache_only)
+{
+ DEFINE_WAIT(wait);
+ struct btrfs_fs_info *fs_info = cache->fs_info;
+ struct btrfs_caching_control *caching_ctl;
+ int ret = 0;
+
+ caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
+ if (!caching_ctl)
+ return -ENOMEM;
+
+ INIT_LIST_HEAD(&caching_ctl->list);
+ mutex_init(&caching_ctl->mutex);
+ init_waitqueue_head(&caching_ctl->wait);
+ caching_ctl->block_group = cache;
+ caching_ctl->progress = cache->key.objectid;
+ refcount_set(&caching_ctl->count, 1);
+ btrfs_init_work(&caching_ctl->work, btrfs_cache_helper,
+ 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;
+ }
+ WARN_ON(cache->caching_ctl);
+ cache->caching_ctl = caching_ctl;
+ cache->cached = BTRFS_CACHE_FAST;
+ 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->key.offset -
+ btrfs_block_group_used(&cache->item);
+ 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);
+ refcount_inc(&caching_ctl->count);
+ list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
+ up_write(&fs_info->commit_root_sem);
+
+ btrfs_get_block_group(cache);
+
+ btrfs_queue_work(fs_info->caching_workers, &caching_ctl->work);
+
+ return ret;
+}
+
+static void clear_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
+{
+ u64 extra_flags = chunk_to_extended(flags) &
+ BTRFS_EXTENDED_PROFILE_MASK;
+
+ write_seqlock(&fs_info->profiles_lock);
+ if (flags & BTRFS_BLOCK_GROUP_DATA)
+ fs_info->avail_data_alloc_bits &= ~extra_flags;
+ if (flags & BTRFS_BLOCK_GROUP_METADATA)
+ fs_info->avail_metadata_alloc_bits &= ~extra_flags;
+ if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+ fs_info->avail_system_alloc_bits &= ~extra_flags;
+ write_sequnlock(&fs_info->profiles_lock);
+}
+
+/*
+ * Clear incompat bits for the following feature(s):
+ *
+ * - RAID56 - in case there's neither RAID5 nor RAID6 profile block group
+ * in the whole filesystem
+ */
+static void clear_incompat_bg_bits(struct btrfs_fs_info *fs_info, u64 flags)
+{
+ if (flags & BTRFS_BLOCK_GROUP_RAID56_MASK) {
+ struct list_head *head = &fs_info->space_info;
+ struct btrfs_space_info *sinfo;
+
+ list_for_each_entry_rcu(sinfo, head, list) {
+ bool found = false;
+
+ down_read(&sinfo->groups_sem);
+ if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID5]))
+ found = true;
+ if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID6]))
+ found = true;
+ up_read(&sinfo->groups_sem);
+
+ if (found)
+ return;
+ }
+ btrfs_clear_fs_incompat(fs_info, RAID56);
+ }
+}
+
+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_cache *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;
+ int index;
+ int factor;
+ struct btrfs_caching_control *caching_ctl = NULL;
+ bool remove_em;
+ bool remove_rsv = false;
+
+ block_group = btrfs_lookup_block_group(fs_info, group_start);
+ BUG_ON(!block_group);
+ BUG_ON(!block_group->ro);
+
+ trace_btrfs_remove_block_group(block_group);
+ /*
+ * Free the reserved super bytes from this block group before
+ * remove it.
+ */
+ btrfs_free_excluded_extents(block_group);
+ btrfs_free_ref_tree_range(fs_info, block_group->key.objectid,
+ block_group->key.offset);
+
+ memcpy(&key, &block_group->key, sizeof(key));
+ index = btrfs_bg_flags_to_raid_index(block_group->flags);
+ factor = btrfs_bg_type_to_factor(block_group->flags);
+
+ /* make sure this block group isn't part of an allocation cluster */
+ cluster = &fs_info->data_alloc_cluster;
+ spin_lock(&cluster->refill_lock);
+ btrfs_return_cluster_to_free_space(block_group, cluster);
+ spin_unlock(&cluster->refill_lock);
+
+ /*
+ * make sure this block group isn't part of a metadata
+ * allocation cluster
+ */
+ cluster = &fs_info->meta_alloc_cluster;
+ spin_lock(&cluster->refill_lock);
+ btrfs_return_cluster_to_free_space(block_group, cluster);
+ spin_unlock(&cluster->refill_lock);
+
+ path = btrfs_alloc_path();
+ if (!path) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ /*
+ * get the inode first so any iput calls done for the io_list
+ * aren't the final iput (no unlinks allowed now)
+ */
+ inode = lookup_free_space_inode(block_group, path);
+
+ mutex_lock(&trans->transaction->cache_write_mutex);
+ /*
+ * Make sure our free space cache IO is done before removing the
+ * free space inode
+ */
+ spin_lock(&trans->transaction->dirty_bgs_lock);
+ if (!list_empty(&block_group->io_list)) {
+ list_del_init(&block_group->io_list);
+
+ WARN_ON(!IS_ERR(inode) && inode != block_group->io_ctl.inode);
+
+ spin_unlock(&trans->transaction->dirty_bgs_lock);
+ btrfs_wait_cache_io(trans, block_group, path);
+ btrfs_put_block_group(block_group);
+ spin_lock(&trans->transaction->dirty_bgs_lock);
+ }
+
+ if (!list_empty(&block_group->dirty_list)) {
+ list_del_init(&block_group->dirty_list);
+ remove_rsv = true;
+ btrfs_put_block_group(block_group);
+ }
+ 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.offset = block_group->key.objectid;
+ key.type = 0;
+
+ ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
+ if (ret < 0)
+ 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);
+ RB_CLEAR_NODE(&block_group->cache_node);
+
+ if (fs_info->first_logical_byte == block_group->key.objectid)
+ fs_info->first_logical_byte = (u64)-1;
+ spin_unlock(&fs_info->block_group_cache_lock);
+
+ down_write(&block_group->space_info->groups_sem);
+ /*
+ * we must use list_del_init so people can check to see if they
+ * are still on the list after taking the semaphore
+ */
+ list_del_init(&block_group->list);
+ if (list_empty(&block_group->space_info->block_groups[index])) {
+ kobj = block_group->space_info->block_group_kobjs[index];
+ block_group->space_info->block_group_kobjs[index] = NULL;
+ clear_avail_alloc_bits(fs_info, block_group->flags);
+ }
+ up_write(&block_group->space_info->groups_sem);
+ clear_incompat_bg_bits(fs_info, block_group->flags);
+ if (kobj) {
+ kobject_del(kobj);
+ 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);
+ }
+ }
+
+ spin_lock(&trans->transaction->dirty_bgs_lock);
+ WARN_ON(!list_empty(&block_group->dirty_list));
+ WARN_ON(!list_empty(&block_group->io_list));
+ spin_unlock(&trans->transaction->dirty_bgs_lock);
+
+ btrfs_remove_free_space_cache(block_group);
+
+ spin_lock(&block_group->space_info->lock);
+ list_del_init(&block_group->ro_list);
+
+ if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
+ WARN_ON(block_group->space_info->total_bytes
+ < block_group->key.offset);
+ WARN_ON(block_group->space_info->bytes_readonly
+ < block_group->key.offset);
+ WARN_ON(block_group->space_info->disk_total
+ < block_group->key.offset * factor);
+ }
+ block_group->space_info->total_bytes -= block_group->key.offset;
+ block_group->space_info->bytes_readonly -= block_group->key.offset;
+ block_group->space_info->disk_total -= block_group->key.offset * factor;
+
+ spin_unlock(&block_group->space_info->lock);
+
+ memcpy(&key, &block_group->key, sizeof(key));
+
+ mutex_lock(&fs_info->chunk_mutex);
+ spin_lock(&block_group->lock);
+ block_group->removed = 1;
+ /*
+ * 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().
+ *
+ * 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
+ * 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
+ * physical device locations unless we take this special care.
+ *
+ * There may also be an implicit trim operation if the file system
+ * is mounted with -odiscard. The same protections must remain
+ * in place until the extents have been discarded completely when
+ * the transaction commit has completed.
+ */
+ remove_em = (atomic_read(&block_group->trimming) == 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;
+
+ em_tree = &fs_info->mapping_tree;
+ write_lock(&em_tree->lock);
+ remove_extent_mapping(em_tree, em);
+ write_unlock(&em_tree->lock);
+ /* once for the tree */
+ free_extent_map(em);
+ }
+out:
+ if (remove_rsv)
+ btrfs_delayed_refs_rsv_release(fs_info, 1);
+ btrfs_free_path(path);
+ return ret;
+}
+
+struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
+ struct btrfs_fs_info *fs_info, const u64 chunk_offset)
+{
+ struct extent_map_tree *em_tree = &fs_info->mapping_tree;
+ struct extent_map *em;
+ struct map_lookup *map;
+ unsigned int num_items;
+
+ read_lock(&em_tree->lock);
+ em = lookup_extent_mapping(em_tree, chunk_offset, 1);
+ read_unlock(&em_tree->lock);
+ ASSERT(em && em->start == chunk_offset);
+
+ /*
+ * We need to reserve 3 + N units from the metadata space info in order
+ * to remove a block group (done at btrfs_remove_chunk() and at
+ * btrfs_remove_block_group()), which are used for:
+ *
+ * 1 unit for adding the free space inode's orphan (located in the tree
+ * of tree roots).
+ * 1 unit for deleting the block group item (located in the extent
+ * tree).
+ * 1 unit for deleting the free space item (located in tree of tree
+ * roots).
+ * N units for deleting N device extent items corresponding to each
+ * stripe (located in the device tree).
+ *
+ * In order to remove a block group we also need to reserve units in the
+ * system space info in order to update the chunk tree (update one or
+ * more device items and remove one chunk item), but this is done at
+ * btrfs_remove_chunk() through a call to check_system_chunk().
+ */
+ map = em->map_lookup;
+ num_items = 3 + map->num_stripes;
+ free_extent_map(em);
+
+ return btrfs_start_transaction_fallback_global_rsv(fs_info->extent_root,
+ num_items, 1);
+}
+
+/*
+ * Mark block group @cache read-only, so later write won't happen to block
+ * group @cache.
+ *
+ * If @force is not set, this function will only mark the block group readonly
+ * if we have enough free space (1M) in other metadata/system block groups.
+ * If @force is not set, this function will mark the block group readonly
+ * without checking free space.
+ *
+ * NOTE: This function doesn't care if other block groups can contain all the
+ * data in this block group. That check should be done by relocation routine,
+ * not this function.
+ */
+static int inc_block_group_ro(struct btrfs_block_group_cache *cache, int force)
+{
+ struct btrfs_space_info *sinfo = cache->space_info;
+ u64 num_bytes;
+ u64 sinfo_used;
+ u64 min_allocable_bytes;
+ int ret = -ENOSPC;
+
+ /*
+ * We need some metadata space and system metadata space for
+ * allocating chunks in some corner cases until we force to set
+ * it to be readonly.
+ */
+ if ((sinfo->flags &
+ (BTRFS_BLOCK_GROUP_SYSTEM | BTRFS_BLOCK_GROUP_METADATA)) &&
+ !force)
+ min_allocable_bytes = SZ_1M;
+ else
+ min_allocable_bytes = 0;
+
+ spin_lock(&sinfo->lock);
+ spin_lock(&cache->lock);
+
+ if (cache->ro) {
+ cache->ro++;
+ ret = 0;
+ goto out;
+ }
+
+ num_bytes = cache->key.offset - cache->reserved - cache->pinned -
+ cache->bytes_super - btrfs_block_group_used(&cache->item);
+ sinfo_used = btrfs_space_info_used(sinfo, true);
+
+ /*
+ * sinfo_used + num_bytes should always <= sinfo->total_bytes.
+ *
+ * Here we make sure if we mark this bg RO, we still have enough
+ * free space as buffer (if min_allocable_bytes is not 0).
+ */
+ if (sinfo_used + num_bytes + min_allocable_bytes <=
+ sinfo->total_bytes) {
+ sinfo->bytes_readonly += num_bytes;
+ cache->ro++;
+ list_add_tail(&cache->ro_list, &sinfo->ro_bgs);
+ ret = 0;
+ }
+out:
+ spin_unlock(&cache->lock);
+ spin_unlock(&sinfo->lock);
+ if (ret == -ENOSPC && btrfs_test_opt(cache->fs_info, ENOSPC_DEBUG)) {
+ btrfs_info(cache->fs_info,
+ "unable to make block group %llu ro",
+ cache->key.objectid);
+ btrfs_info(cache->fs_info,
+ "sinfo_used=%llu bg_num_bytes=%llu min_allocable=%llu",
+ sinfo_used, num_bytes, min_allocable_bytes);
+ btrfs_dump_space_info(cache->fs_info, cache->space_info, 0, 0);
+ }
+ return ret;
+}
+
+/*
+ * Process the unused_bgs list and remove any that don't have any allocated
+ * space inside of them.
+ */
+void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info)
+{
+ struct btrfs_block_group_cache *block_group;
+ struct btrfs_space_info *space_info;
+ struct btrfs_trans_handle *trans;
+ int ret = 0;
+
+ if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
+ 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,
+ struct btrfs_block_group_cache,
+ bg_list);
+ list_del_init(&block_group->bg_list);
+
+ space_info = block_group->space_info;
+
+ if (ret || btrfs_mixed_space_info(space_info)) {
+ btrfs_put_block_group(block_group);
+ continue;
+ }
+ spin_unlock(&fs_info->unused_bgs_lock);
+
+ 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);
+ spin_lock(&block_group->lock);
+ if (block_group->reserved || block_group->pinned ||
+ btrfs_block_group_used(&block_group->item) ||
+ block_group->ro ||
+ list_is_singular(&block_group->list)) {
+ /*
+ * 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.
+ */
+ trace_btrfs_skip_unused_block_group(block_group);
+ spin_unlock(&block_group->lock);
+ up_write(&space_info->groups_sem);
+ goto next;
+ }
+ spin_unlock(&block_group->lock);
+
+ /* We don't want to force the issue, only flip if it's ok. */
+ ret = inc_block_group_ro(block_group, 0);
+ up_write(&space_info->groups_sem);
+ if (ret < 0) {
+ 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.
+ */
+ trans = btrfs_start_trans_remove_block_group(fs_info,
+ block_group->key.objectid);
+ if (IS_ERR(trans)) {
+ btrfs_dec_block_group_ro(block_group);
+ ret = PTR_ERR(trans);
+ goto next;
+ }
+
+ /*
+ * We could have pending pinned extents for this block group,
+ * just delete them, we don't care about them anymore.
+ */
+ start = block_group->key.objectid;
+ end = start + block_group->key.offset - 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);
+ 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);
+
+ /* Reset pinned so btrfs_put_block_group doesn't complain */
+ spin_lock(&space_info->lock);
+ spin_lock(&block_group->lock);
+
+ 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);
+ spin_unlock(&space_info->lock);
+
+ /* DISCARD can flip during remount */
+ trimming = btrfs_test_opt(fs_info, DISCARD);
+
+ /* Implicit trim during transaction commit. */
+ if (trimming)
+ btrfs_get_block_group_trimming(block_group);
+
+ /*
+ * Btrfs_remove_chunk will abort the transaction if things go
+ * horribly wrong.
+ */
+ ret = btrfs_remove_chunk(trans, block_group->key.objectid);
+
+ if (ret) {
+ if (trimming)
+ btrfs_put_block_group_trimming(block_group);
+ goto end_trans;
+ }
+
+ /*
+ * If we're not mounted with -odiscard, we can just forget
+ * about this block group. Otherwise we'll need to wait
+ * until transaction commit to do the actual discard.
+ */
+ if (trimming) {
+ spin_lock(&fs_info->unused_bgs_lock);
+ /*
+ * A concurrent scrub might have added us to the list
+ * fs_info->unused_bgs, so use a list_move operation
+ * to add the block group to the deleted_bgs list.
+ */
+ list_move(&block_group->bg_list,
+ &trans->transaction->deleted_bgs);
+ spin_unlock(&fs_info->unused_bgs_lock);
+ btrfs_get_block_group(block_group);
+ }
+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);
+}
+
+void btrfs_mark_bg_unused(struct btrfs_block_group_cache *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_unused_block_group(bg);
+ list_add_tail(&bg->bg_list, &fs_info->unused_bgs);
+ }
+ 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)
+{
+ struct btrfs_root *root = fs_info->extent_root;
+ int ret = 0;
+ struct btrfs_key found_key;
+ struct extent_buffer *leaf;
+ struct btrfs_block_group_item bg;
+ u64 flags;
+ int slot;
+
+ ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
+ if (ret < 0)
+ goto out;
+
+ 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);
+
+ 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_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;
+ }
+ path->slots[0]++;
+ }
+out:
+ return ret;
+}
+
+static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
+{
+ u64 extra_flags = chunk_to_extended(flags) &
+ BTRFS_EXTENDED_PROFILE_MASK;
+
+ write_seqlock(&fs_info->profiles_lock);
+ if (flags & BTRFS_BLOCK_GROUP_DATA)
+ fs_info->avail_data_alloc_bits |= extra_flags;
+ if (flags & BTRFS_BLOCK_GROUP_METADATA)
+ fs_info->avail_metadata_alloc_bits |= extra_flags;
+ if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+ fs_info->avail_system_alloc_bits |= extra_flags;
+ write_sequnlock(&fs_info->profiles_lock);
+}
+
+static int exclude_super_stripes(struct btrfs_block_group_cache *cache)
+{
+ struct btrfs_fs_info *fs_info = cache->fs_info;
+ u64 bytenr;
+ u64 *logical;
+ int stripe_len;
+ int i, nr, ret;
+
+ if (cache->key.objectid < BTRFS_SUPER_INFO_OFFSET) {
+ stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->key.objectid;
+ cache->bytes_super += stripe_len;
+ ret = btrfs_add_excluded_extent(fs_info, cache->key.objectid,
+ stripe_len);
+ if (ret)
+ return ret;
+ }
+
+ for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+ bytenr = btrfs_sb_offset(i);
+ ret = btrfs_rmap_block(fs_info, cache->key.objectid,
+ bytenr, &logical, &nr, &stripe_len);
+ if (ret)
+ return ret;
+
+ while (nr--) {
+ u64 start, len;
+
+ if (logical[nr] > cache->key.objectid +
+ cache->key.offset)
+ continue;
+
+ if (logical[nr] + stripe_len <= cache->key.objectid)
+ continue;
+
+ start = logical[nr];
+ if (start < cache->key.objectid) {
+ start = cache->key.objectid;
+ len = (logical[nr] + stripe_len) - start;
+ } else {
+ len = min_t(u64, stripe_len,
+ cache->key.objectid +
+ cache->key.offset - start);
+ }
+
+ cache->bytes_super += len;
+ ret = btrfs_add_excluded_extent(fs_info, start, len);
+ if (ret) {
+ kfree(logical);
+ return ret;
+ }
+ }
+
+ kfree(logical);
+ }
+ return 0;
+}
+
+static void link_block_group(struct btrfs_block_group_cache *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_cache *btrfs_create_block_group_cache(
+ struct btrfs_fs_info *fs_info, u64 start, u64 size)
+{
+ struct btrfs_block_group_cache *cache;
+
+ cache = kzalloc(sizeof(*cache), GFP_NOFS);
+ if (!cache)
+ return NULL;
+
+ cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
+ GFP_NOFS);
+ if (!cache->free_space_ctl) {
+ kfree(cache);
+ return NULL;
+ }
+
+ cache->key.objectid = start;
+ cache->key.offset = size;
+ cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
+
+ cache->fs_info = fs_info;
+ cache->full_stripe_len = btrfs_full_stripe_len(fs_info, start);
+ set_free_space_tree_thresholds(cache);
+
+ atomic_set(&cache->count, 1);
+ spin_lock_init(&cache->lock);
+ init_rwsem(&cache->data_rwsem);
+ INIT_LIST_HEAD(&cache->list);
+ INIT_LIST_HEAD(&cache->cluster_list);
+ INIT_LIST_HEAD(&cache->bg_list);
+ INIT_LIST_HEAD(&cache->ro_list);
+ INIT_LIST_HEAD(&cache->dirty_list);
+ INIT_LIST_HEAD(&cache->io_list);
+ btrfs_init_free_space_ctl(cache);
+ atomic_set(&cache->trimming, 0);
+ mutex_init(&cache->free_space_lock);
+ btrfs_init_full_stripe_locks_tree(&cache->full_stripe_locks_root);
+
+ return cache;
+}
+
+/*
+ * Iterate all chunks and verify that each of them has the corresponding block
+ * group
+ */
+static int check_chunk_block_group_mappings(struct btrfs_fs_info *fs_info)
+{
+ struct extent_map_tree *map_tree = &fs_info->mapping_tree;
+ struct extent_map *em;
+ struct btrfs_block_group_cache *bg;
+ u64 start = 0;
+ int ret = 0;
+
+ while (1) {
+ read_lock(&map_tree->lock);
+ /*
+ * lookup_extent_mapping will return the first extent map
+ * intersecting the range, so setting @len to 1 is enough to
+ * get the first chunk.
+ */
+ em = lookup_extent_mapping(map_tree, start, 1);
+ read_unlock(&map_tree->lock);
+ if (!em)
+ break;
+
+ bg = btrfs_lookup_block_group(fs_info, em->start);
+ if (!bg) {
+ btrfs_err(fs_info,
+ "chunk start=%llu len=%llu doesn't have corresponding block group",
+ em->start, em->len);
+ ret = -EUCLEAN;
+ free_extent_map(em);
+ break;
+ }
+ if (bg->key.objectid != em->start ||
+ bg->key.offset != em->len ||
+ (bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK) !=
+ (em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK)) {
+ btrfs_err(fs_info,
+"chunk start=%llu len=%llu flags=0x%llx doesn't match block group start=%llu len=%llu flags=0x%llx",
+ em->start, em->len,
+ em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK,
+ bg->key.objectid, bg->key.offset,
+ bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK);
+ ret = -EUCLEAN;
+ free_extent_map(em);
+ btrfs_put_block_group(bg);
+ break;
+ }
+ start = em->start + em->len;
+ free_extent_map(em);
+ btrfs_put_block_group(bg);
+ }
+ return ret;
+}
+
+int btrfs_read_block_groups(struct btrfs_fs_info *info)
+{
+ struct btrfs_path *path;
+ int ret;
+ struct btrfs_block_group_cache *cache;
+ struct btrfs_space_info *space_info;
+ struct btrfs_key key;
+ struct btrfs_key found_key;
+ struct extent_buffer *leaf;
+ int need_clear = 0;
+ u64 cache_gen;
+ u64 feature;
+ int mixed;
+
+ feature = btrfs_super_incompat_flags(info->super_copy);
+ mixed = !!(feature & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS);
+
+ 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) &&
+ btrfs_super_generation(info->super_copy) != cache_gen)
+ need_clear = 1;
+ if (btrfs_test_opt(info, CLEAR_CACHE))
+ need_clear = 1;
+
+ while (1) {
+ ret = find_first_block_group(info, path, &key);
+ if (ret > 0)
+ break;
+ if (ret != 0)
+ goto error;
+
+ leaf = path->nodes[0];
+ btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+
+ cache = btrfs_create_block_group_cache(info, found_key.objectid,
+ found_key.offset);
+ if (!cache) {
+ ret = -ENOMEM;
+ goto error;
+ }
+
+ if (need_clear) {
+ /*
+ * When we mount with old space cache, we need to
+ * set BTRFS_DC_CLEAR and set dirty flag.
+ *
+ * a) Setting 'BTRFS_DC_CLEAR' makes sure that we
+ * truncate the old free space cache inode and
+ * setup a new one.
+ * b) Setting 'dirty flag' makes sure that we flush
+ * the new space cache info onto disk.
+ */
+ if (btrfs_test_opt(info, SPACE_CACHE))
+ cache->disk_cache_state = BTRFS_DC_CLEAR;
+ }
+
+ read_extent_buffer(leaf, &cache->item,
+ btrfs_item_ptr_offset(leaf, path->slots[0]),
+ sizeof(cache->item));
+ cache->flags = btrfs_block_group_flags(&cache->item);
+ if (!mixed &&
+ ((cache->flags & BTRFS_BLOCK_GROUP_METADATA) &&
+ (cache->flags & BTRFS_BLOCK_GROUP_DATA))) {
+ btrfs_err(info,
+"bg %llu is a mixed block group but filesystem hasn't enabled mixed block groups",
+ cache->key.objectid);
+ ret = -EINVAL;
+ goto error;
+ }
+
+ key.objectid = found_key.objectid + found_key.offset;
+ btrfs_release_path(path);
+
+ /*
+ * 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 than we actually do.
+ */
+ 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);
+ goto error;
+ }
+
+ /*
+ * 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 (found_key.offset == btrfs_block_group_used(&cache->item)) {
+ cache->last_byte_to_unpin = (u64)-1;
+ cache->cached = BTRFS_CACHE_FINISHED;
+ btrfs_free_excluded_extents(cache);
+ } else if (btrfs_block_group_used(&cache->item) == 0) {
+ cache->last_byte_to_unpin = (u64)-1;
+ cache->cached = BTRFS_CACHE_FINISHED;
+ add_new_free_space(cache, found_key.objectid,
+ found_key.objectid +
+ found_key.offset);
+ btrfs_free_excluded_extents(cache);
+ }
+
+ ret = btrfs_add_block_group_cache(info, cache);
+ if (ret) {
+ btrfs_remove_free_space_cache(cache);
+ btrfs_put_block_group(cache);
+ goto error;
+ }
+
+ trace_btrfs_add_block_group(info, cache, 0);
+ btrfs_update_space_info(info, cache->flags, found_key.offset,
+ btrfs_block_group_used(&cache->item),
+ cache->bytes_super, &space_info);
+
+ cache->space_info = space_info;
+
+ link_block_group(cache);
+
+ set_avail_alloc_bits(info, cache->flags);
+ if (btrfs_chunk_readonly(info, cache->key.objectid)) {
+ inc_block_group_ro(cache, 1);
+ } else if (btrfs_block_group_used(&cache->item) == 0) {
+ ASSERT(list_empty(&cache->bg_list));
+ btrfs_mark_bg_unused(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 |
+ BTRFS_BLOCK_GROUP_RAID56_MASK |
+ BTRFS_BLOCK_GROUP_DUP)))
+ continue;
+ /*
+ * Avoid allocating from un-mirrored block group if there are
+ * mirrored block groups.
+ */
+ list_for_each_entry(cache,
+ &space_info->block_groups[BTRFS_RAID_RAID0],
+ list)
+ inc_block_group_ro(cache, 1);
+ list_for_each_entry(cache,
+ &space_info->block_groups[BTRFS_RAID_SINGLE],
+ list)
+ inc_block_group_ro(cache, 1);
+ }
+
+ btrfs_init_global_block_rsv(info);
+ ret = check_chunk_block_group_mappings(info);
+error:
+ btrfs_free_path(path);
+ return ret;
+}
+
+void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_block_group_cache *block_group;
+ struct btrfs_root *extent_root = fs_info->extent_root;
+ struct btrfs_block_group_item item;
+ struct btrfs_key key;
+ int ret = 0;
+
+ if (!trans->can_flush_pending_bgs)
+ return;
+
+ while (!list_empty(&trans->new_bgs)) {
+ block_group = list_first_entry(&trans->new_bgs,
+ struct btrfs_block_group_cache,
+ bg_list);
+ if (ret)
+ goto next;
+
+ spin_lock(&block_group->lock);
+ memcpy(&item, &block_group->item, sizeof(item));
+ memcpy(&key, &block_group->key, sizeof(key));
+ spin_unlock(&block_group->lock);
+
+ ret = btrfs_insert_item(trans, extent_root, &key, &item,
+ sizeof(item));
+ if (ret)
+ btrfs_abort_transaction(trans, ret);
+ ret = btrfs_finish_chunk_alloc(trans, key.objectid, key.offset);
+ if (ret)
+ btrfs_abort_transaction(trans, ret);
+ add_block_group_free_space(trans, block_group);
+ /* Already aborted the transaction if it failed. */
+next:
+ btrfs_delayed_refs_rsv_release(fs_info, 1);
+ list_del_init(&block_group->bg_list);
+ }
+ 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)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_block_group_cache *cache;
+ int ret;
+
+ btrfs_set_log_full_commit(trans);
+
+ cache = btrfs_create_block_group_cache(fs_info, chunk_offset, size);
+ if (!cache)
+ return -ENOMEM;
+
+ btrfs_set_block_group_used(&cache->item, bytes_used);
+ btrfs_set_block_group_chunk_objectid(&cache->item,
+ BTRFS_FIRST_CHUNK_TREE_OBJECTID);
+ btrfs_set_block_group_flags(&cache->item, type);
+
+ cache->flags = type;
+ cache->last_byte_to_unpin = (u64)-1;
+ cache->cached = BTRFS_CACHE_FINISHED;
+ cache->needs_free_space = 1;
+ 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;
+ }
+
+ 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
+ * with its ->space_info set.
+ */
+ cache->space_info = btrfs_find_space_info(fs_info, cache->flags);
+ ASSERT(cache->space_info);
+
+ ret = btrfs_add_block_group_cache(fs_info, cache);
+ if (ret) {
+ btrfs_remove_free_space_cache(cache);
+ btrfs_put_block_group(cache);
+ return ret;
+ }
+
+ /*
+ * Now that our block group has its ->space_info set and is inserted in
+ * 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_update_global_block_rsv(fs_info);
+
+ link_block_group(cache);
+
+ 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;
+}
+
+int btrfs_inc_block_group_ro(struct btrfs_block_group_cache *cache)
+
+{
+ struct btrfs_fs_info *fs_info = cache->fs_info;
+ struct btrfs_trans_handle *trans;
+ u64 alloc_flags;
+ int ret;
+
+again:
+ trans = btrfs_join_transaction(fs_info->extent_root);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
+
+ /*
+ * 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
+ */
+ 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;
+ goto again;
+ }
+
+ /*
+ * 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);
+ if (alloc_flags != cache->flags) {
+ ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
+ /*
+ * ENOSPC is allowed here, we may have enough space
+ * already allocated at the new raid level to
+ * carry on
+ */
+ if (ret == -ENOSPC)
+ ret = 0;
+ if (ret < 0)
+ goto out;
+ }
+
+ ret = inc_block_group_ro(cache, 0);
+ if (!ret)
+ goto out;
+ alloc_flags = btrfs_get_alloc_profile(fs_info, cache->space_info->flags);
+ ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
+ if (ret < 0)
+ goto out;
+ ret = inc_block_group_ro(cache, 0);
+out:
+ if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {
+ alloc_flags = update_block_group_flags(fs_info, cache->flags);
+ mutex_lock(&fs_info->chunk_mutex);
+ check_system_chunk(trans, alloc_flags);
+ mutex_unlock(&fs_info->chunk_mutex);
+ }
+ mutex_unlock(&fs_info->ro_block_group_mutex);
+
+ btrfs_end_transaction(trans);
+ return ret;
+}
+
+void btrfs_dec_block_group_ro(struct btrfs_block_group_cache *cache)
+{
+ struct btrfs_space_info *sinfo = cache->space_info;
+ u64 num_bytes;
+
+ BUG_ON(!cache->ro);
+
+ spin_lock(&sinfo->lock);
+ spin_lock(&cache->lock);
+ if (!--cache->ro) {
+ num_bytes = cache->key.offset - cache->reserved -
+ cache->pinned - cache->bytes_super -
+ btrfs_block_group_used(&cache->item);
+ sinfo->bytes_readonly -= num_bytes;
+ list_del_init(&cache->ro_list);
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&sinfo->lock);
+}
+
+static int write_one_cache_group(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ struct btrfs_block_group_cache *cache)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ int ret;
+ struct btrfs_root *extent_root = fs_info->extent_root;
+ unsigned long bi;
+ struct extent_buffer *leaf;
+
+ ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
+ if (ret) {
+ if (ret > 0)
+ ret = -ENOENT;
+ goto fail;
+ }
+
+ leaf = path->nodes[0];
+ bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
+ write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
+ btrfs_mark_buffer_dirty(leaf);
+fail:
+ btrfs_release_path(path);
+ return ret;
+
+}
+
+static int cache_save_setup(struct btrfs_block_group_cache *block_group,
+ struct btrfs_trans_handle *trans,
+ struct btrfs_path *path)
+{
+ struct btrfs_fs_info *fs_info = block_group->fs_info;
+ struct btrfs_root *root = fs_info->tree_root;
+ struct inode *inode = NULL;
+ struct extent_changeset *data_reserved = NULL;
+ u64 alloc_hint = 0;
+ int dcs = BTRFS_DC_ERROR;
+ u64 num_pages = 0;
+ int retries = 0;
+ int ret = 0;
+
+ /*
+ * If this block group is smaller than 100 megs don't bother caching the
+ * block group.
+ */
+ if (block_group->key.offset < (100 * SZ_1M)) {
+ spin_lock(&block_group->lock);
+ block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+ spin_unlock(&block_group->lock);
+ return 0;
+ }
+
+ if (trans->aborted)
+ return 0;
+again:
+ inode = lookup_free_space_inode(block_group, path);
+ if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {
+ ret = PTR_ERR(inode);
+ btrfs_release_path(path);
+ goto out;
+ }
+
+ if (IS_ERR(inode)) {
+ BUG_ON(retries);
+ retries++;
+
+ if (block_group->ro)
+ goto out_free;
+
+ ret = create_free_space_inode(trans, block_group, path);
+ if (ret)
+ goto out_free;
+ goto again;
+ }
+
+ /*
+ * We want to set the generation to 0, that way if anything goes wrong
+ * from here on out we know not to trust this cache when we load up next
+ * time.
+ */
+ BTRFS_I(inode)->generation = 0;
+ ret = btrfs_update_inode(trans, root, inode);
+ if (ret) {
+ /*
+ * So theoretically we could recover from this, simply set the
+ * super cache generation to 0 so we know to invalidate the
+ * cache, but then we'd have to keep track of the block groups
+ * that fail this way so we know we _have_ to reset this cache
+ * before the next commit or risk reading stale cache. So to
+ * limit our exposure to horrible edge cases lets just abort the
+ * transaction, this only happens in really bad situations
+ * anyway.
+ */
+ btrfs_abort_transaction(trans, ret);
+ goto out_put;
+ }
+ WARN_ON(ret);
+
+ /* We've already setup this transaction, go ahead and exit */
+ if (block_group->cache_generation == trans->transid &&
+ i_size_read(inode)) {
+ dcs = BTRFS_DC_SETUP;
+ goto out_put;
+ }
+
+ if (i_size_read(inode) > 0) {
+ ret = btrfs_check_trunc_cache_free_space(fs_info,
+ &fs_info->global_block_rsv);
+ if (ret)
+ goto out_put;
+
+ ret = btrfs_truncate_free_space_cache(trans, NULL, inode);
+ if (ret)
+ goto out_put;
+ }
+
+ spin_lock(&block_group->lock);
+ if (block_group->cached != BTRFS_CACHE_FINISHED ||
+ !btrfs_test_opt(fs_info, SPACE_CACHE)) {
+ /*
+ * don't bother trying to write stuff out _if_
+ * a) we're not cached,
+ * b) we're with nospace_cache mount option,
+ * c) we're with v2 space_cache (FREE_SPACE_TREE).
+ */
+ dcs = BTRFS_DC_WRITTEN;
+ spin_unlock(&block_group->lock);
+ goto out_put;
+ }
+ spin_unlock(&block_group->lock);
+
+ /*
+ * We hit an ENOSPC when setting up the cache in this transaction, just
+ * skip doing the setup, we've already cleared the cache so we're safe.
+ */
+ if (test_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags)) {
+ ret = -ENOSPC;
+ goto out_put;
+ }
+
+ /*
+ * Try to preallocate enough space based on how big the block group is.
+ * Keep in mind this has to include any pinned space which could end up
+ * taking up quite a bit since it's not folded into the other space
+ * cache.
+ */
+ num_pages = div_u64(block_group->key.offset, SZ_256M);
+ if (!num_pages)
+ num_pages = 1;
+
+ num_pages *= 16;
+ num_pages *= PAGE_SIZE;
+
+ ret = btrfs_check_data_free_space(inode, &data_reserved, 0, num_pages);
+ if (ret)
+ goto out_put;
+
+ ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
+ num_pages, num_pages,
+ &alloc_hint);
+ /*
+ * Our cache requires contiguous chunks so that we don't modify a bunch
+ * of metadata or split extents when writing the cache out, which means
+ * we can enospc if we are heavily fragmented in addition to just normal
+ * out of space conditions. So if we hit this just skip setting up any
+ * other block groups for this transaction, maybe we'll unpin enough
+ * space the next time around.
+ */
+ if (!ret)
+ dcs = BTRFS_DC_SETUP;
+ else if (ret == -ENOSPC)
+ set_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags);
+
+out_put:
+ iput(inode);
+out_free:
+ btrfs_release_path(path);
+out:
+ spin_lock(&block_group->lock);
+ if (!ret && dcs == BTRFS_DC_SETUP)
+ block_group->cache_generation = trans->transid;
+ block_group->disk_cache_state = dcs;
+ spin_unlock(&block_group->lock);
+
+ extent_changeset_free(data_reserved);
+ return ret;
+}
+
+int btrfs_setup_space_cache(struct btrfs_trans_handle *trans)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_block_group_cache *cache, *tmp;
+ struct btrfs_transaction *cur_trans = trans->transaction;
+ struct btrfs_path *path;
+
+ if (list_empty(&cur_trans->dirty_bgs) ||
+ !btrfs_test_opt(fs_info, SPACE_CACHE))
+ return 0;
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ /* Could add new block groups, use _safe just in case */
+ list_for_each_entry_safe(cache, tmp, &cur_trans->dirty_bgs,
+ dirty_list) {
+ if (cache->disk_cache_state == BTRFS_DC_CLEAR)
+ cache_save_setup(cache, trans, path);
+ }
+
+ btrfs_free_path(path);
+ return 0;
+}
+
+/*
+ * Transaction commit does final block group cache writeback during a critical
+ * section where nothing is allowed to change the FS. This is required in
+ * order for the cache to actually match the block group, but can introduce a
+ * lot of latency into the commit.
+ *
+ * So, btrfs_start_dirty_block_groups is here to kick off block group cache IO.
+ * There's a chance we'll have to redo some of it if the block group changes
+ * again during the commit, but it greatly reduces the commit latency by
+ * getting rid of the easy block groups while we're still allowing others to
+ * join the commit.
+ */
+int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_block_group_cache *cache;
+ struct btrfs_transaction *cur_trans = trans->transaction;
+ int ret = 0;
+ int should_put;
+ 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);
+ if (list_empty(&cur_trans->dirty_bgs)) {
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+ return 0;
+ }
+ list_splice_init(&cur_trans->dirty_bgs, &dirty);
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+
+again:
+ /* Make sure all the block groups on our dirty list actually exist */
+ btrfs_create_pending_block_groups(trans);
+
+ if (!path) {
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+ }
+
+ /*
+ * cache_write_mutex is here only to save us from balance or automatic
+ * removal of empty block groups deleting this block group while we are
+ * writing out the cache
+ */
+ mutex_lock(&trans->transaction->cache_write_mutex);
+ while (!list_empty(&dirty)) {
+ bool drop_reserve = true;
+
+ cache = list_first_entry(&dirty,
+ struct btrfs_block_group_cache,
+ dirty_list);
+ /*
+ * This can happen if something re-dirties a block group that
+ * is already under IO. Just wait for it to finish and then do
+ * it all again
+ */
+ if (!list_empty(&cache->io_list)) {
+ list_del_init(&cache->io_list);
+ btrfs_wait_cache_io(trans, cache, path);
+ btrfs_put_block_group(cache);
+ }
+
+
+ /*
+ * btrfs_wait_cache_io uses the cache->dirty_list to decide if
+ * it should update the cache_state. Don't delete until after
+ * we wait.
+ *
+ * Since we're not running in the commit critical section
+ * we need the dirty_bgs_lock to protect from update_block_group
+ */
+ spin_lock(&cur_trans->dirty_bgs_lock);
+ list_del_init(&cache->dirty_list);
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+
+ should_put = 1;
+
+ cache_save_setup(cache, trans, path);
+
+ if (cache->disk_cache_state == BTRFS_DC_SETUP) {
+ 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;
+
+ /*
+ * The cache_write_mutex is protecting the
+ * io_list, also refer to the definition of
+ * btrfs_transaction::io_bgs for more details
+ */
+ list_add_tail(&cache->io_list, io);
+ } else {
+ /*
+ * If we failed to write the cache, the
+ * generation will be bad and life goes on
+ */
+ ret = 0;
+ }
+ }
+ if (!ret) {
+ ret = write_one_cache_group(trans, path, cache);
+ /*
+ * Our block group might still be attached to the list
+ * of new block groups in the transaction handle of some
+ * other task (struct btrfs_trans_handle->new_bgs). This
+ * means its block group item isn't yet in the extent
+ * tree. If this happens ignore the error, as we will
+ * try again later in the critical section of the
+ * transaction commit.
+ */
+ if (ret == -ENOENT) {
+ ret = 0;
+ spin_lock(&cur_trans->dirty_bgs_lock);
+ if (list_empty(&cache->dirty_list)) {
+ list_add_tail(&cache->dirty_list,
+ &cur_trans->dirty_bgs);
+ btrfs_get_block_group(cache);
+ drop_reserve = false;
+ }
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+ } else if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ }
+ }
+
+ /* If it's not on the io list, we need to put the block group */
+ if (should_put)
+ 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);
+ mutex_lock(&trans->transaction->cache_write_mutex);
+ }
+ mutex_unlock(&trans->transaction->cache_write_mutex);
+
+ /*
+ * 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 && loops == 0) {
+ loops++;
+ spin_lock(&cur_trans->dirty_bgs_lock);
+ list_splice_init(&cur_trans->dirty_bgs, &dirty);
+ /*
+ * dirty_bgs_lock protects us from concurrent block group
+ * deletes too (not just cache_write_mutex).
+ */
+ if (!list_empty(&dirty)) {
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+ goto again;
+ }
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+ } else if (ret < 0) {
+ btrfs_cleanup_dirty_bgs(cur_trans, fs_info);
+ }
+
+ btrfs_free_path(path);
+ return ret;
+}
+
+int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_block_group_cache *cache;
+ struct btrfs_transaction *cur_trans = trans->transaction;
+ int ret = 0;
+ 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)
+ return -ENOMEM;
+
+ /*
+ * Even though we are in the critical section of the transaction commit,
+ * we can still have concurrent tasks adding elements to this
+ * transaction's list of dirty block groups. These tasks correspond to
+ * endio free space workers started when writeback finishes for a
+ * space cache, which run inode.c:btrfs_finish_ordered_io(), and can
+ * allocate new block groups as a result of COWing nodes of the root
+ * tree when updating the free space inode. The writeback for the space
+ * caches is triggered by an earlier call to
+ * btrfs_start_dirty_block_groups() and iterations of the following
+ * loop.
+ * Also we want to do the cache_save_setup first and then run the
+ * delayed refs to make sure we have the best chance at doing this all
+ * in one shot.
+ */
+ spin_lock(&cur_trans->dirty_bgs_lock);
+ while (!list_empty(&cur_trans->dirty_bgs)) {
+ cache = list_first_entry(&cur_trans->dirty_bgs,
+ struct btrfs_block_group_cache,
+ dirty_list);
+
+ /*
+ * This can happen if cache_save_setup re-dirties a block group
+ * that is already under IO. Just wait for it to finish and
+ * then do it all again
+ */
+ if (!list_empty(&cache->io_list)) {
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+ list_del_init(&cache->io_list);
+ btrfs_wait_cache_io(trans, cache, path);
+ btrfs_put_block_group(cache);
+ spin_lock(&cur_trans->dirty_bgs_lock);
+ }
+
+ /*
+ * Don't remove from the dirty list until after we've waited on
+ * any pending IO
+ */
+ list_del_init(&cache->dirty_list);
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+ should_put = 1;
+
+ cache_save_setup(cache, trans, path);
+
+ if (!ret)
+ ret = btrfs_run_delayed_refs(trans,
+ (unsigned long) -1);
+
+ if (!ret && cache->disk_cache_state == BTRFS_DC_SETUP) {
+ 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 {
+ /*
+ * If we failed to write the cache, the
+ * generation will be bad and life goes on
+ */
+ ret = 0;
+ }
+ }
+ if (!ret) {
+ ret = write_one_cache_group(trans, path, cache);
+ /*
+ * One of the free space endio workers might have
+ * created a new block group while updating a free space
+ * cache's inode (at inode.c:btrfs_finish_ordered_io())
+ * and hasn't released its transaction handle yet, in
+ * which case the new block group is still attached to
+ * its transaction handle and its creation has not
+ * 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
+ * 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);
+ }
+ if (ret)
+ btrfs_abort_transaction(trans, ret);
+ }
+
+ /* If its not on the io list, we need to put the block group */
+ if (should_put)
+ btrfs_put_block_group(cache);
+ btrfs_delayed_refs_rsv_release(fs_info, 1);
+ spin_lock(&cur_trans->dirty_bgs_lock);
+ }
+ spin_unlock(&cur_trans->dirty_bgs_lock);
+
+ /*
+ * Refer to the definition of io_bgs member for details why it's safe
+ * to use it without any locking
+ */
+ while (!list_empty(io)) {
+ cache = list_first_entry(io, struct btrfs_block_group_cache,
+ io_list);
+ list_del_init(&cache->io_list);
+ btrfs_wait_cache_io(trans, cache, path);
+ btrfs_put_block_group(cache);
+ }
+
+ btrfs_free_path(path);
+ return ret;
+}
+
+int btrfs_update_block_group(struct btrfs_trans_handle *trans,
+ u64 bytenr, u64 num_bytes, int alloc)
+{
+ struct btrfs_fs_info *info = trans->fs_info;
+ struct btrfs_block_group_cache *cache = NULL;
+ u64 total = num_bytes;
+ u64 old_val;
+ u64 byte_in_group;
+ int factor;
+ int ret = 0;
+
+ /* Block accounting for super block */
+ spin_lock(&info->delalloc_root_lock);
+ old_val = btrfs_super_bytes_used(info->super_copy);
+ if (alloc)
+ old_val += num_bytes;
+ else
+ old_val -= num_bytes;
+ btrfs_set_super_bytes_used(info->super_copy, old_val);
+ spin_unlock(&info->delalloc_root_lock);
+
+ while (total) {
+ cache = btrfs_lookup_block_group(info, bytenr);
+ if (!cache) {
+ ret = -ENOENT;
+ break;
+ }
+ factor = btrfs_bg_type_to_factor(cache->flags);
+
+ /*
+ * If this block group has free space cache written out, we
+ * need to make sure to load it if we are removing space. This
+ * is because we need the unpinning stage to actually add the
+ * space back to the block group, otherwise we will leak space.
+ */
+ if (!alloc && cache->cached == BTRFS_CACHE_NO)
+ btrfs_cache_block_group(cache, 1);
+
+ byte_in_group = bytenr - cache->key.objectid;
+ WARN_ON(byte_in_group > cache->key.offset);
+
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+
+ if (btrfs_test_opt(info, SPACE_CACHE) &&
+ cache->disk_cache_state < BTRFS_DC_CLEAR)
+ cache->disk_cache_state = BTRFS_DC_CLEAR;
+
+ old_val = btrfs_block_group_used(&cache->item);
+ num_bytes = min(total, cache->key.offset - byte_in_group);
+ if (alloc) {
+ old_val += num_bytes;
+ btrfs_set_block_group_used(&cache->item, old_val);
+ cache->reserved -= num_bytes;
+ cache->space_info->bytes_reserved -= num_bytes;
+ cache->space_info->bytes_used += num_bytes;
+ cache->space_info->disk_used += num_bytes * factor;
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+ } else {
+ old_val -= num_bytes;
+ btrfs_set_block_group_used(&cache->item, old_val);
+ cache->pinned += num_bytes;
+ btrfs_space_info_update_bytes_pinned(info,
+ cache->space_info, num_bytes);
+ cache->space_info->bytes_used -= num_bytes;
+ cache->space_info->disk_used -= num_bytes * factor;
+ 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,
+ bytenr, bytenr + num_bytes - 1,
+ GFP_NOFS | __GFP_NOFAIL);
+ }
+
+ spin_lock(&trans->transaction->dirty_bgs_lock);
+ if (list_empty(&cache->dirty_list)) {
+ list_add_tail(&cache->dirty_list,
+ &trans->transaction->dirty_bgs);
+ trans->delayed_ref_updates++;
+ btrfs_get_block_group(cache);
+ }
+ spin_unlock(&trans->transaction->dirty_bgs_lock);
+
+ /*
+ * No longer have used bytes in this block group, queue it for
+ * deletion. We do this after adding the block group to the
+ * dirty list to avoid races between cleaner kthread and space
+ * cache writeout.
+ */
+ if (!alloc && old_val == 0)
+ btrfs_mark_bg_unused(cache);
+
+ btrfs_put_block_group(cache);
+ total -= num_bytes;
+ bytenr += num_bytes;
+ }
+
+ /* Modified block groups are accounted for in the delayed_refs_rsv. */
+ btrfs_update_delayed_refs_rsv(trans);
+ return ret;
+}
+
+/**
+ * btrfs_add_reserved_bytes - update the block_group and space info counters
+ * @cache: The cache we are manipulating
+ * @ram_bytes: The number of bytes of file content, and will be same to
+ * @num_bytes except for the compress path.
+ * @num_bytes: The number of bytes in question
+ * @delalloc: The blocks are allocated for the delalloc write
+ *
+ * This is called by the allocator when it reserves space. If this is a
+ * reservation and the block group has become read only we cannot make the
+ * reservation and return -EAGAIN, otherwise this function always succeeds.
+ */
+int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
+ u64 ram_bytes, u64 num_bytes, int delalloc)
+{
+ struct btrfs_space_info *space_info = cache->space_info;
+ int ret = 0;
+
+ spin_lock(&space_info->lock);
+ spin_lock(&cache->lock);
+ if (cache->ro) {
+ ret = -EAGAIN;
+ } else {
+ cache->reserved += num_bytes;
+ space_info->bytes_reserved += num_bytes;
+ trace_btrfs_space_reservation(cache->fs_info, "space_info",
+ space_info->flags, num_bytes, 1);
+ btrfs_space_info_update_bytes_may_use(cache->fs_info,
+ space_info, -ram_bytes);
+ if (delalloc)
+ cache->delalloc_bytes += num_bytes;
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&space_info->lock);
+ return ret;
+}
+
+/**
+ * btrfs_free_reserved_bytes - update the block_group and space info counters
+ * @cache: The cache we are manipulating
+ * @num_bytes: The number of bytes in question
+ * @delalloc: The blocks are allocated for the delalloc write
+ *
+ * This is called by somebody who is freeing space that was never actually used
+ * on disk. For example if you reserve some space for a new leaf in transaction
+ * A and before transaction A commits you free that leaf, you call this with
+ * reserve set to 0 in order to clear the reservation.
+ */
+void btrfs_free_reserved_bytes(struct btrfs_block_group_cache *cache,
+ u64 num_bytes, int delalloc)
+{
+ struct btrfs_space_info *space_info = cache->space_info;
+
+ spin_lock(&space_info->lock);
+ spin_lock(&cache->lock);
+ if (cache->ro)
+ space_info->bytes_readonly += num_bytes;
+ cache->reserved -= num_bytes;
+ space_info->bytes_reserved -= num_bytes;
+ space_info->max_extent_size = 0;
+
+ if (delalloc)
+ cache->delalloc_bytes -= num_bytes;
+ spin_unlock(&cache->lock);
+ spin_unlock(&space_info->lock);
+}
+
+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) {
+ 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,
+ struct btrfs_space_info *sinfo, int force)
+{
+ u64 bytes_used = btrfs_space_info_used(sinfo, false);
+ u64 thresh;
+
+ if (force == CHUNK_ALLOC_FORCE)
+ return 1;
+
+ /*
+ * in limited mode, we want to have some free space up to
+ * about 1% of the FS size.
+ */
+ if (force == CHUNK_ALLOC_LIMITED) {
+ thresh = btrfs_super_total_bytes(fs_info->super_copy);
+ thresh = max_t(u64, SZ_64M, div_factor_fine(thresh, 1));
+
+ if (sinfo->total_bytes - bytes_used < thresh)
+ return 1;
+ }
+
+ if (bytes_used + SZ_2M < div_factor(sinfo->total_bytes, 8))
+ return 0;
+ return 1;
+}
+
+int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type)
+{
+ u64 alloc_flags = btrfs_get_alloc_profile(trans->fs_info, type);
+
+ return btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
+}
+
+/*
+ * 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:
+ * - 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.
+ */
+int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
+ enum btrfs_chunk_alloc_enum force)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_space_info *space_info;
+ bool wait_for_alloc = false;
+ bool should_alloc = false;
+ int ret = 0;
+
+ /* Don't re-enter if we're already allocating a chunk */
+ if (trans->allocating_chunk)
+ return -ENOSPC;
+
+ space_info = btrfs_find_space_info(fs_info, flags);
+ ASSERT(space_info);
+
+ do {
+ spin_lock(&space_info->lock);
+ if (force < space_info->force_alloc)
+ force = space_info->force_alloc;
+ should_alloc = should_alloc_chunk(fs_info, space_info, force);
+ if (space_info->full) {
+ /* No more free physical space */
+ if (should_alloc)
+ ret = -ENOSPC;
+ else
+ ret = 0;
+ spin_unlock(&space_info->lock);
+ return ret;
+ } else if (!should_alloc) {
+ spin_unlock(&space_info->lock);
+ return 0;
+ } else if (space_info->chunk_alloc) {
+ /*
+ * Someone is already allocating, so we need to block
+ * until this someone is finished and then loop to
+ * recheck if we should continue with our allocation
+ * attempt.
+ */
+ wait_for_alloc = true;
+ spin_unlock(&space_info->lock);
+ mutex_lock(&fs_info->chunk_mutex);
+ mutex_unlock(&fs_info->chunk_mutex);
+ } else {
+ /* Proceed with allocation */
+ space_info->chunk_alloc = 1;
+ wait_for_alloc = false;
+ spin_unlock(&space_info->lock);
+ }
+
+ cond_resched();
+ } while (wait_for_alloc);
+
+ mutex_lock(&fs_info->chunk_mutex);
+ trans->allocating_chunk = true;
+
+ /*
+ * If we have mixed data/metadata chunks we want to make sure we keep
+ * allocating mixed chunks instead of individual chunks.
+ */
+ if (btrfs_mixed_space_info(space_info))
+ flags |= (BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA);
+
+ /*
+ * if we're doing a data chunk, go ahead and make sure that
+ * we keep a reasonable number of metadata chunks allocated in the
+ * FS as well.
+ */
+ if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
+ fs_info->data_chunk_allocations++;
+ if (!(fs_info->data_chunk_allocations %
+ fs_info->metadata_ratio))
+ 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);
+ trans->allocating_chunk = false;
+
+ spin_lock(&space_info->lock);
+ if (ret < 0) {
+ if (ret == -ENOSPC)
+ space_info->full = 1;
+ else
+ goto out;
+ } else {
+ ret = 1;
+ space_info->max_extent_size = 0;
+ }
+
+ space_info->force_alloc = CHUNK_ALLOC_NO_FORCE;
+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;
+}
+
+static u64 get_profile_num_devs(struct btrfs_fs_info *fs_info, u64 type)
+{
+ u64 num_dev;
+
+ num_dev = btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)].devs_max;
+ if (!num_dev)
+ num_dev = fs_info->fs_devices->rw_devices;
+
+ return num_dev;
+}
+
+/*
+ * If @is_allocation is true, reserve space in the system space info necessary
+ * for allocating a chunk, otherwise if it's false, reserve space necessary for
+ * removing a chunk.
+ */
+void check_system_chunk(struct btrfs_trans_handle *trans, 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
+ * atomic and race free space reservation in the chunk block reserve.
+ */
+ lockdep_assert_held(&fs_info->chunk_mutex);
+
+ info = btrfs_find_space_info(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
+ spin_lock(&info->lock);
+ 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)) {
+ btrfs_info(fs_info, "left=%llu, need=%llu, flags=%llu",
+ left, thresh, type);
+ btrfs_dump_space_info(fs_info, info, 0, 0);
+ }
+
+ if (left < thresh) {
+ u64 flags = btrfs_system_alloc_profile(fs_info);
+
+ /*
+ * Ignore failure to create system chunk. We might end up not
+ * needing it, as we might not need to COW all nodes/leafs from
+ * 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);
+ }
+
+ if (!ret) {
+ ret = btrfs_block_rsv_add(fs_info->chunk_root,
+ &fs_info->chunk_block_rsv,
+ thresh, BTRFS_RESERVE_NO_FLUSH);
+ if (!ret)
+ trans->chunk_bytes_reserved += thresh;
+ }
+}
+
+void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
+{
+ struct btrfs_block_group_cache *block_group;
+ u64 last = 0;
+
+ while (1) {
+ struct inode *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;
+ 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->key.objectid + block_group->key.offset;
+ btrfs_put_block_group(block_group);
+ }
+}
+
+/*
+ * Must be called only after stopping all workers, since we could have block
+ * group caching kthreads running, and therefore they could race with us if we
+ * freed the block groups before stopping them.
+ */
+int btrfs_free_block_groups(struct btrfs_fs_info *info)
+{
+ struct btrfs_block_group_cache *block_group;
+ struct btrfs_space_info *space_info;
+ struct btrfs_caching_control *caching_ctl;
+ struct rb_node *n;
+
+ down_write(&info->commit_root_sem);
+ 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);
+
+ spin_lock(&info->unused_bgs_lock);
+ while (!list_empty(&info->unused_bgs)) {
+ block_group = list_first_entry(&info->unused_bgs,
+ struct btrfs_block_group_cache,
+ 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) {
+ block_group = rb_entry(n, struct btrfs_block_group_cache,
+ cache_node);
+ rb_erase(&block_group->cache_node,
+ &info->block_group_cache_tree);
+ RB_CLEAR_NODE(&block_group->cache_node);
+ spin_unlock(&info->block_group_cache_lock);
+
+ down_write(&block_group->space_info->groups_sem);
+ list_del(&block_group->list);
+ up_write(&block_group->space_info->groups_sem);
+
+ /*
+ * We haven't cached this block group, which means we could
+ * possibly have excluded extents on this block group.
+ */
+ if (block_group->cached == BTRFS_CACHE_NO ||
+ block_group->cached == BTRFS_CACHE_ERROR)
+ btrfs_free_excluded_extents(block_group);
+
+ btrfs_remove_free_space_cache(block_group);
+ ASSERT(block_group->cached != BTRFS_CACHE_STARTED);
+ 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);
+ btrfs_put_block_group(block_group);
+
+ spin_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();
+
+ btrfs_release_global_block_rsv(info);
+
+ while (!list_empty(&info->space_info)) {
+ space_info = list_entry(info->space_info.next,
+ struct btrfs_space_info,
+ list);
+
+ /*
+ * Do not hide this behind enospc_debug, this is actually
+ * 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);
+ list_del(&space_info->list);
+ btrfs_sysfs_remove_space_info(space_info);
+ }
+ return 0;
+}
*/
#include <linux/sched.h>
-#include <linux/sched/mm.h>
#include <linux/sched/signal.h>
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/percpu_counter.h>
#include <linux/lockdep.h>
#include <linux/crc32c.h>
+#include "misc.h"
#include "tree-log.h"
#include "disk-io.h"
#include "print-tree.h"
#include "locking.h"
#include "free-space-cache.h"
#include "free-space-tree.h"
-#include "math.h"
#include "sysfs.h"
#include "qgroup.h"
#include "ref-verify.h"
#include "space-info.h"
#include "block-rsv.h"
#include "delalloc-space.h"
+#include "block-group.h"
#undef SCRAMBLE_DELAYED_REFS
static int find_next_key(struct btrfs_path *path, int level,
struct btrfs_key *key);
-static noinline int
-block_group_cache_done(struct btrfs_block_group_cache *cache)
-{
- smp_mb();
- return cache->cached == BTRFS_CACHE_FINISHED ||
- cache->cached == BTRFS_CACHE_ERROR;
-}
-
static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
return (cache->flags & bits) == bits;
}
-void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
-{
- atomic_inc(&cache->count);
-}
-
-void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
-{
- if (atomic_dec_and_test(&cache->count)) {
- WARN_ON(cache->pinned > 0);
- WARN_ON(cache->reserved > 0);
-
- /*
- * If not empty, someone is still holding mutex of
- * full_stripe_lock, which can only be released by caller.
- * And it will definitely cause use-after-free when caller
- * tries to release full stripe lock.
- *
- * No better way to resolve, but only to warn.
- */
- WARN_ON(!RB_EMPTY_ROOT(&cache->full_stripe_locks_root.root));
- kfree(cache->free_space_ctl);
- kfree(cache);
- }
-}
-
-/*
- * this adds the block group to the fs_info rb tree for the block group
- * cache
- */
-static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
- struct btrfs_block_group_cache *block_group)
-{
- struct rb_node **p;
- struct rb_node *parent = NULL;
- struct btrfs_block_group_cache *cache;
-
- spin_lock(&info->block_group_cache_lock);
- p = &info->block_group_cache_tree.rb_node;
-
- while (*p) {
- parent = *p;
- cache = rb_entry(parent, struct btrfs_block_group_cache,
- cache_node);
- if (block_group->key.objectid < cache->key.objectid) {
- p = &(*p)->rb_left;
- } else if (block_group->key.objectid > cache->key.objectid) {
- p = &(*p)->rb_right;
- } else {
- spin_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->key.objectid)
- info->first_logical_byte = block_group->key.objectid;
-
- spin_unlock(&info->block_group_cache_lock);
-
- return 0;
-}
-
-/*
- * This will return the block group at or after bytenr if contains is 0, else
- * it will return the block group that contains the bytenr
- */
-static struct btrfs_block_group_cache *
-block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
- int contains)
-{
- struct btrfs_block_group_cache *cache, *ret = NULL;
- struct rb_node *n;
- u64 end, start;
-
- spin_lock(&info->block_group_cache_lock);
- n = info->block_group_cache_tree.rb_node;
-
- while (n) {
- cache = rb_entry(n, struct btrfs_block_group_cache,
- cache_node);
- end = cache->key.objectid + cache->key.offset - 1;
- start = cache->key.objectid;
-
- if (bytenr < start) {
- if (!contains && (!ret || start < ret->key.objectid))
- ret = cache;
- n = n->rb_left;
- } else if (bytenr > start) {
- if (contains && bytenr <= end) {
- ret = cache;
- break;
- }
- n = n->rb_right;
- } else {
- ret = cache;
- break;
- }
- }
- if (ret) {
- btrfs_get_block_group(ret);
- if (bytenr == 0 && info->first_logical_byte > ret->key.objectid)
- info->first_logical_byte = ret->key.objectid;
- }
- spin_unlock(&info->block_group_cache_lock);
-
- return ret;
-}
-
-static int add_excluded_extent(struct btrfs_fs_info *fs_info,
- u64 start, u64 num_bytes)
+int btrfs_add_excluded_extent(struct btrfs_fs_info *fs_info,
+ u64 start, u64 num_bytes)
{
u64 end = start + num_bytes - 1;
set_extent_bits(&fs_info->freed_extents[0],
return 0;
}
-static void free_excluded_extents(struct btrfs_block_group_cache *cache)
+void btrfs_free_excluded_extents(struct btrfs_block_group_cache *cache)
{
struct btrfs_fs_info *fs_info = cache->fs_info;
u64 start, end;
start, end, EXTENT_UPTODATE);
}
-static int exclude_super_stripes(struct btrfs_block_group_cache *cache)
+static u64 generic_ref_to_space_flags(struct btrfs_ref *ref)
{
- struct btrfs_fs_info *fs_info = cache->fs_info;
- u64 bytenr;
- u64 *logical;
- int stripe_len;
- int i, nr, ret;
-
- if (cache->key.objectid < BTRFS_SUPER_INFO_OFFSET) {
- stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->key.objectid;
- cache->bytes_super += stripe_len;
- ret = add_excluded_extent(fs_info, cache->key.objectid,
- stripe_len);
- if (ret)
- return ret;
- }
-
- for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
- bytenr = btrfs_sb_offset(i);
- ret = btrfs_rmap_block(fs_info, cache->key.objectid,
- bytenr, &logical, &nr, &stripe_len);
- if (ret)
- return ret;
-
- while (nr--) {
- u64 start, len;
-
- if (logical[nr] > cache->key.objectid +
- cache->key.offset)
- continue;
-
- if (logical[nr] + stripe_len <= cache->key.objectid)
- continue;
-
- start = logical[nr];
- if (start < cache->key.objectid) {
- start = cache->key.objectid;
- len = (logical[nr] + stripe_len) - start;
- } else {
- len = min_t(u64, stripe_len,
- cache->key.objectid +
- cache->key.offset - start);
- }
-
- cache->bytes_super += len;
- ret = add_excluded_extent(fs_info, start, len);
- if (ret) {
- kfree(logical);
- return ret;
- }
- }
-
- kfree(logical);
+ if (ref->type == BTRFS_REF_METADATA) {
+ if (ref->tree_ref.root == BTRFS_CHUNK_TREE_OBJECTID)
+ return BTRFS_BLOCK_GROUP_SYSTEM;
+ else
+ return BTRFS_BLOCK_GROUP_METADATA;
}
- return 0;
+ return BTRFS_BLOCK_GROUP_DATA;
}
-static struct btrfs_caching_control *
-get_caching_control(struct btrfs_block_group_cache *cache)
+static void add_pinned_bytes(struct btrfs_fs_info *fs_info,
+ struct btrfs_ref *ref)
{
- struct btrfs_caching_control *ctl;
-
- spin_lock(&cache->lock);
- if (!cache->caching_ctl) {
- spin_unlock(&cache->lock);
- return NULL;
- }
+ struct btrfs_space_info *space_info;
+ u64 flags = generic_ref_to_space_flags(ref);
- ctl = cache->caching_ctl;
- refcount_inc(&ctl->count);
- spin_unlock(&cache->lock);
- return ctl;
+ space_info = btrfs_find_space_info(fs_info, flags);
+ ASSERT(space_info);
+ percpu_counter_add_batch(&space_info->total_bytes_pinned, ref->len,
+ BTRFS_TOTAL_BYTES_PINNED_BATCH);
}
-static void put_caching_control(struct btrfs_caching_control *ctl)
+static void sub_pinned_bytes(struct btrfs_fs_info *fs_info,
+ struct btrfs_ref *ref)
{
- if (refcount_dec_and_test(&ctl->count))
- kfree(ctl);
-}
+ struct btrfs_space_info *space_info;
+ u64 flags = generic_ref_to_space_flags(ref);
-#ifdef CONFIG_BTRFS_DEBUG
-static void fragment_free_space(struct btrfs_block_group_cache *block_group)
-{
- struct btrfs_fs_info *fs_info = block_group->fs_info;
- u64 start = block_group->key.objectid;
- u64 len = block_group->key.offset;
- u64 chunk = block_group->flags & BTRFS_BLOCK_GROUP_METADATA ?
- fs_info->nodesize : fs_info->sectorsize;
- u64 step = chunk << 1;
-
- while (len > chunk) {
- btrfs_remove_free_space(block_group, start, chunk);
- start += step;
- if (len < step)
- len = 0;
- else
- len -= step;
- }
+ space_info = btrfs_find_space_info(fs_info, flags);
+ ASSERT(space_info);
+ percpu_counter_add_batch(&space_info->total_bytes_pinned, -ref->len,
+ BTRFS_TOTAL_BYTES_PINNED_BATCH);
}
-#endif
-/*
- * this is only called by cache_block_group, since we could have freed extents
- * we need to check the pinned_extents for any extents that can't be used yet
- * since their free space will be released as soon as the transaction commits.
- */
-u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
- u64 start, u64 end)
+/* simple helper to search for an existing data extent at a given offset */
+int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
{
- struct btrfs_fs_info *info = block_group->fs_info;
- u64 extent_start, extent_end, size, total_added = 0;
int ret;
+ struct btrfs_key key;
+ struct btrfs_path *path;
- while (start < end) {
- ret = find_first_extent_bit(info->pinned_extents, start,
- &extent_start, &extent_end,
- EXTENT_DIRTY | EXTENT_UPTODATE,
- NULL);
- if (ret)
- break;
-
- if (extent_start <= start) {
- start = extent_end + 1;
- } else if (extent_start > start && extent_start < end) {
- size = extent_start - start;
- total_added += size;
- ret = btrfs_add_free_space(block_group, start,
- size);
- BUG_ON(ret); /* -ENOMEM or logic error */
- start = extent_end + 1;
- } else {
- break;
- }
- }
-
- if (start < end) {
- size = end - start;
- total_added += size;
- ret = btrfs_add_free_space(block_group, start, size);
- BUG_ON(ret); /* -ENOMEM or logic error */
- }
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
- return total_added;
+ key.objectid = start;
+ key.offset = len;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
+ btrfs_free_path(path);
+ return ret;
}
-static int load_extent_tree_free(struct btrfs_caching_control *caching_ctl)
+/*
+ * helper function to lookup reference count and flags of a tree block.
+ *
+ * the head node for delayed ref is used to store the sum of all the
+ * reference count modifications queued up in the rbtree. the head
+ * node may also store the extent flags to set. This way you can check
+ * to see what the reference count and extent flags would be if all of
+ * the delayed refs are not processed.
+ */
+int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
+ struct btrfs_fs_info *fs_info, u64 bytenr,
+ u64 offset, int metadata, u64 *refs, u64 *flags)
{
- struct btrfs_block_group_cache *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_delayed_ref_head *head;
+ struct btrfs_delayed_ref_root *delayed_refs;
struct btrfs_path *path;
+ struct btrfs_extent_item *ei;
struct extent_buffer *leaf;
struct btrfs_key key;
- u64 total_found = 0;
- u64 last = 0;
- u32 nritems;
+ u32 item_size;
+ u64 num_refs;
+ u64 extent_flags;
int ret;
- bool wakeup = true;
+
+ /*
+ * If we don't have skinny metadata, don't bother doing anything
+ * different
+ */
+ if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
+ offset = fs_info->nodesize;
+ metadata = 0;
+ }
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
- last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
-
-#ifdef CONFIG_BTRFS_DEBUG
- /*
- * If we're fragmenting we don't want to make anybody think we can
- * allocate from this block group until we've had a chance to fragment
- * the free space.
- */
- if (btrfs_should_fragment_free_space(block_group))
- wakeup = false;
-#endif
- /*
- * We don't want to deadlock with somebody trying to allocate a new
- * extent for the extent root while also trying to search the extent
- * root to add free space. So we skip locking and search the commit
- * root, since its read-only
- */
- path->skip_locking = 1;
- path->search_commit_root = 1;
- path->reada = READA_FORWARD;
+ if (!trans) {
+ path->skip_locking = 1;
+ path->search_commit_root = 1;
+ }
- key.objectid = last;
- key.offset = 0;
- key.type = BTRFS_EXTENT_ITEM_KEY;
+search_again:
+ key.objectid = bytenr;
+ key.offset = offset;
+ if (metadata)
+ key.type = BTRFS_METADATA_ITEM_KEY;
+ else
+ key.type = BTRFS_EXTENT_ITEM_KEY;
-next:
- ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
+ ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
if (ret < 0)
- goto out;
-
- leaf = path->nodes[0];
- nritems = btrfs_header_nritems(leaf);
+ goto out_free;
- while (1) {
- if (btrfs_fs_closing(fs_info) > 1) {
- last = (u64)-1;
- break;
+ if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
+ if (path->slots[0]) {
+ path->slots[0]--;
+ btrfs_item_key_to_cpu(path->nodes[0], &key,
+ path->slots[0]);
+ if (key.objectid == bytenr &&
+ key.type == BTRFS_EXTENT_ITEM_KEY &&
+ key.offset == fs_info->nodesize)
+ ret = 0;
}
+ }
- if (path->slots[0] < nritems) {
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+ if (ret == 0) {
+ leaf = path->nodes[0];
+ item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+ if (item_size >= sizeof(*ei)) {
+ ei = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_item);
+ num_refs = btrfs_extent_refs(leaf, ei);
+ extent_flags = btrfs_extent_flags(leaf, ei);
} else {
- ret = find_next_key(path, 0, &key);
- if (ret)
- break;
-
- 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);
- cond_resched();
- mutex_lock(&caching_ctl->mutex);
- down_read(&fs_info->commit_root_sem);
- goto next;
- }
-
- ret = btrfs_next_leaf(extent_root, path);
- if (ret < 0)
- goto out;
- if (ret)
- break;
- leaf = path->nodes[0];
- nritems = btrfs_header_nritems(leaf);
- continue;
- }
-
- if (key.objectid < last) {
- key.objectid = last;
- key.offset = 0;
- key.type = BTRFS_EXTENT_ITEM_KEY;
+ ret = -EINVAL;
+ btrfs_print_v0_err(fs_info);
+ if (trans)
+ btrfs_abort_transaction(trans, ret);
+ else
+ btrfs_handle_fs_error(fs_info, ret, NULL);
- if (wakeup)
- caching_ctl->progress = last;
- btrfs_release_path(path);
- goto next;
+ goto out_free;
}
- if (key.objectid < block_group->key.objectid) {
- path->slots[0]++;
- continue;
- }
+ BUG_ON(num_refs == 0);
+ } else {
+ num_refs = 0;
+ extent_flags = 0;
+ ret = 0;
+ }
- if (key.objectid >= block_group->key.objectid +
- block_group->key.offset)
- break;
+ if (!trans)
+ goto out;
- if (key.type == BTRFS_EXTENT_ITEM_KEY ||
- key.type == BTRFS_METADATA_ITEM_KEY) {
- total_found += add_new_free_space(block_group, last,
- key.objectid);
- if (key.type == BTRFS_METADATA_ITEM_KEY)
- last = key.objectid +
- fs_info->nodesize;
- else
- last = key.objectid + key.offset;
+ delayed_refs = &trans->transaction->delayed_refs;
+ spin_lock(&delayed_refs->lock);
+ head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
+ if (head) {
+ if (!mutex_trylock(&head->mutex)) {
+ refcount_inc(&head->refs);
+ spin_unlock(&delayed_refs->lock);
- if (total_found > CACHING_CTL_WAKE_UP) {
- total_found = 0;
- if (wakeup)
- wake_up(&caching_ctl->wait);
- }
- }
- path->slots[0]++;
- }
- ret = 0;
+ btrfs_release_path(path);
- total_found += add_new_free_space(block_group, last,
- block_group->key.objectid +
- block_group->key.offset);
- caching_ctl->progress = (u64)-1;
+ /*
+ * Mutex was contended, block until it's released and try
+ * again
+ */
+ mutex_lock(&head->mutex);
+ mutex_unlock(&head->mutex);
+ btrfs_put_delayed_ref_head(head);
+ goto search_again;
+ }
+ spin_lock(&head->lock);
+ if (head->extent_op && head->extent_op->update_flags)
+ extent_flags |= head->extent_op->flags_to_set;
+ else
+ BUG_ON(num_refs == 0);
+ num_refs += head->ref_mod;
+ spin_unlock(&head->lock);
+ mutex_unlock(&head->mutex);
+ }
+ spin_unlock(&delayed_refs->lock);
out:
+ WARN_ON(num_refs == 0);
+ if (refs)
+ *refs = num_refs;
+ if (flags)
+ *flags = extent_flags;
+out_free:
btrfs_free_path(path);
return ret;
}
-static noinline void caching_thread(struct btrfs_work *work)
-{
- struct btrfs_block_group_cache *block_group;
- struct btrfs_fs_info *fs_info;
- struct btrfs_caching_control *caching_ctl;
- int ret;
-
- caching_ctl = container_of(work, struct btrfs_caching_control, work);
- block_group = caching_ctl->block_group;
- fs_info = block_group->fs_info;
-
- mutex_lock(&caching_ctl->mutex);
- down_read(&fs_info->commit_root_sem);
-
- if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
- ret = load_free_space_tree(caching_ctl);
- else
- ret = load_extent_tree_free(caching_ctl);
-
- spin_lock(&block_group->lock);
- block_group->caching_ctl = NULL;
- block_group->cached = ret ? BTRFS_CACHE_ERROR : BTRFS_CACHE_FINISHED;
- spin_unlock(&block_group->lock);
+/*
+ * Back reference rules. Back refs have three main goals:
+ *
+ * 1) differentiate between all holders of references to an extent so that
+ * when a reference is dropped we can make sure it was a valid reference
+ * before freeing the extent.
+ *
+ * 2) Provide enough information to quickly find the holders of an extent
+ * if we notice a given block is corrupted or bad.
+ *
+ * 3) Make it easy to migrate blocks for FS shrinking or storage pool
+ * maintenance. This is actually the same as #2, but with a slightly
+ * different use case.
+ *
+ * There are two kinds of back refs. The implicit back refs is optimized
+ * for pointers in non-shared tree blocks. For a given pointer in a block,
+ * back refs of this kind provide information about the block's owner tree
+ * and the pointer's key. These information allow us to find the block by
+ * b-tree searching. The full back refs is for pointers in tree blocks not
+ * referenced by their owner trees. The location of tree block is recorded
+ * in the back refs. Actually the full back refs is generic, and can be
+ * used in all cases the implicit back refs is used. The major shortcoming
+ * of the full back refs is its overhead. Every time a tree block gets
+ * COWed, we have to update back refs entry for all pointers in it.
+ *
+ * For a newly allocated tree block, we use implicit back refs for
+ * pointers in it. This means most tree related operations only involve
+ * implicit back refs. For a tree block created in old transaction, the
+ * only way to drop a reference to it is COW it. So we can detect the
+ * event that tree block loses its owner tree's reference and do the
+ * back refs conversion.
+ *
+ * When a tree block is COWed through a tree, there are four cases:
+ *
+ * The reference count of the block is one and the tree is the block's
+ * owner tree. Nothing to do in this case.
+ *
+ * The reference count of the block is one and the tree is not the
+ * block's owner tree. In this case, full back refs is used for pointers
+ * in the block. Remove these full back refs, add implicit back refs for
+ * every pointers in the new block.
+ *
+ * The reference count of the block is greater than one and the tree is
+ * the block's owner tree. In this case, implicit back refs is used for
+ * pointers in the block. Add full back refs for every pointers in the
+ * block, increase lower level extents' reference counts. The original
+ * implicit back refs are entailed to the new block.
+ *
+ * The reference count of the block is greater than one and the tree is
+ * not the block's owner tree. Add implicit back refs for every pointer in
+ * the new block, increase lower level extents' reference count.
+ *
+ * Back Reference Key composing:
+ *
+ * The key objectid corresponds to the first byte in the extent,
+ * The key type is used to differentiate between types of back refs.
+ * There are different meanings of the key offset for different types
+ * of back refs.
+ *
+ * File extents can be referenced by:
+ *
+ * - multiple snapshots, subvolumes, or different generations in one subvol
+ * - different files inside a single subvolume
+ * - different offsets inside a file (bookend extents in file.c)
+ *
+ * The extent ref structure for the implicit back refs has fields for:
+ *
+ * - Objectid of the subvolume root
+ * - objectid of the file holding the reference
+ * - original offset in the file
+ * - how many bookend extents
+ *
+ * The key offset for the implicit back refs is hash of the first
+ * three fields.
+ *
+ * The extent ref structure for the full back refs has field for:
+ *
+ * - number of pointers in the tree leaf
+ *
+ * The key offset for the implicit back refs is the first byte of
+ * the tree leaf
+ *
+ * When a file extent is allocated, The implicit back refs is used.
+ * the fields are filled in:
+ *
+ * (root_key.objectid, inode objectid, offset in file, 1)
+ *
+ * When a file extent is removed file truncation, we find the
+ * corresponding implicit back refs and check the following fields:
+ *
+ * (btrfs_header_owner(leaf), inode objectid, offset in file)
+ *
+ * Btree extents can be referenced by:
+ *
+ * - Different subvolumes
+ *
+ * Both the implicit back refs and the full back refs for tree blocks
+ * only consist of key. The key offset for the implicit back refs is
+ * objectid of block's owner tree. The key offset for the full back refs
+ * is the first byte of parent block.
+ *
+ * When implicit back refs is used, information about the lowest key and
+ * level of the tree block are required. These information are stored in
+ * tree block info structure.
+ */
-#ifdef CONFIG_BTRFS_DEBUG
- if (btrfs_should_fragment_free_space(block_group)) {
- u64 bytes_used;
+/*
+ * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
+ * is_data == BTRFS_REF_TYPE_DATA, data type is requiried,
+ * is_data == BTRFS_REF_TYPE_ANY, either type is OK.
+ */
+int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
+ struct btrfs_extent_inline_ref *iref,
+ enum btrfs_inline_ref_type is_data)
+{
+ int type = btrfs_extent_inline_ref_type(eb, iref);
+ u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
- spin_lock(&block_group->space_info->lock);
- spin_lock(&block_group->lock);
- bytes_used = block_group->key.offset -
- btrfs_block_group_used(&block_group->item);
- block_group->space_info->bytes_used += bytes_used >> 1;
- spin_unlock(&block_group->lock);
- spin_unlock(&block_group->space_info->lock);
- fragment_free_space(block_group);
+ if (type == BTRFS_TREE_BLOCK_REF_KEY ||
+ type == BTRFS_SHARED_BLOCK_REF_KEY ||
+ type == BTRFS_SHARED_DATA_REF_KEY ||
+ type == BTRFS_EXTENT_DATA_REF_KEY) {
+ if (is_data == BTRFS_REF_TYPE_BLOCK) {
+ if (type == BTRFS_TREE_BLOCK_REF_KEY)
+ return type;
+ if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
+ ASSERT(eb->fs_info);
+ /*
+ * Every shared one has parent tree
+ * block, which must be aligned to
+ * nodesize.
+ */
+ if (offset &&
+ IS_ALIGNED(offset, eb->fs_info->nodesize))
+ return type;
+ }
+ } else if (is_data == BTRFS_REF_TYPE_DATA) {
+ if (type == BTRFS_EXTENT_DATA_REF_KEY)
+ return type;
+ if (type == BTRFS_SHARED_DATA_REF_KEY) {
+ ASSERT(eb->fs_info);
+ /*
+ * Every shared one has parent tree
+ * block, which must be aligned to
+ * nodesize.
+ */
+ if (offset &&
+ IS_ALIGNED(offset, eb->fs_info->nodesize))
+ return type;
+ }
+ } else {
+ ASSERT(is_data == BTRFS_REF_TYPE_ANY);
+ return type;
+ }
}
-#endif
-
- caching_ctl->progress = (u64)-1;
- up_read(&fs_info->commit_root_sem);
- free_excluded_extents(block_group);
- mutex_unlock(&caching_ctl->mutex);
-
- wake_up(&caching_ctl->wait);
+ btrfs_print_leaf((struct extent_buffer *)eb);
+ btrfs_err(eb->fs_info, "eb %llu invalid extent inline ref type %d",
+ eb->start, type);
+ WARN_ON(1);
- put_caching_control(caching_ctl);
- btrfs_put_block_group(block_group);
+ return BTRFS_REF_TYPE_INVALID;
}
-static int cache_block_group(struct btrfs_block_group_cache *cache,
- int load_cache_only)
+u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
{
- DEFINE_WAIT(wait);
- struct btrfs_fs_info *fs_info = cache->fs_info;
- struct btrfs_caching_control *caching_ctl;
- int ret = 0;
-
- caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
- if (!caching_ctl)
- return -ENOMEM;
-
- INIT_LIST_HEAD(&caching_ctl->list);
- mutex_init(&caching_ctl->mutex);
- init_waitqueue_head(&caching_ctl->wait);
- caching_ctl->block_group = cache;
- caching_ctl->progress = cache->key.objectid;
- refcount_set(&caching_ctl->count, 1);
- btrfs_init_work(&caching_ctl->work, btrfs_cache_helper,
- 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;
+ u32 high_crc = ~(u32)0;
+ u32 low_crc = ~(u32)0;
+ __le64 lenum;
- ctl = cache->caching_ctl;
- refcount_inc(&ctl->count);
- prepare_to_wait(&ctl->wait, &wait, TASK_UNINTERRUPTIBLE);
- spin_unlock(&cache->lock);
+ lenum = cpu_to_le64(root_objectid);
+ high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
+ lenum = cpu_to_le64(owner);
+ low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
+ lenum = cpu_to_le64(offset);
+ low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
- schedule();
+ return ((u64)high_crc << 31) ^ (u64)low_crc;
+}
- finish_wait(&ctl->wait, &wait);
- put_caching_control(ctl);
- spin_lock(&cache->lock);
- }
+static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
+ struct btrfs_extent_data_ref *ref)
+{
+ return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
+ btrfs_extent_data_ref_objectid(leaf, ref),
+ btrfs_extent_data_ref_offset(leaf, ref));
+}
- if (cache->cached != BTRFS_CACHE_NO) {
- spin_unlock(&cache->lock);
- kfree(caching_ctl);
+static int match_extent_data_ref(struct extent_buffer *leaf,
+ struct btrfs_extent_data_ref *ref,
+ u64 root_objectid, u64 owner, u64 offset)
+{
+ if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
+ btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
+ btrfs_extent_data_ref_offset(leaf, ref) != offset)
return 0;
- }
- WARN_ON(cache->caching_ctl);
- cache->caching_ctl = caching_ctl;
- cache->cached = BTRFS_CACHE_FAST;
- 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->key.offset -
- btrfs_block_group_used(&cache->item);
- 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);
+ return 1;
+}
- wake_up(&caching_ctl->wait);
- if (ret == 1) {
- put_caching_control(caching_ctl);
- free_excluded_extents(cache);
- return 0;
- }
+static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ u64 bytenr, u64 parent,
+ u64 root_objectid,
+ u64 owner, u64 offset)
+{
+ struct btrfs_root *root = trans->fs_info->extent_root;
+ struct btrfs_key key;
+ struct btrfs_extent_data_ref *ref;
+ struct extent_buffer *leaf;
+ u32 nritems;
+ int ret;
+ int recow;
+ int err = -ENOENT;
+
+ key.objectid = bytenr;
+ if (parent) {
+ key.type = BTRFS_SHARED_DATA_REF_KEY;
+ key.offset = parent;
} 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);
+ key.type = BTRFS_EXTENT_DATA_REF_KEY;
+ key.offset = hash_extent_data_ref(root_objectid,
+ owner, offset);
+ }
+again:
+ recow = 0;
+ ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+ if (ret < 0) {
+ err = ret;
+ goto fail;
}
- if (load_cache_only) {
- put_caching_control(caching_ctl);
- return 0;
+ if (parent) {
+ if (!ret)
+ return 0;
+ goto fail;
}
- down_write(&fs_info->commit_root_sem);
- refcount_inc(&caching_ctl->count);
- list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
- up_write(&fs_info->commit_root_sem);
+ leaf = path->nodes[0];
+ nritems = btrfs_header_nritems(leaf);
+ while (1) {
+ if (path->slots[0] >= nritems) {
+ ret = btrfs_next_leaf(root, path);
+ if (ret < 0)
+ err = ret;
+ if (ret)
+ goto fail;
- btrfs_get_block_group(cache);
+ leaf = path->nodes[0];
+ nritems = btrfs_header_nritems(leaf);
+ recow = 1;
+ }
- btrfs_queue_work(fs_info->caching_workers, &caching_ctl->work);
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+ if (key.objectid != bytenr ||
+ key.type != BTRFS_EXTENT_DATA_REF_KEY)
+ goto fail;
- return ret;
-}
+ ref = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_data_ref);
-/*
- * return the block group that starts at or after bytenr
- */
-static struct btrfs_block_group_cache *
-btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
-{
- return block_group_cache_tree_search(info, bytenr, 0);
+ if (match_extent_data_ref(leaf, ref, root_objectid,
+ owner, offset)) {
+ if (recow) {
+ btrfs_release_path(path);
+ goto again;
+ }
+ err = 0;
+ break;
+ }
+ path->slots[0]++;
+ }
+fail:
+ return err;
}
-/*
- * return the block group that contains the given bytenr
- */
-struct btrfs_block_group_cache *btrfs_lookup_block_group(
- struct btrfs_fs_info *info,
- u64 bytenr)
+static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ u64 bytenr, u64 parent,
+ u64 root_objectid, u64 owner,
+ u64 offset, int refs_to_add)
{
- return block_group_cache_tree_search(info, bytenr, 1);
-}
+ struct btrfs_root *root = trans->fs_info->extent_root;
+ struct btrfs_key key;
+ struct extent_buffer *leaf;
+ u32 size;
+ u32 num_refs;
+ int ret;
-static u64 generic_ref_to_space_flags(struct btrfs_ref *ref)
-{
- if (ref->type == BTRFS_REF_METADATA) {
- if (ref->tree_ref.root == BTRFS_CHUNK_TREE_OBJECTID)
- return BTRFS_BLOCK_GROUP_SYSTEM;
- else
- return BTRFS_BLOCK_GROUP_METADATA;
+ key.objectid = bytenr;
+ if (parent) {
+ key.type = BTRFS_SHARED_DATA_REF_KEY;
+ key.offset = parent;
+ size = sizeof(struct btrfs_shared_data_ref);
+ } else {
+ key.type = BTRFS_EXTENT_DATA_REF_KEY;
+ key.offset = hash_extent_data_ref(root_objectid,
+ owner, offset);
+ size = sizeof(struct btrfs_extent_data_ref);
}
- return BTRFS_BLOCK_GROUP_DATA;
-}
-
-static void add_pinned_bytes(struct btrfs_fs_info *fs_info,
- struct btrfs_ref *ref)
-{
- struct btrfs_space_info *space_info;
- u64 flags = generic_ref_to_space_flags(ref);
- space_info = btrfs_find_space_info(fs_info, flags);
- ASSERT(space_info);
- percpu_counter_add_batch(&space_info->total_bytes_pinned, ref->len,
- BTRFS_TOTAL_BYTES_PINNED_BATCH);
-}
+ ret = btrfs_insert_empty_item(trans, root, path, &key, size);
+ if (ret && ret != -EEXIST)
+ goto fail;
-static void sub_pinned_bytes(struct btrfs_fs_info *fs_info,
- struct btrfs_ref *ref)
-{
- struct btrfs_space_info *space_info;
- u64 flags = generic_ref_to_space_flags(ref);
+ leaf = path->nodes[0];
+ if (parent) {
+ struct btrfs_shared_data_ref *ref;
+ ref = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_shared_data_ref);
+ if (ret == 0) {
+ btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
+ } else {
+ num_refs = btrfs_shared_data_ref_count(leaf, ref);
+ num_refs += refs_to_add;
+ btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
+ }
+ } else {
+ struct btrfs_extent_data_ref *ref;
+ while (ret == -EEXIST) {
+ ref = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_data_ref);
+ if (match_extent_data_ref(leaf, ref, root_objectid,
+ owner, offset))
+ break;
+ btrfs_release_path(path);
+ key.offset++;
+ ret = btrfs_insert_empty_item(trans, root, path, &key,
+ size);
+ if (ret && ret != -EEXIST)
+ goto fail;
- space_info = btrfs_find_space_info(fs_info, flags);
- ASSERT(space_info);
- percpu_counter_add_batch(&space_info->total_bytes_pinned, -ref->len,
- BTRFS_TOTAL_BYTES_PINNED_BATCH);
+ leaf = path->nodes[0];
+ }
+ ref = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_data_ref);
+ if (ret == 0) {
+ btrfs_set_extent_data_ref_root(leaf, ref,
+ root_objectid);
+ btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
+ btrfs_set_extent_data_ref_offset(leaf, ref, offset);
+ btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
+ } else {
+ num_refs = btrfs_extent_data_ref_count(leaf, ref);
+ num_refs += refs_to_add;
+ btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
+ }
+ }
+ btrfs_mark_buffer_dirty(leaf);
+ ret = 0;
+fail:
+ btrfs_release_path(path);
+ return ret;
}
-/* simple helper to search for an existing data extent at a given offset */
-int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
+static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ int refs_to_drop, int *last_ref)
{
- int ret;
struct btrfs_key key;
- struct btrfs_path *path;
+ struct btrfs_extent_data_ref *ref1 = NULL;
+ struct btrfs_shared_data_ref *ref2 = NULL;
+ struct extent_buffer *leaf;
+ u32 num_refs = 0;
+ int ret = 0;
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
+ leaf = path->nodes[0];
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- key.objectid = start;
- key.offset = len;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
- btrfs_free_path(path);
+ if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
+ ref1 = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_data_ref);
+ num_refs = btrfs_extent_data_ref_count(leaf, ref1);
+ } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
+ ref2 = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_shared_data_ref);
+ num_refs = btrfs_shared_data_ref_count(leaf, ref2);
+ } else if (unlikely(key.type == BTRFS_EXTENT_REF_V0_KEY)) {
+ btrfs_print_v0_err(trans->fs_info);
+ btrfs_abort_transaction(trans, -EINVAL);
+ return -EINVAL;
+ } else {
+ BUG();
+ }
+
+ BUG_ON(num_refs < refs_to_drop);
+ num_refs -= refs_to_drop;
+
+ if (num_refs == 0) {
+ ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
+ *last_ref = 1;
+ } else {
+ if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
+ btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
+ else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
+ btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
+ btrfs_mark_buffer_dirty(leaf);
+ }
return ret;
}
-/*
- * helper function to lookup reference count and flags of a tree block.
- *
- * the head node for delayed ref is used to store the sum of all the
- * reference count modifications queued up in the rbtree. the head
- * node may also store the extent flags to set. This way you can check
- * to see what the reference count and extent flags would be if all of
- * the delayed refs are not processed.
- */
-int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
- struct btrfs_fs_info *fs_info, u64 bytenr,
- u64 offset, int metadata, u64 *refs, u64 *flags)
+static noinline u32 extent_data_ref_count(struct btrfs_path *path,
+ struct btrfs_extent_inline_ref *iref)
{
- struct btrfs_delayed_ref_head *head;
- struct btrfs_delayed_ref_root *delayed_refs;
- struct btrfs_path *path;
- struct btrfs_extent_item *ei;
- struct extent_buffer *leaf;
struct btrfs_key key;
- u32 item_size;
- u64 num_refs;
- u64 extent_flags;
+ struct extent_buffer *leaf;
+ struct btrfs_extent_data_ref *ref1;
+ struct btrfs_shared_data_ref *ref2;
+ u32 num_refs = 0;
+ int type;
+
+ leaf = path->nodes[0];
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+
+ BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
+ if (iref) {
+ /*
+ * If type is invalid, we should have bailed out earlier than
+ * this call.
+ */
+ type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
+ ASSERT(type != BTRFS_REF_TYPE_INVALID);
+ if (type == BTRFS_EXTENT_DATA_REF_KEY) {
+ ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
+ num_refs = btrfs_extent_data_ref_count(leaf, ref1);
+ } else {
+ ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
+ num_refs = btrfs_shared_data_ref_count(leaf, ref2);
+ }
+ } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
+ ref1 = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_data_ref);
+ num_refs = btrfs_extent_data_ref_count(leaf, ref1);
+ } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
+ ref2 = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_shared_data_ref);
+ num_refs = btrfs_shared_data_ref_count(leaf, ref2);
+ } else {
+ WARN_ON(1);
+ }
+ return num_refs;
+}
+
+static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ u64 bytenr, u64 parent,
+ u64 root_objectid)
+{
+ struct btrfs_root *root = trans->fs_info->extent_root;
+ struct btrfs_key key;
int ret;
- /*
- * If we don't have skinny metadata, don't bother doing anything
- * different
- */
- if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
- offset = fs_info->nodesize;
- metadata = 0;
+ key.objectid = bytenr;
+ if (parent) {
+ key.type = BTRFS_SHARED_BLOCK_REF_KEY;
+ key.offset = parent;
+ } else {
+ key.type = BTRFS_TREE_BLOCK_REF_KEY;
+ key.offset = root_objectid;
}
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
+ ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+ if (ret > 0)
+ ret = -ENOENT;
+ return ret;
+}
- if (!trans) {
- path->skip_locking = 1;
- path->search_commit_root = 1;
+static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ u64 bytenr, u64 parent,
+ u64 root_objectid)
+{
+ struct btrfs_key key;
+ int ret;
+
+ key.objectid = bytenr;
+ if (parent) {
+ key.type = BTRFS_SHARED_BLOCK_REF_KEY;
+ key.offset = parent;
+ } else {
+ key.type = BTRFS_TREE_BLOCK_REF_KEY;
+ key.offset = root_objectid;
}
-search_again:
+ ret = btrfs_insert_empty_item(trans, trans->fs_info->extent_root,
+ path, &key, 0);
+ btrfs_release_path(path);
+ return ret;
+}
+
+static inline int extent_ref_type(u64 parent, u64 owner)
+{
+ int type;
+ if (owner < BTRFS_FIRST_FREE_OBJECTID) {
+ if (parent > 0)
+ type = BTRFS_SHARED_BLOCK_REF_KEY;
+ else
+ type = BTRFS_TREE_BLOCK_REF_KEY;
+ } else {
+ if (parent > 0)
+ type = BTRFS_SHARED_DATA_REF_KEY;
+ else
+ type = BTRFS_EXTENT_DATA_REF_KEY;
+ }
+ return type;
+}
+
+static int find_next_key(struct btrfs_path *path, int level,
+ struct btrfs_key *key)
+
+{
+ for (; level < BTRFS_MAX_LEVEL; level++) {
+ if (!path->nodes[level])
+ break;
+ if (path->slots[level] + 1 >=
+ btrfs_header_nritems(path->nodes[level]))
+ continue;
+ if (level == 0)
+ btrfs_item_key_to_cpu(path->nodes[level], key,
+ path->slots[level] + 1);
+ else
+ btrfs_node_key_to_cpu(path->nodes[level], key,
+ path->slots[level] + 1);
+ return 0;
+ }
+ return 1;
+}
+
+/*
+ * look for inline back ref. if back ref is found, *ref_ret is set
+ * to the address of inline back ref, and 0 is returned.
+ *
+ * if back ref isn't found, *ref_ret is set to the address where it
+ * should be inserted, and -ENOENT is returned.
+ *
+ * if insert is true and there are too many inline back refs, the path
+ * points to the extent item, and -EAGAIN is returned.
+ *
+ * NOTE: inline back refs are ordered in the same way that back ref
+ * items in the tree are ordered.
+ */
+static noinline_for_stack
+int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ struct btrfs_extent_inline_ref **ref_ret,
+ u64 bytenr, u64 num_bytes,
+ u64 parent, u64 root_objectid,
+ u64 owner, u64 offset, int insert)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_root *root = fs_info->extent_root;
+ struct btrfs_key key;
+ struct extent_buffer *leaf;
+ struct btrfs_extent_item *ei;
+ struct btrfs_extent_inline_ref *iref;
+ u64 flags;
+ u64 item_size;
+ unsigned long ptr;
+ unsigned long end;
+ int extra_size;
+ int type;
+ int want;
+ int ret;
+ int err = 0;
+ bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
+ int needed;
+
key.objectid = bytenr;
- key.offset = offset;
- if (metadata)
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = num_bytes;
+
+ want = extent_ref_type(parent, owner);
+ if (insert) {
+ extra_size = btrfs_extent_inline_ref_size(want);
+ path->keep_locks = 1;
+ } else
+ extra_size = -1;
+
+ /*
+ * Owner is our level, so we can just add one to get the level for the
+ * block we are interested in.
+ */
+ if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
key.type = BTRFS_METADATA_ITEM_KEY;
- else
- key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = owner;
+ }
- ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
- if (ret < 0)
- goto out_free;
+again:
+ ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
+ if (ret < 0) {
+ err = ret;
+ goto out;
+ }
- if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
+ /*
+ * We may be a newly converted file system which still has the old fat
+ * extent entries for metadata, so try and see if we have one of those.
+ */
+ if (ret > 0 && skinny_metadata) {
+ skinny_metadata = false;
if (path->slots[0]) {
path->slots[0]--;
btrfs_item_key_to_cpu(path->nodes[0], &key,
path->slots[0]);
if (key.objectid == bytenr &&
key.type == BTRFS_EXTENT_ITEM_KEY &&
- key.offset == fs_info->nodesize)
+ key.offset == num_bytes)
ret = 0;
}
- }
-
- if (ret == 0) {
- leaf = path->nodes[0];
- item_size = btrfs_item_size_nr(leaf, path->slots[0]);
- if (item_size >= sizeof(*ei)) {
- ei = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_item);
- num_refs = btrfs_extent_refs(leaf, ei);
- extent_flags = btrfs_extent_flags(leaf, ei);
- } else {
- ret = -EINVAL;
- btrfs_print_v0_err(fs_info);
- if (trans)
- btrfs_abort_transaction(trans, ret);
- else
- btrfs_handle_fs_error(fs_info, ret, NULL);
-
- goto out_free;
+ if (ret) {
+ key.objectid = bytenr;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = num_bytes;
+ btrfs_release_path(path);
+ goto again;
}
+ }
- BUG_ON(num_refs == 0);
- } else {
- num_refs = 0;
- extent_flags = 0;
- ret = 0;
+ if (ret && !insert) {
+ err = -ENOENT;
+ goto out;
+ } else if (WARN_ON(ret)) {
+ err = -EIO;
+ goto out;
}
- if (!trans)
+ leaf = path->nodes[0];
+ item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+ if (unlikely(item_size < sizeof(*ei))) {
+ err = -EINVAL;
+ btrfs_print_v0_err(fs_info);
+ btrfs_abort_transaction(trans, err);
goto out;
+ }
- delayed_refs = &trans->transaction->delayed_refs;
- spin_lock(&delayed_refs->lock);
- head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
- if (head) {
- if (!mutex_trylock(&head->mutex)) {
- refcount_inc(&head->refs);
- spin_unlock(&delayed_refs->lock);
-
- btrfs_release_path(path);
-
- /*
- * Mutex was contended, block until it's released and try
- * again
- */
- mutex_lock(&head->mutex);
- mutex_unlock(&head->mutex);
- btrfs_put_delayed_ref_head(head);
- goto search_again;
- }
- spin_lock(&head->lock);
- if (head->extent_op && head->extent_op->update_flags)
- extent_flags |= head->extent_op->flags_to_set;
- else
- BUG_ON(num_refs == 0);
-
- num_refs += head->ref_mod;
- spin_unlock(&head->lock);
- mutex_unlock(&head->mutex);
- }
- spin_unlock(&delayed_refs->lock);
-out:
- WARN_ON(num_refs == 0);
- if (refs)
- *refs = num_refs;
- if (flags)
- *flags = extent_flags;
-out_free:
- btrfs_free_path(path);
- return ret;
-}
-
-/*
- * Back reference rules. Back refs have three main goals:
- *
- * 1) differentiate between all holders of references to an extent so that
- * when a reference is dropped we can make sure it was a valid reference
- * before freeing the extent.
- *
- * 2) Provide enough information to quickly find the holders of an extent
- * if we notice a given block is corrupted or bad.
- *
- * 3) Make it easy to migrate blocks for FS shrinking or storage pool
- * maintenance. This is actually the same as #2, but with a slightly
- * different use case.
- *
- * There are two kinds of back refs. The implicit back refs is optimized
- * for pointers in non-shared tree blocks. For a given pointer in a block,
- * back refs of this kind provide information about the block's owner tree
- * and the pointer's key. These information allow us to find the block by
- * b-tree searching. The full back refs is for pointers in tree blocks not
- * referenced by their owner trees. The location of tree block is recorded
- * in the back refs. Actually the full back refs is generic, and can be
- * used in all cases the implicit back refs is used. The major shortcoming
- * of the full back refs is its overhead. Every time a tree block gets
- * COWed, we have to update back refs entry for all pointers in it.
- *
- * For a newly allocated tree block, we use implicit back refs for
- * pointers in it. This means most tree related operations only involve
- * implicit back refs. For a tree block created in old transaction, the
- * only way to drop a reference to it is COW it. So we can detect the
- * event that tree block loses its owner tree's reference and do the
- * back refs conversion.
- *
- * When a tree block is COWed through a tree, there are four cases:
- *
- * The reference count of the block is one and the tree is the block's
- * owner tree. Nothing to do in this case.
- *
- * The reference count of the block is one and the tree is not the
- * block's owner tree. In this case, full back refs is used for pointers
- * in the block. Remove these full back refs, add implicit back refs for
- * every pointers in the new block.
- *
- * The reference count of the block is greater than one and the tree is
- * the block's owner tree. In this case, implicit back refs is used for
- * pointers in the block. Add full back refs for every pointers in the
- * block, increase lower level extents' reference counts. The original
- * implicit back refs are entailed to the new block.
- *
- * The reference count of the block is greater than one and the tree is
- * not the block's owner tree. Add implicit back refs for every pointer in
- * the new block, increase lower level extents' reference count.
- *
- * Back Reference Key composing:
- *
- * The key objectid corresponds to the first byte in the extent,
- * The key type is used to differentiate between types of back refs.
- * There are different meanings of the key offset for different types
- * of back refs.
- *
- * File extents can be referenced by:
- *
- * - multiple snapshots, subvolumes, or different generations in one subvol
- * - different files inside a single subvolume
- * - different offsets inside a file (bookend extents in file.c)
- *
- * The extent ref structure for the implicit back refs has fields for:
- *
- * - Objectid of the subvolume root
- * - objectid of the file holding the reference
- * - original offset in the file
- * - how many bookend extents
- *
- * The key offset for the implicit back refs is hash of the first
- * three fields.
- *
- * The extent ref structure for the full back refs has field for:
- *
- * - number of pointers in the tree leaf
- *
- * The key offset for the implicit back refs is the first byte of
- * the tree leaf
- *
- * When a file extent is allocated, The implicit back refs is used.
- * the fields are filled in:
- *
- * (root_key.objectid, inode objectid, offset in file, 1)
- *
- * When a file extent is removed file truncation, we find the
- * corresponding implicit back refs and check the following fields:
- *
- * (btrfs_header_owner(leaf), inode objectid, offset in file)
- *
- * Btree extents can be referenced by:
- *
- * - Different subvolumes
- *
- * Both the implicit back refs and the full back refs for tree blocks
- * only consist of key. The key offset for the implicit back refs is
- * objectid of block's owner tree. The key offset for the full back refs
- * is the first byte of parent block.
- *
- * When implicit back refs is used, information about the lowest key and
- * level of the tree block are required. These information are stored in
- * tree block info structure.
- */
-
-/*
- * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
- * is_data == BTRFS_REF_TYPE_DATA, data type is requiried,
- * is_data == BTRFS_REF_TYPE_ANY, either type is OK.
- */
-int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
- struct btrfs_extent_inline_ref *iref,
- enum btrfs_inline_ref_type is_data)
-{
- int type = btrfs_extent_inline_ref_type(eb, iref);
- u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
-
- if (type == BTRFS_TREE_BLOCK_REF_KEY ||
- type == BTRFS_SHARED_BLOCK_REF_KEY ||
- type == BTRFS_SHARED_DATA_REF_KEY ||
- type == BTRFS_EXTENT_DATA_REF_KEY) {
- if (is_data == BTRFS_REF_TYPE_BLOCK) {
- if (type == BTRFS_TREE_BLOCK_REF_KEY)
- return type;
- if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
- ASSERT(eb->fs_info);
- /*
- * Every shared one has parent tree
- * block, which must be aligned to
- * nodesize.
- */
- if (offset &&
- IS_ALIGNED(offset, eb->fs_info->nodesize))
- return type;
- }
- } else if (is_data == BTRFS_REF_TYPE_DATA) {
- if (type == BTRFS_EXTENT_DATA_REF_KEY)
- return type;
- if (type == BTRFS_SHARED_DATA_REF_KEY) {
- ASSERT(eb->fs_info);
- /*
- * Every shared one has parent tree
- * block, which must be aligned to
- * nodesize.
- */
- if (offset &&
- IS_ALIGNED(offset, eb->fs_info->nodesize))
- return type;
- }
- } else {
- ASSERT(is_data == BTRFS_REF_TYPE_ANY);
- return type;
- }
- }
-
- btrfs_print_leaf((struct extent_buffer *)eb);
- btrfs_err(eb->fs_info, "eb %llu invalid extent inline ref type %d",
- eb->start, type);
- WARN_ON(1);
-
- return BTRFS_REF_TYPE_INVALID;
-}
-
-static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
-{
- u32 high_crc = ~(u32)0;
- u32 low_crc = ~(u32)0;
- __le64 lenum;
-
- lenum = cpu_to_le64(root_objectid);
- high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
- lenum = cpu_to_le64(owner);
- low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
- lenum = cpu_to_le64(offset);
- low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
-
- return ((u64)high_crc << 31) ^ (u64)low_crc;
-}
-
-static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
- struct btrfs_extent_data_ref *ref)
-{
- return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
- btrfs_extent_data_ref_objectid(leaf, ref),
- btrfs_extent_data_ref_offset(leaf, ref));
-}
-
-static int match_extent_data_ref(struct extent_buffer *leaf,
- struct btrfs_extent_data_ref *ref,
- u64 root_objectid, u64 owner, u64 offset)
-{
- if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
- btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
- btrfs_extent_data_ref_offset(leaf, ref) != offset)
- return 0;
- return 1;
-}
-
-static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- u64 bytenr, u64 parent,
- u64 root_objectid,
- u64 owner, u64 offset)
-{
- struct btrfs_root *root = trans->fs_info->extent_root;
- struct btrfs_key key;
- struct btrfs_extent_data_ref *ref;
- struct extent_buffer *leaf;
- u32 nritems;
- int ret;
- int recow;
- int err = -ENOENT;
-
- key.objectid = bytenr;
- if (parent) {
- key.type = BTRFS_SHARED_DATA_REF_KEY;
- key.offset = parent;
- } else {
- key.type = BTRFS_EXTENT_DATA_REF_KEY;
- key.offset = hash_extent_data_ref(root_objectid,
- owner, offset);
- }
-again:
- recow = 0;
- ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
- if (ret < 0) {
- err = ret;
- goto fail;
- }
-
- if (parent) {
- if (!ret)
- return 0;
- goto fail;
- }
-
- leaf = path->nodes[0];
- nritems = btrfs_header_nritems(leaf);
- while (1) {
- if (path->slots[0] >= nritems) {
- ret = btrfs_next_leaf(root, path);
- if (ret < 0)
- err = ret;
- if (ret)
- goto fail;
-
- leaf = path->nodes[0];
- nritems = btrfs_header_nritems(leaf);
- recow = 1;
- }
-
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- if (key.objectid != bytenr ||
- key.type != BTRFS_EXTENT_DATA_REF_KEY)
- goto fail;
-
- ref = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_data_ref);
-
- if (match_extent_data_ref(leaf, ref, root_objectid,
- owner, offset)) {
- if (recow) {
- btrfs_release_path(path);
- goto again;
- }
- err = 0;
- break;
- }
- path->slots[0]++;
- }
-fail:
- return err;
-}
-
-static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- u64 bytenr, u64 parent,
- u64 root_objectid, u64 owner,
- u64 offset, int refs_to_add)
-{
- struct btrfs_root *root = trans->fs_info->extent_root;
- struct btrfs_key key;
- struct extent_buffer *leaf;
- u32 size;
- u32 num_refs;
- int ret;
-
- key.objectid = bytenr;
- if (parent) {
- key.type = BTRFS_SHARED_DATA_REF_KEY;
- key.offset = parent;
- size = sizeof(struct btrfs_shared_data_ref);
- } else {
- key.type = BTRFS_EXTENT_DATA_REF_KEY;
- key.offset = hash_extent_data_ref(root_objectid,
- owner, offset);
- size = sizeof(struct btrfs_extent_data_ref);
- }
-
- ret = btrfs_insert_empty_item(trans, root, path, &key, size);
- if (ret && ret != -EEXIST)
- goto fail;
-
- leaf = path->nodes[0];
- if (parent) {
- struct btrfs_shared_data_ref *ref;
- ref = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_shared_data_ref);
- if (ret == 0) {
- btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
- } else {
- num_refs = btrfs_shared_data_ref_count(leaf, ref);
- num_refs += refs_to_add;
- btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
- }
- } else {
- struct btrfs_extent_data_ref *ref;
- while (ret == -EEXIST) {
- ref = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_data_ref);
- if (match_extent_data_ref(leaf, ref, root_objectid,
- owner, offset))
- break;
- btrfs_release_path(path);
- key.offset++;
- ret = btrfs_insert_empty_item(trans, root, path, &key,
- size);
- if (ret && ret != -EEXIST)
- goto fail;
-
- leaf = path->nodes[0];
- }
- ref = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_data_ref);
- if (ret == 0) {
- btrfs_set_extent_data_ref_root(leaf, ref,
- root_objectid);
- btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
- btrfs_set_extent_data_ref_offset(leaf, ref, offset);
- btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
- } else {
- num_refs = btrfs_extent_data_ref_count(leaf, ref);
- num_refs += refs_to_add;
- btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
- }
- }
- btrfs_mark_buffer_dirty(leaf);
- ret = 0;
-fail:
- btrfs_release_path(path);
- return ret;
-}
-
-static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- int refs_to_drop, int *last_ref)
-{
- struct btrfs_key key;
- struct btrfs_extent_data_ref *ref1 = NULL;
- struct btrfs_shared_data_ref *ref2 = NULL;
- struct extent_buffer *leaf;
- u32 num_refs = 0;
- int ret = 0;
-
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-
- if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
- ref1 = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_data_ref);
- num_refs = btrfs_extent_data_ref_count(leaf, ref1);
- } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
- ref2 = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_shared_data_ref);
- num_refs = btrfs_shared_data_ref_count(leaf, ref2);
- } else if (unlikely(key.type == BTRFS_EXTENT_REF_V0_KEY)) {
- btrfs_print_v0_err(trans->fs_info);
- btrfs_abort_transaction(trans, -EINVAL);
- return -EINVAL;
- } else {
- BUG();
- }
-
- BUG_ON(num_refs < refs_to_drop);
- num_refs -= refs_to_drop;
-
- if (num_refs == 0) {
- ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
- *last_ref = 1;
- } else {
- if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
- btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
- else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
- btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
- btrfs_mark_buffer_dirty(leaf);
- }
- return ret;
-}
-
-static noinline u32 extent_data_ref_count(struct btrfs_path *path,
- struct btrfs_extent_inline_ref *iref)
-{
- struct btrfs_key key;
- struct extent_buffer *leaf;
- struct btrfs_extent_data_ref *ref1;
- struct btrfs_shared_data_ref *ref2;
- u32 num_refs = 0;
- int type;
-
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-
- BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
- if (iref) {
- /*
- * If type is invalid, we should have bailed out earlier than
- * this call.
- */
- type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
- ASSERT(type != BTRFS_REF_TYPE_INVALID);
- if (type == BTRFS_EXTENT_DATA_REF_KEY) {
- ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
- num_refs = btrfs_extent_data_ref_count(leaf, ref1);
- } else {
- ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
- num_refs = btrfs_shared_data_ref_count(leaf, ref2);
- }
- } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
- ref1 = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_data_ref);
- num_refs = btrfs_extent_data_ref_count(leaf, ref1);
- } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
- ref2 = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_shared_data_ref);
- num_refs = btrfs_shared_data_ref_count(leaf, ref2);
- } else {
- WARN_ON(1);
- }
- return num_refs;
-}
-
-static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- u64 bytenr, u64 parent,
- u64 root_objectid)
-{
- struct btrfs_root *root = trans->fs_info->extent_root;
- struct btrfs_key key;
- int ret;
-
- key.objectid = bytenr;
- if (parent) {
- key.type = BTRFS_SHARED_BLOCK_REF_KEY;
- key.offset = parent;
- } else {
- key.type = BTRFS_TREE_BLOCK_REF_KEY;
- key.offset = root_objectid;
- }
-
- ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
- if (ret > 0)
- ret = -ENOENT;
- return ret;
-}
-
-static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- u64 bytenr, u64 parent,
- u64 root_objectid)
-{
- struct btrfs_key key;
- int ret;
-
- key.objectid = bytenr;
- if (parent) {
- key.type = BTRFS_SHARED_BLOCK_REF_KEY;
- key.offset = parent;
- } else {
- key.type = BTRFS_TREE_BLOCK_REF_KEY;
- key.offset = root_objectid;
- }
-
- ret = btrfs_insert_empty_item(trans, trans->fs_info->extent_root,
- path, &key, 0);
- btrfs_release_path(path);
- return ret;
-}
-
-static inline int extent_ref_type(u64 parent, u64 owner)
-{
- int type;
- if (owner < BTRFS_FIRST_FREE_OBJECTID) {
- if (parent > 0)
- type = BTRFS_SHARED_BLOCK_REF_KEY;
- else
- type = BTRFS_TREE_BLOCK_REF_KEY;
- } else {
- if (parent > 0)
- type = BTRFS_SHARED_DATA_REF_KEY;
- else
- type = BTRFS_EXTENT_DATA_REF_KEY;
- }
- return type;
-}
-
-static int find_next_key(struct btrfs_path *path, int level,
- struct btrfs_key *key)
-
-{
- for (; level < BTRFS_MAX_LEVEL; level++) {
- if (!path->nodes[level])
- break;
- if (path->slots[level] + 1 >=
- btrfs_header_nritems(path->nodes[level]))
- continue;
- if (level == 0)
- btrfs_item_key_to_cpu(path->nodes[level], key,
- path->slots[level] + 1);
- else
- btrfs_node_key_to_cpu(path->nodes[level], key,
- path->slots[level] + 1);
- return 0;
- }
- return 1;
-}
-
-/*
- * look for inline back ref. if back ref is found, *ref_ret is set
- * to the address of inline back ref, and 0 is returned.
- *
- * if back ref isn't found, *ref_ret is set to the address where it
- * should be inserted, and -ENOENT is returned.
- *
- * if insert is true and there are too many inline back refs, the path
- * points to the extent item, and -EAGAIN is returned.
- *
- * NOTE: inline back refs are ordered in the same way that back ref
- * items in the tree are ordered.
- */
-static noinline_for_stack
-int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- struct btrfs_extent_inline_ref **ref_ret,
- u64 bytenr, u64 num_bytes,
- u64 parent, u64 root_objectid,
- u64 owner, u64 offset, int insert)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_root *root = fs_info->extent_root;
- struct btrfs_key key;
- struct extent_buffer *leaf;
- struct btrfs_extent_item *ei;
- struct btrfs_extent_inline_ref *iref;
- u64 flags;
- u64 item_size;
- unsigned long ptr;
- unsigned long end;
- int extra_size;
- int type;
- int want;
- int ret;
- int err = 0;
- bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
- int needed;
-
- key.objectid = bytenr;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- key.offset = num_bytes;
-
- want = extent_ref_type(parent, owner);
- if (insert) {
- extra_size = btrfs_extent_inline_ref_size(want);
- path->keep_locks = 1;
- } else
- extra_size = -1;
-
- /*
- * Owner is our level, so we can just add one to get the level for the
- * block we are interested in.
- */
- if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
- key.type = BTRFS_METADATA_ITEM_KEY;
- key.offset = owner;
- }
-
-again:
- ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
- if (ret < 0) {
- err = ret;
- goto out;
- }
-
- /*
- * We may be a newly converted file system which still has the old fat
- * extent entries for metadata, so try and see if we have one of those.
- */
- if (ret > 0 && skinny_metadata) {
- skinny_metadata = false;
- if (path->slots[0]) {
- path->slots[0]--;
- btrfs_item_key_to_cpu(path->nodes[0], &key,
- path->slots[0]);
- if (key.objectid == bytenr &&
- key.type == BTRFS_EXTENT_ITEM_KEY &&
- key.offset == num_bytes)
- ret = 0;
- }
- if (ret) {
- key.objectid = bytenr;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- key.offset = num_bytes;
- btrfs_release_path(path);
- goto again;
- }
- }
-
- if (ret && !insert) {
- err = -ENOENT;
- goto out;
- } else if (WARN_ON(ret)) {
- err = -EIO;
- goto out;
- }
-
- leaf = path->nodes[0];
- item_size = btrfs_item_size_nr(leaf, path->slots[0]);
- if (unlikely(item_size < sizeof(*ei))) {
- err = -EINVAL;
- btrfs_print_v0_err(fs_info);
- btrfs_abort_transaction(trans, err);
- goto out;
- }
-
- ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- flags = btrfs_extent_flags(leaf, ei);
-
- ptr = (unsigned long)(ei + 1);
- end = (unsigned long)ei + item_size;
-
- if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
- ptr += sizeof(struct btrfs_tree_block_info);
- BUG_ON(ptr > end);
- }
-
- if (owner >= BTRFS_FIRST_FREE_OBJECTID)
- needed = BTRFS_REF_TYPE_DATA;
- else
- needed = BTRFS_REF_TYPE_BLOCK;
-
- err = -ENOENT;
- while (1) {
- if (ptr >= end) {
- WARN_ON(ptr > end);
- break;
- }
- iref = (struct btrfs_extent_inline_ref *)ptr;
- type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
- if (type == BTRFS_REF_TYPE_INVALID) {
- err = -EUCLEAN;
- goto out;
- }
-
- if (want < type)
- break;
- if (want > type) {
- ptr += btrfs_extent_inline_ref_size(type);
- continue;
- }
-
- if (type == BTRFS_EXTENT_DATA_REF_KEY) {
- struct btrfs_extent_data_ref *dref;
- dref = (struct btrfs_extent_data_ref *)(&iref->offset);
- if (match_extent_data_ref(leaf, dref, root_objectid,
- owner, offset)) {
- err = 0;
- break;
- }
- if (hash_extent_data_ref_item(leaf, dref) <
- hash_extent_data_ref(root_objectid, owner, offset))
- break;
- } else {
- u64 ref_offset;
- ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
- if (parent > 0) {
- if (parent == ref_offset) {
- err = 0;
- break;
- }
- if (ref_offset < parent)
- break;
- } else {
- if (root_objectid == ref_offset) {
- err = 0;
- break;
- }
- if (ref_offset < root_objectid)
- break;
- }
- }
- ptr += btrfs_extent_inline_ref_size(type);
- }
- if (err == -ENOENT && insert) {
- if (item_size + extra_size >=
- BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
- err = -EAGAIN;
- goto out;
- }
- /*
- * To add new inline back ref, we have to make sure
- * there is no corresponding back ref item.
- * For simplicity, we just do not add new inline back
- * ref if there is any kind of item for this block
- */
- if (find_next_key(path, 0, &key) == 0 &&
- key.objectid == bytenr &&
- key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
- err = -EAGAIN;
- goto out;
- }
- }
- *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
-out:
- if (insert) {
- path->keep_locks = 0;
- btrfs_unlock_up_safe(path, 1);
- }
- return err;
-}
-
-/*
- * helper to add new inline back ref
- */
-static noinline_for_stack
-void setup_inline_extent_backref(struct btrfs_fs_info *fs_info,
- struct btrfs_path *path,
- struct btrfs_extent_inline_ref *iref,
- u64 parent, u64 root_objectid,
- u64 owner, u64 offset, int refs_to_add,
- struct btrfs_delayed_extent_op *extent_op)
-{
- struct extent_buffer *leaf;
- struct btrfs_extent_item *ei;
- unsigned long ptr;
- unsigned long end;
- unsigned long item_offset;
- u64 refs;
- int size;
- int type;
-
- leaf = path->nodes[0];
- ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- item_offset = (unsigned long)iref - (unsigned long)ei;
-
- type = extent_ref_type(parent, owner);
- size = btrfs_extent_inline_ref_size(type);
-
- btrfs_extend_item(path, size);
-
- ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- refs = btrfs_extent_refs(leaf, ei);
- refs += refs_to_add;
- btrfs_set_extent_refs(leaf, ei, refs);
- if (extent_op)
- __run_delayed_extent_op(extent_op, leaf, ei);
-
- ptr = (unsigned long)ei + item_offset;
- end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
- if (ptr < end - size)
- memmove_extent_buffer(leaf, ptr + size, ptr,
- end - size - ptr);
-
- iref = (struct btrfs_extent_inline_ref *)ptr;
- btrfs_set_extent_inline_ref_type(leaf, iref, type);
- if (type == BTRFS_EXTENT_DATA_REF_KEY) {
- struct btrfs_extent_data_ref *dref;
- dref = (struct btrfs_extent_data_ref *)(&iref->offset);
- btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
- btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
- btrfs_set_extent_data_ref_offset(leaf, dref, offset);
- btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
- } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
- struct btrfs_shared_data_ref *sref;
- sref = (struct btrfs_shared_data_ref *)(iref + 1);
- btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
- btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
- } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
- btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
- } else {
- btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
- }
- btrfs_mark_buffer_dirty(leaf);
-}
-
-static int lookup_extent_backref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- struct btrfs_extent_inline_ref **ref_ret,
- u64 bytenr, u64 num_bytes, u64 parent,
- u64 root_objectid, u64 owner, u64 offset)
-{
- int ret;
-
- ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr,
- num_bytes, parent, root_objectid,
- owner, offset, 0);
- if (ret != -ENOENT)
- return ret;
-
- btrfs_release_path(path);
- *ref_ret = NULL;
-
- if (owner < BTRFS_FIRST_FREE_OBJECTID) {
- ret = lookup_tree_block_ref(trans, path, bytenr, parent,
- root_objectid);
- } else {
- ret = lookup_extent_data_ref(trans, path, bytenr, parent,
- root_objectid, owner, offset);
- }
- return ret;
-}
-
-/*
- * helper to update/remove inline back ref
- */
-static noinline_for_stack
-void update_inline_extent_backref(struct btrfs_path *path,
- struct btrfs_extent_inline_ref *iref,
- int refs_to_mod,
- struct btrfs_delayed_extent_op *extent_op,
- int *last_ref)
-{
- struct extent_buffer *leaf = path->nodes[0];
- struct btrfs_extent_item *ei;
- struct btrfs_extent_data_ref *dref = NULL;
- struct btrfs_shared_data_ref *sref = NULL;
- unsigned long ptr;
- unsigned long end;
- u32 item_size;
- int size;
- int type;
- u64 refs;
-
- ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- refs = btrfs_extent_refs(leaf, ei);
- WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
- refs += refs_to_mod;
- btrfs_set_extent_refs(leaf, ei, refs);
- if (extent_op)
- __run_delayed_extent_op(extent_op, leaf, ei);
-
- /*
- * If type is invalid, we should have bailed out after
- * lookup_inline_extent_backref().
- */
- type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
- ASSERT(type != BTRFS_REF_TYPE_INVALID);
-
- if (type == BTRFS_EXTENT_DATA_REF_KEY) {
- dref = (struct btrfs_extent_data_ref *)(&iref->offset);
- refs = btrfs_extent_data_ref_count(leaf, dref);
- } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
- sref = (struct btrfs_shared_data_ref *)(iref + 1);
- refs = btrfs_shared_data_ref_count(leaf, sref);
- } else {
- refs = 1;
- BUG_ON(refs_to_mod != -1);
- }
-
- BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
- refs += refs_to_mod;
-
- if (refs > 0) {
- if (type == BTRFS_EXTENT_DATA_REF_KEY)
- btrfs_set_extent_data_ref_count(leaf, dref, refs);
- else
- btrfs_set_shared_data_ref_count(leaf, sref, refs);
- } else {
- *last_ref = 1;
- size = btrfs_extent_inline_ref_size(type);
- item_size = btrfs_item_size_nr(leaf, path->slots[0]);
- ptr = (unsigned long)iref;
- end = (unsigned long)ei + item_size;
- if (ptr + size < end)
- memmove_extent_buffer(leaf, ptr, ptr + size,
- end - ptr - size);
- item_size -= size;
- btrfs_truncate_item(path, item_size, 1);
- }
- btrfs_mark_buffer_dirty(leaf);
-}
-
-static noinline_for_stack
-int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- u64 bytenr, u64 num_bytes, u64 parent,
- u64 root_objectid, u64 owner,
- u64 offset, int refs_to_add,
- struct btrfs_delayed_extent_op *extent_op)
-{
- struct btrfs_extent_inline_ref *iref;
- int ret;
-
- ret = lookup_inline_extent_backref(trans, path, &iref, bytenr,
- num_bytes, parent, root_objectid,
- owner, offset, 1);
- if (ret == 0) {
- BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
- update_inline_extent_backref(path, iref, refs_to_add,
- extent_op, NULL);
- } else if (ret == -ENOENT) {
- setup_inline_extent_backref(trans->fs_info, path, iref, parent,
- root_objectid, owner, offset,
- refs_to_add, extent_op);
- ret = 0;
- }
- return ret;
-}
-
-static int insert_extent_backref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- u64 bytenr, u64 parent, u64 root_objectid,
- u64 owner, u64 offset, int refs_to_add)
-{
- int ret;
- if (owner < BTRFS_FIRST_FREE_OBJECTID) {
- BUG_ON(refs_to_add != 1);
- ret = insert_tree_block_ref(trans, path, bytenr, parent,
- root_objectid);
- } else {
- ret = insert_extent_data_ref(trans, path, bytenr, parent,
- root_objectid, owner, offset,
- refs_to_add);
- }
- return ret;
-}
-
-static int remove_extent_backref(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- struct btrfs_extent_inline_ref *iref,
- int refs_to_drop, int is_data, int *last_ref)
-{
- int ret = 0;
-
- BUG_ON(!is_data && refs_to_drop != 1);
- if (iref) {
- update_inline_extent_backref(path, iref, -refs_to_drop, NULL,
- last_ref);
- } else if (is_data) {
- ret = remove_extent_data_ref(trans, path, refs_to_drop,
- last_ref);
- } else {
- *last_ref = 1;
- ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
- }
- return ret;
-}
-
-static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
- u64 *discarded_bytes)
-{
- int j, ret = 0;
- u64 bytes_left, end;
- u64 aligned_start = ALIGN(start, 1 << 9);
-
- if (WARN_ON(start != aligned_start)) {
- len -= aligned_start - start;
- len = round_down(len, 1 << 9);
- start = aligned_start;
- }
-
- *discarded_bytes = 0;
-
- if (!len)
- return 0;
-
- end = start + len;
- bytes_left = len;
-
- /* Skip any superblocks on this device. */
- for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
- u64 sb_start = btrfs_sb_offset(j);
- u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
- u64 size = sb_start - start;
-
- if (!in_range(sb_start, start, bytes_left) &&
- !in_range(sb_end, start, bytes_left) &&
- !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
- continue;
-
- /*
- * Superblock spans beginning of range. Adjust start and
- * try again.
- */
- if (sb_start <= start) {
- start += sb_end - start;
- if (start > end) {
- bytes_left = 0;
- break;
- }
- bytes_left = end - start;
- continue;
- }
-
- if (size) {
- ret = blkdev_issue_discard(bdev, start >> 9, size >> 9,
- GFP_NOFS, 0);
- if (!ret)
- *discarded_bytes += size;
- else if (ret != -EOPNOTSUPP)
- return ret;
- }
-
- start = sb_end;
- if (start > end) {
- bytes_left = 0;
- break;
- }
- bytes_left = end - start;
- }
-
- if (bytes_left) {
- ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9,
- GFP_NOFS, 0);
- if (!ret)
- *discarded_bytes += bytes_left;
- }
- return ret;
-}
-
-int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
- u64 num_bytes, u64 *actual_bytes)
-{
- int ret;
- u64 discarded_bytes = 0;
- struct btrfs_bio *bbio = NULL;
-
-
- /*
- * Avoid races with device replace and make sure our bbio has devices
- * associated to its stripes that don't go away while we are discarding.
- */
- btrfs_bio_counter_inc_blocked(fs_info);
- /* Tell the block device(s) that the sectors can be discarded */
- ret = btrfs_map_block(fs_info, BTRFS_MAP_DISCARD, bytenr, &num_bytes,
- &bbio, 0);
- /* Error condition is -ENOMEM */
- if (!ret) {
- struct btrfs_bio_stripe *stripe = bbio->stripes;
- int i;
-
-
- for (i = 0; i < bbio->num_stripes; i++, stripe++) {
- u64 bytes;
- struct request_queue *req_q;
-
- if (!stripe->dev->bdev) {
- ASSERT(btrfs_test_opt(fs_info, DEGRADED));
- continue;
- }
- req_q = bdev_get_queue(stripe->dev->bdev);
- if (!blk_queue_discard(req_q))
- continue;
-
- ret = btrfs_issue_discard(stripe->dev->bdev,
- stripe->physical,
- stripe->length,
- &bytes);
- if (!ret)
- discarded_bytes += bytes;
- else if (ret != -EOPNOTSUPP)
- break; /* Logic errors or -ENOMEM, or -EIO but I don't know how that could happen JDM */
-
- /*
- * Just in case we get back EOPNOTSUPP for some reason,
- * just ignore the return value so we don't screw up
- * people calling discard_extent.
- */
- ret = 0;
- }
- btrfs_put_bbio(bbio);
- }
- btrfs_bio_counter_dec(fs_info);
-
- if (actual_bytes)
- *actual_bytes = discarded_bytes;
-
-
- if (ret == -EOPNOTSUPP)
- ret = 0;
- return ret;
-}
-
-/* Can return -ENOMEM */
-int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
- struct btrfs_ref *generic_ref)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- int old_ref_mod, new_ref_mod;
- int ret;
-
- ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
- generic_ref->action);
- BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
- generic_ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID);
-
- if (generic_ref->type == BTRFS_REF_METADATA)
- ret = btrfs_add_delayed_tree_ref(trans, generic_ref,
- NULL, &old_ref_mod, &new_ref_mod);
- else
- ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0,
- &old_ref_mod, &new_ref_mod);
-
- btrfs_ref_tree_mod(fs_info, generic_ref);
-
- if (ret == 0 && old_ref_mod < 0 && new_ref_mod >= 0)
- sub_pinned_bytes(fs_info, generic_ref);
-
- return ret;
-}
-
-/*
- * __btrfs_inc_extent_ref - insert backreference for a given extent
- *
- * @trans: Handle of transaction
- *
- * @node: The delayed ref node used to get the bytenr/length for
- * extent whose references are incremented.
- *
- * @parent: If this is a shared extent (BTRFS_SHARED_DATA_REF_KEY/
- * BTRFS_SHARED_BLOCK_REF_KEY) then it holds the logical
- * bytenr of the parent block. Since new extents are always
- * created with indirect references, this will only be the case
- * when relocating a shared extent. In that case, root_objectid
- * will be BTRFS_TREE_RELOC_OBJECTID. Otheriwse, parent must
- * be 0
- *
- * @root_objectid: The id of the root where this modification has originated,
- * this can be either one of the well-known metadata trees or
- * the subvolume id which references this extent.
- *
- * @owner: For data extents it is the inode number of the owning file.
- * For metadata extents this parameter holds the level in the
- * tree of the extent.
- *
- * @offset: For metadata extents the offset is ignored and is currently
- * always passed as 0. For data extents it is the fileoffset
- * this extent belongs to.
- *
- * @refs_to_add Number of references to add
- *
- * @extent_op Pointer to a structure, holding information necessary when
- * updating a tree block's flags
- *
- */
-static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_node *node,
- u64 parent, u64 root_objectid,
- u64 owner, u64 offset, int refs_to_add,
- struct btrfs_delayed_extent_op *extent_op)
-{
- struct btrfs_path *path;
- struct extent_buffer *leaf;
- struct btrfs_extent_item *item;
- struct btrfs_key key;
- u64 bytenr = node->bytenr;
- u64 num_bytes = node->num_bytes;
- u64 refs;
- int ret;
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
-
- path->reada = READA_FORWARD;
- path->leave_spinning = 1;
- /* this will setup the path even if it fails to insert the back ref */
- ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes,
- parent, root_objectid, owner,
- offset, refs_to_add, extent_op);
- if ((ret < 0 && ret != -EAGAIN) || !ret)
- goto out;
-
- /*
- * Ok we had -EAGAIN which means we didn't have space to insert and
- * inline extent ref, so just update the reference count and add a
- * normal backref.
- */
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- refs = btrfs_extent_refs(leaf, item);
- btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
- if (extent_op)
- __run_delayed_extent_op(extent_op, leaf, item);
-
- btrfs_mark_buffer_dirty(leaf);
- btrfs_release_path(path);
-
- path->reada = READA_FORWARD;
- path->leave_spinning = 1;
- /* now insert the actual backref */
- ret = insert_extent_backref(trans, path, bytenr, parent, root_objectid,
- owner, offset, refs_to_add);
- if (ret)
- btrfs_abort_transaction(trans, ret);
-out:
- btrfs_free_path(path);
- return ret;
-}
-
-static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_node *node,
- struct btrfs_delayed_extent_op *extent_op,
- int insert_reserved)
-{
- int ret = 0;
- struct btrfs_delayed_data_ref *ref;
- struct btrfs_key ins;
- u64 parent = 0;
- u64 ref_root = 0;
- u64 flags = 0;
-
- ins.objectid = node->bytenr;
- ins.offset = node->num_bytes;
- ins.type = BTRFS_EXTENT_ITEM_KEY;
-
- ref = btrfs_delayed_node_to_data_ref(node);
- trace_run_delayed_data_ref(trans->fs_info, node, ref, node->action);
-
- if (node->type == BTRFS_SHARED_DATA_REF_KEY)
- parent = ref->parent;
- ref_root = ref->root;
-
- if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
- if (extent_op)
- flags |= extent_op->flags_to_set;
- ret = alloc_reserved_file_extent(trans, parent, ref_root,
- flags, ref->objectid,
- ref->offset, &ins,
- node->ref_mod);
- } else if (node->action == BTRFS_ADD_DELAYED_REF) {
- ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
- ref->objectid, ref->offset,
- node->ref_mod, extent_op);
- } else if (node->action == BTRFS_DROP_DELAYED_REF) {
- ret = __btrfs_free_extent(trans, node, parent,
- ref_root, ref->objectid,
- ref->offset, node->ref_mod,
- extent_op);
- } else {
- BUG();
- }
- return ret;
-}
-
-static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
- struct extent_buffer *leaf,
- struct btrfs_extent_item *ei)
-{
- u64 flags = btrfs_extent_flags(leaf, ei);
- if (extent_op->update_flags) {
- flags |= extent_op->flags_to_set;
- btrfs_set_extent_flags(leaf, ei, flags);
- }
-
- if (extent_op->update_key) {
- struct btrfs_tree_block_info *bi;
- BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
- bi = (struct btrfs_tree_block_info *)(ei + 1);
- btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
- }
-}
-
-static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_head *head,
- struct btrfs_delayed_extent_op *extent_op)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_key key;
- struct btrfs_path *path;
- struct btrfs_extent_item *ei;
- struct extent_buffer *leaf;
- u32 item_size;
- int ret;
- int err = 0;
- int metadata = !extent_op->is_data;
-
- if (trans->aborted)
- return 0;
-
- if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA))
- metadata = 0;
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
-
- key.objectid = head->bytenr;
-
- if (metadata) {
- key.type = BTRFS_METADATA_ITEM_KEY;
- key.offset = extent_op->level;
- } else {
- key.type = BTRFS_EXTENT_ITEM_KEY;
- key.offset = head->num_bytes;
- }
-
-again:
- path->reada = READA_FORWARD;
- path->leave_spinning = 1;
- ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 1);
- if (ret < 0) {
- err = ret;
- goto out;
- }
- if (ret > 0) {
- if (metadata) {
- if (path->slots[0] > 0) {
- path->slots[0]--;
- btrfs_item_key_to_cpu(path->nodes[0], &key,
- path->slots[0]);
- if (key.objectid == head->bytenr &&
- key.type == BTRFS_EXTENT_ITEM_KEY &&
- key.offset == head->num_bytes)
- ret = 0;
- }
- if (ret > 0) {
- btrfs_release_path(path);
- metadata = 0;
-
- key.objectid = head->bytenr;
- key.offset = head->num_bytes;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- goto again;
- }
- } else {
- err = -EIO;
- goto out;
- }
- }
-
- leaf = path->nodes[0];
- item_size = btrfs_item_size_nr(leaf, path->slots[0]);
-
- if (unlikely(item_size < sizeof(*ei))) {
- err = -EINVAL;
- btrfs_print_v0_err(fs_info);
- btrfs_abort_transaction(trans, err);
- goto out;
- }
-
- ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- __run_delayed_extent_op(extent_op, leaf, ei);
-
- btrfs_mark_buffer_dirty(leaf);
-out:
- btrfs_free_path(path);
- return err;
-}
-
-static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_node *node,
- struct btrfs_delayed_extent_op *extent_op,
- int insert_reserved)
-{
- int ret = 0;
- struct btrfs_delayed_tree_ref *ref;
- u64 parent = 0;
- u64 ref_root = 0;
-
- ref = btrfs_delayed_node_to_tree_ref(node);
- trace_run_delayed_tree_ref(trans->fs_info, node, ref, node->action);
-
- if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
- parent = ref->parent;
- ref_root = ref->root;
-
- if (node->ref_mod != 1) {
- btrfs_err(trans->fs_info,
- "btree block(%llu) has %d references rather than 1: action %d ref_root %llu parent %llu",
- node->bytenr, node->ref_mod, node->action, ref_root,
- parent);
- return -EIO;
- }
- if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
- BUG_ON(!extent_op || !extent_op->update_flags);
- ret = alloc_reserved_tree_block(trans, node, extent_op);
- } else if (node->action == BTRFS_ADD_DELAYED_REF) {
- ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
- ref->level, 0, 1, extent_op);
- } else if (node->action == BTRFS_DROP_DELAYED_REF) {
- ret = __btrfs_free_extent(trans, node, parent, ref_root,
- ref->level, 0, 1, extent_op);
- } else {
- BUG();
- }
- return ret;
-}
-
-/* helper function to actually process a single delayed ref entry */
-static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_node *node,
- struct btrfs_delayed_extent_op *extent_op,
- int insert_reserved)
-{
- int ret = 0;
-
- if (trans->aborted) {
- if (insert_reserved)
- btrfs_pin_extent(trans->fs_info, node->bytenr,
- node->num_bytes, 1);
- return 0;
- }
-
- if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
- node->type == BTRFS_SHARED_BLOCK_REF_KEY)
- ret = run_delayed_tree_ref(trans, node, extent_op,
- insert_reserved);
- else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
- node->type == BTRFS_SHARED_DATA_REF_KEY)
- ret = run_delayed_data_ref(trans, node, extent_op,
- insert_reserved);
- else
- BUG();
- if (ret && insert_reserved)
- btrfs_pin_extent(trans->fs_info, node->bytenr,
- node->num_bytes, 1);
- return ret;
-}
-
-static inline struct btrfs_delayed_ref_node *
-select_delayed_ref(struct btrfs_delayed_ref_head *head)
-{
- struct btrfs_delayed_ref_node *ref;
-
- if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
- return NULL;
-
- /*
- * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
- * This is to prevent a ref count from going down to zero, which deletes
- * the extent item from the extent tree, when there still are references
- * to add, which would fail because they would not find the extent item.
- */
- if (!list_empty(&head->ref_add_list))
- return list_first_entry(&head->ref_add_list,
- struct btrfs_delayed_ref_node, add_list);
-
- ref = rb_entry(rb_first_cached(&head->ref_tree),
- struct btrfs_delayed_ref_node, ref_node);
- ASSERT(list_empty(&ref->add_list));
- return ref;
-}
-
-static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
- struct btrfs_delayed_ref_head *head)
-{
- spin_lock(&delayed_refs->lock);
- head->processing = 0;
- delayed_refs->num_heads_ready++;
- spin_unlock(&delayed_refs->lock);
- btrfs_delayed_ref_unlock(head);
-}
-
-static struct btrfs_delayed_extent_op *cleanup_extent_op(
- struct btrfs_delayed_ref_head *head)
-{
- struct btrfs_delayed_extent_op *extent_op = head->extent_op;
-
- if (!extent_op)
- return NULL;
-
- if (head->must_insert_reserved) {
- head->extent_op = NULL;
- btrfs_free_delayed_extent_op(extent_op);
- return NULL;
- }
- return extent_op;
-}
-
-static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_head *head)
-{
- struct btrfs_delayed_extent_op *extent_op;
- int ret;
-
- extent_op = cleanup_extent_op(head);
- if (!extent_op)
- return 0;
- head->extent_op = NULL;
- spin_unlock(&head->lock);
- ret = run_delayed_extent_op(trans, head, extent_op);
- btrfs_free_delayed_extent_op(extent_op);
- return ret ? ret : 1;
-}
-
-void btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info,
- struct btrfs_delayed_ref_root *delayed_refs,
- struct btrfs_delayed_ref_head *head)
-{
- int nr_items = 1; /* Dropping this ref head update. */
-
- if (head->total_ref_mod < 0) {
- struct btrfs_space_info *space_info;
- u64 flags;
-
- if (head->is_data)
- flags = BTRFS_BLOCK_GROUP_DATA;
- else if (head->is_system)
- flags = BTRFS_BLOCK_GROUP_SYSTEM;
- else
- flags = BTRFS_BLOCK_GROUP_METADATA;
- space_info = btrfs_find_space_info(fs_info, flags);
- ASSERT(space_info);
- percpu_counter_add_batch(&space_info->total_bytes_pinned,
- -head->num_bytes,
- BTRFS_TOTAL_BYTES_PINNED_BATCH);
-
- /*
- * We had csum deletions accounted for in our delayed refs rsv,
- * we need to drop the csum leaves for this update from our
- * delayed_refs_rsv.
- */
- if (head->is_data) {
- spin_lock(&delayed_refs->lock);
- delayed_refs->pending_csums -= head->num_bytes;
- spin_unlock(&delayed_refs->lock);
- nr_items += btrfs_csum_bytes_to_leaves(fs_info,
- head->num_bytes);
- }
- }
-
- btrfs_delayed_refs_rsv_release(fs_info, nr_items);
-}
-
-static int cleanup_ref_head(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_head *head)
-{
-
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_delayed_ref_root *delayed_refs;
- int ret;
-
- delayed_refs = &trans->transaction->delayed_refs;
-
- ret = run_and_cleanup_extent_op(trans, head);
- if (ret < 0) {
- unselect_delayed_ref_head(delayed_refs, head);
- btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
- return ret;
- } else if (ret) {
- return ret;
- }
-
- /*
- * Need to drop our head ref lock and re-acquire the delayed ref lock
- * and then re-check to make sure nobody got added.
- */
- spin_unlock(&head->lock);
- spin_lock(&delayed_refs->lock);
- spin_lock(&head->lock);
- if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) {
- spin_unlock(&head->lock);
- spin_unlock(&delayed_refs->lock);
- return 1;
- }
- btrfs_delete_ref_head(delayed_refs, head);
- spin_unlock(&head->lock);
- spin_unlock(&delayed_refs->lock);
-
- if (head->must_insert_reserved) {
- btrfs_pin_extent(fs_info, head->bytenr,
- head->num_bytes, 1);
- if (head->is_data) {
- ret = btrfs_del_csums(trans, fs_info, head->bytenr,
- head->num_bytes);
- }
- }
-
- btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
-
- trace_run_delayed_ref_head(fs_info, head, 0);
- btrfs_delayed_ref_unlock(head);
- btrfs_put_delayed_ref_head(head);
- return 0;
-}
-
-static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head(
- struct btrfs_trans_handle *trans)
-{
- struct btrfs_delayed_ref_root *delayed_refs =
- &trans->transaction->delayed_refs;
- struct btrfs_delayed_ref_head *head = NULL;
- int ret;
-
- spin_lock(&delayed_refs->lock);
- head = btrfs_select_ref_head(delayed_refs);
- if (!head) {
- spin_unlock(&delayed_refs->lock);
- return head;
- }
-
- /*
- * Grab the lock that says we are going to process all the refs for
- * this head
- */
- ret = btrfs_delayed_ref_lock(delayed_refs, head);
- spin_unlock(&delayed_refs->lock);
-
- /*
- * We may have dropped the spin lock to get the head mutex lock, and
- * that might have given someone else time to free the head. If that's
- * true, it has been removed from our list and we can move on.
- */
- if (ret == -EAGAIN)
- head = ERR_PTR(-EAGAIN);
-
- return head;
-}
-
-static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_head *locked_ref,
- unsigned long *run_refs)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_delayed_ref_root *delayed_refs;
- struct btrfs_delayed_extent_op *extent_op;
- struct btrfs_delayed_ref_node *ref;
- int must_insert_reserved = 0;
- int ret;
-
- delayed_refs = &trans->transaction->delayed_refs;
-
- lockdep_assert_held(&locked_ref->mutex);
- lockdep_assert_held(&locked_ref->lock);
-
- while ((ref = select_delayed_ref(locked_ref))) {
- if (ref->seq &&
- btrfs_check_delayed_seq(fs_info, ref->seq)) {
- spin_unlock(&locked_ref->lock);
- unselect_delayed_ref_head(delayed_refs, locked_ref);
- return -EAGAIN;
- }
-
- (*run_refs)++;
- ref->in_tree = 0;
- rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree);
- RB_CLEAR_NODE(&ref->ref_node);
- if (!list_empty(&ref->add_list))
- list_del(&ref->add_list);
- /*
- * When we play the delayed ref, also correct the ref_mod on
- * head
- */
- switch (ref->action) {
- case BTRFS_ADD_DELAYED_REF:
- case BTRFS_ADD_DELAYED_EXTENT:
- locked_ref->ref_mod -= ref->ref_mod;
- break;
- case BTRFS_DROP_DELAYED_REF:
- locked_ref->ref_mod += ref->ref_mod;
- break;
- default:
- WARN_ON(1);
- }
- atomic_dec(&delayed_refs->num_entries);
-
- /*
- * Record the must_insert_reserved flag before we drop the
- * spin lock.
- */
- must_insert_reserved = locked_ref->must_insert_reserved;
- locked_ref->must_insert_reserved = 0;
-
- extent_op = locked_ref->extent_op;
- locked_ref->extent_op = NULL;
- spin_unlock(&locked_ref->lock);
-
- ret = run_one_delayed_ref(trans, ref, extent_op,
- must_insert_reserved);
-
- btrfs_free_delayed_extent_op(extent_op);
- if (ret) {
- unselect_delayed_ref_head(delayed_refs, locked_ref);
- btrfs_put_delayed_ref(ref);
- btrfs_debug(fs_info, "run_one_delayed_ref returned %d",
- ret);
- return ret;
- }
-
- btrfs_put_delayed_ref(ref);
- cond_resched();
-
- spin_lock(&locked_ref->lock);
- btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
- }
-
- return 0;
-}
-
-/*
- * Returns 0 on success or if called with an already aborted transaction.
- * Returns -ENOMEM or -EIO on failure and will abort the transaction.
- */
-static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
- unsigned long nr)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_delayed_ref_root *delayed_refs;
- struct btrfs_delayed_ref_head *locked_ref = NULL;
- ktime_t start = ktime_get();
- int ret;
- unsigned long count = 0;
- unsigned long actual_count = 0;
-
- delayed_refs = &trans->transaction->delayed_refs;
- do {
- if (!locked_ref) {
- locked_ref = btrfs_obtain_ref_head(trans);
- if (IS_ERR_OR_NULL(locked_ref)) {
- if (PTR_ERR(locked_ref) == -EAGAIN) {
- continue;
- } else {
- break;
- }
- }
- count++;
- }
- /*
- * We need to try and merge add/drops of the same ref since we
- * can run into issues with relocate dropping the implicit ref
- * and then it being added back again before the drop can
- * finish. If we merged anything we need to re-loop so we can
- * get a good ref.
- * Or we can get node references of the same type that weren't
- * merged when created due to bumps in the tree mod seq, and
- * we need to merge them to prevent adding an inline extent
- * backref before dropping it (triggering a BUG_ON at
- * insert_inline_extent_backref()).
- */
- spin_lock(&locked_ref->lock);
- btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
-
- ret = btrfs_run_delayed_refs_for_head(trans, locked_ref,
- &actual_count);
- if (ret < 0 && ret != -EAGAIN) {
- /*
- * Error, btrfs_run_delayed_refs_for_head already
- * unlocked everything so just bail out
- */
- return ret;
- } else if (!ret) {
- /*
- * Success, perform the usual cleanup of a processed
- * head
- */
- ret = cleanup_ref_head(trans, locked_ref);
- if (ret > 0 ) {
- /* We dropped our lock, we need to loop. */
- ret = 0;
- continue;
- } else if (ret) {
- return ret;
- }
- }
-
- /*
- * Either success case or btrfs_run_delayed_refs_for_head
- * returned -EAGAIN, meaning we need to select another head
- */
-
- locked_ref = NULL;
- cond_resched();
- } while ((nr != -1 && count < nr) || locked_ref);
-
- /*
- * We don't want to include ref heads since we can have empty ref heads
- * and those will drastically skew our runtime down since we just do
- * accounting, no actual extent tree updates.
- */
- if (actual_count > 0) {
- u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
- u64 avg;
-
- /*
- * We weigh the current average higher than our current runtime
- * to avoid large swings in the average.
- */
- spin_lock(&delayed_refs->lock);
- avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
- fs_info->avg_delayed_ref_runtime = avg >> 2; /* div by 4 */
- spin_unlock(&delayed_refs->lock);
- }
- return 0;
-}
-
-#ifdef SCRAMBLE_DELAYED_REFS
-/*
- * Normally delayed refs get processed in ascending bytenr order. This
- * correlates in most cases to the order added. To expose dependencies on this
- * order, we start to process the tree in the middle instead of the beginning
- */
-static u64 find_middle(struct rb_root *root)
-{
- struct rb_node *n = root->rb_node;
- struct btrfs_delayed_ref_node *entry;
- int alt = 1;
- u64 middle;
- u64 first = 0, last = 0;
-
- n = rb_first(root);
- if (n) {
- entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
- first = entry->bytenr;
- }
- n = rb_last(root);
- if (n) {
- entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
- last = entry->bytenr;
- }
- n = root->rb_node;
-
- while (n) {
- entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
- WARN_ON(!entry->in_tree);
-
- middle = entry->bytenr;
-
- if (alt)
- n = n->rb_left;
- else
- n = n->rb_right;
-
- alt = 1 - alt;
- }
- return middle;
-}
-#endif
-
-static inline u64 heads_to_leaves(struct btrfs_fs_info *fs_info, u64 heads)
-{
- u64 num_bytes;
-
- num_bytes = heads * (sizeof(struct btrfs_extent_item) +
- sizeof(struct btrfs_extent_inline_ref));
- if (!btrfs_fs_incompat(fs_info, SKINNY_METADATA))
- num_bytes += heads * sizeof(struct btrfs_tree_block_info);
-
- /*
- * We don't ever fill up leaves all the way so multiply by 2 just to be
- * closer to what we're really going to want to use.
- */
- return div_u64(num_bytes, BTRFS_LEAF_DATA_SIZE(fs_info));
-}
-
-/*
- * Takes the number of bytes to be csumm'ed and figures out how many leaves it
- * would require to store the csums for that many bytes.
- */
-u64 btrfs_csum_bytes_to_leaves(struct btrfs_fs_info *fs_info, u64 csum_bytes)
-{
- u64 csum_size;
- u64 num_csums_per_leaf;
- u64 num_csums;
-
- csum_size = BTRFS_MAX_ITEM_SIZE(fs_info);
- num_csums_per_leaf = div64_u64(csum_size,
- (u64)btrfs_super_csum_size(fs_info->super_copy));
- num_csums = div64_u64(csum_bytes, fs_info->sectorsize);
- num_csums += num_csums_per_leaf - 1;
- num_csums = div64_u64(num_csums, num_csums_per_leaf);
- return num_csums;
-}
-
-/*
- * this starts processing the delayed reference count updates and
- * extent insertions we have queued up so far. count can be
- * 0, which means to process everything in the tree at the start
- * of the run (but not newly added entries), or it can be some target
- * number you'd like to process.
- *
- * Returns 0 on success or if called with an aborted transaction
- * Returns <0 on error and aborts the transaction
- */
-int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
- unsigned long count)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct rb_node *node;
- struct btrfs_delayed_ref_root *delayed_refs;
- struct btrfs_delayed_ref_head *head;
- int ret;
- int run_all = count == (unsigned long)-1;
-
- /* We'll clean this up in btrfs_cleanup_transaction */
- if (trans->aborted)
- return 0;
-
- if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
- return 0;
-
- delayed_refs = &trans->transaction->delayed_refs;
- if (count == 0)
- count = atomic_read(&delayed_refs->num_entries) * 2;
-
-again:
-#ifdef SCRAMBLE_DELAYED_REFS
- delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
-#endif
- ret = __btrfs_run_delayed_refs(trans, count);
- if (ret < 0) {
- btrfs_abort_transaction(trans, ret);
- return ret;
- }
-
- if (run_all) {
- btrfs_create_pending_block_groups(trans);
-
- spin_lock(&delayed_refs->lock);
- node = rb_first_cached(&delayed_refs->href_root);
- if (!node) {
- spin_unlock(&delayed_refs->lock);
- goto out;
- }
- head = rb_entry(node, struct btrfs_delayed_ref_head,
- href_node);
- refcount_inc(&head->refs);
- spin_unlock(&delayed_refs->lock);
-
- /* Mutex was contended, block until it's released and retry. */
- mutex_lock(&head->mutex);
- mutex_unlock(&head->mutex);
-
- btrfs_put_delayed_ref_head(head);
- cond_resched();
- goto again;
- }
-out:
- return 0;
-}
-
-int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
- u64 bytenr, u64 num_bytes, u64 flags,
- int level, int is_data)
-{
- struct btrfs_delayed_extent_op *extent_op;
- int ret;
-
- extent_op = btrfs_alloc_delayed_extent_op();
- if (!extent_op)
- return -ENOMEM;
-
- extent_op->flags_to_set = flags;
- extent_op->update_flags = true;
- extent_op->update_key = false;
- extent_op->is_data = is_data ? true : false;
- extent_op->level = level;
-
- ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
- if (ret)
- btrfs_free_delayed_extent_op(extent_op);
- return ret;
-}
-
-static noinline int check_delayed_ref(struct btrfs_root *root,
- struct btrfs_path *path,
- u64 objectid, u64 offset, u64 bytenr)
-{
- struct btrfs_delayed_ref_head *head;
- struct btrfs_delayed_ref_node *ref;
- struct btrfs_delayed_data_ref *data_ref;
- struct btrfs_delayed_ref_root *delayed_refs;
- struct btrfs_transaction *cur_trans;
- struct rb_node *node;
- int ret = 0;
-
- spin_lock(&root->fs_info->trans_lock);
- cur_trans = root->fs_info->running_transaction;
- if (cur_trans)
- refcount_inc(&cur_trans->use_count);
- spin_unlock(&root->fs_info->trans_lock);
- if (!cur_trans)
- return 0;
-
- delayed_refs = &cur_trans->delayed_refs;
- spin_lock(&delayed_refs->lock);
- head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
- if (!head) {
- spin_unlock(&delayed_refs->lock);
- btrfs_put_transaction(cur_trans);
- return 0;
- }
-
- if (!mutex_trylock(&head->mutex)) {
- refcount_inc(&head->refs);
- spin_unlock(&delayed_refs->lock);
-
- btrfs_release_path(path);
-
- /*
- * Mutex was contended, block until it's released and let
- * caller try again
- */
- mutex_lock(&head->mutex);
- mutex_unlock(&head->mutex);
- btrfs_put_delayed_ref_head(head);
- btrfs_put_transaction(cur_trans);
- return -EAGAIN;
- }
- spin_unlock(&delayed_refs->lock);
-
- spin_lock(&head->lock);
- /*
- * XXX: We should replace this with a proper search function in the
- * future.
- */
- for (node = rb_first_cached(&head->ref_tree); node;
- node = rb_next(node)) {
- ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
- /* If it's a shared ref we know a cross reference exists */
- if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
- ret = 1;
- break;
- }
-
- data_ref = btrfs_delayed_node_to_data_ref(ref);
-
- /*
- * If our ref doesn't match the one we're currently looking at
- * then we have a cross reference.
- */
- if (data_ref->root != root->root_key.objectid ||
- data_ref->objectid != objectid ||
- data_ref->offset != offset) {
- ret = 1;
- break;
- }
- }
- spin_unlock(&head->lock);
- mutex_unlock(&head->mutex);
- btrfs_put_transaction(cur_trans);
- return ret;
-}
-
-static noinline int check_committed_ref(struct btrfs_root *root,
- struct btrfs_path *path,
- u64 objectid, u64 offset, u64 bytenr)
-{
- struct btrfs_fs_info *fs_info = root->fs_info;
- struct btrfs_root *extent_root = fs_info->extent_root;
- struct extent_buffer *leaf;
- struct btrfs_extent_data_ref *ref;
- struct btrfs_extent_inline_ref *iref;
- struct btrfs_extent_item *ei;
- struct btrfs_key key;
- u32 item_size;
- int type;
- int ret;
-
- key.objectid = bytenr;
- key.offset = (u64)-1;
- key.type = BTRFS_EXTENT_ITEM_KEY;
-
- ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
- if (ret < 0)
- goto out;
- BUG_ON(ret == 0); /* Corruption */
-
- ret = -ENOENT;
- if (path->slots[0] == 0)
- goto out;
-
- path->slots[0]--;
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-
- if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
- goto out;
-
- ret = 1;
- item_size = btrfs_item_size_nr(leaf, path->slots[0]);
- ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
-
- if (item_size != sizeof(*ei) +
- btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
- goto out;
-
- if (btrfs_extent_generation(leaf, ei) <=
- btrfs_root_last_snapshot(&root->root_item))
- goto out;
-
- iref = (struct btrfs_extent_inline_ref *)(ei + 1);
-
- type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
- if (type != BTRFS_EXTENT_DATA_REF_KEY)
- goto out;
-
- ref = (struct btrfs_extent_data_ref *)(&iref->offset);
- if (btrfs_extent_refs(leaf, ei) !=
- btrfs_extent_data_ref_count(leaf, ref) ||
- btrfs_extent_data_ref_root(leaf, ref) !=
- root->root_key.objectid ||
- btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
- btrfs_extent_data_ref_offset(leaf, ref) != offset)
- goto out;
-
- ret = 0;
-out:
- return ret;
-}
-
-int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset,
- u64 bytenr)
-{
- struct btrfs_path *path;
- int ret;
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
-
- do {
- ret = check_committed_ref(root, path, objectid,
- offset, bytenr);
- if (ret && ret != -ENOENT)
- goto out;
-
- ret = check_delayed_ref(root, path, objectid, offset, bytenr);
- } while (ret == -EAGAIN);
-
-out:
- btrfs_free_path(path);
- if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
- WARN_ON(ret > 0);
- return ret;
-}
-
-static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct extent_buffer *buf,
- int full_backref, int inc)
-{
- struct btrfs_fs_info *fs_info = root->fs_info;
- u64 bytenr;
- u64 num_bytes;
- u64 parent;
- u64 ref_root;
- u32 nritems;
- struct btrfs_key key;
- struct btrfs_file_extent_item *fi;
- struct btrfs_ref generic_ref = { 0 };
- bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC);
- int i;
- int action;
- int level;
- int ret = 0;
-
- if (btrfs_is_testing(fs_info))
- return 0;
-
- ref_root = btrfs_header_owner(buf);
- nritems = btrfs_header_nritems(buf);
- level = btrfs_header_level(buf);
-
- if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state) && level == 0)
- return 0;
-
- if (full_backref)
- parent = buf->start;
- else
- parent = 0;
- if (inc)
- action = BTRFS_ADD_DELAYED_REF;
- else
- action = BTRFS_DROP_DELAYED_REF;
-
- for (i = 0; i < nritems; i++) {
- if (level == 0) {
- btrfs_item_key_to_cpu(buf, &key, i);
- if (key.type != BTRFS_EXTENT_DATA_KEY)
- continue;
- fi = btrfs_item_ptr(buf, i,
- struct btrfs_file_extent_item);
- if (btrfs_file_extent_type(buf, fi) ==
- BTRFS_FILE_EXTENT_INLINE)
- continue;
- bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
- if (bytenr == 0)
- continue;
-
- num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
- key.offset -= btrfs_file_extent_offset(buf, fi);
- btrfs_init_generic_ref(&generic_ref, action, bytenr,
- num_bytes, parent);
- generic_ref.real_root = root->root_key.objectid;
- btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
- key.offset);
- generic_ref.skip_qgroup = for_reloc;
- if (inc)
- ret = btrfs_inc_extent_ref(trans, &generic_ref);
- else
- ret = btrfs_free_extent(trans, &generic_ref);
- if (ret)
- goto fail;
- } else {
- bytenr = btrfs_node_blockptr(buf, i);
- num_bytes = fs_info->nodesize;
- btrfs_init_generic_ref(&generic_ref, action, bytenr,
- num_bytes, parent);
- generic_ref.real_root = root->root_key.objectid;
- btrfs_init_tree_ref(&generic_ref, level - 1, ref_root);
- generic_ref.skip_qgroup = for_reloc;
- if (inc)
- ret = btrfs_inc_extent_ref(trans, &generic_ref);
- else
- ret = btrfs_free_extent(trans, &generic_ref);
- if (ret)
- goto fail;
- }
- }
- return 0;
-fail:
- return ret;
-}
-
-int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
- struct extent_buffer *buf, int full_backref)
-{
- return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
-}
-
-int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
- struct extent_buffer *buf, int full_backref)
-{
- return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
-}
-
-static int write_one_cache_group(struct btrfs_trans_handle *trans,
- struct btrfs_path *path,
- struct btrfs_block_group_cache *cache)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- int ret;
- struct btrfs_root *extent_root = fs_info->extent_root;
- unsigned long bi;
- struct extent_buffer *leaf;
-
- ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
- if (ret) {
- if (ret > 0)
- ret = -ENOENT;
- goto fail;
- }
-
- leaf = path->nodes[0];
- bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
- write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
- btrfs_mark_buffer_dirty(leaf);
-fail:
- btrfs_release_path(path);
- return ret;
-
-}
-
-static struct btrfs_block_group_cache *next_block_group(
- struct btrfs_block_group_cache *cache)
-{
- struct btrfs_fs_info *fs_info = cache->fs_info;
- struct rb_node *node;
-
- spin_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->key.objectid + cache->key.offset;
-
- spin_unlock(&fs_info->block_group_cache_lock);
- btrfs_put_block_group(cache);
- cache = btrfs_lookup_first_block_group(fs_info, next_bytenr); return cache;
- }
- node = rb_next(&cache->cache_node);
- btrfs_put_block_group(cache);
- if (node) {
- cache = rb_entry(node, struct btrfs_block_group_cache,
- cache_node);
- btrfs_get_block_group(cache);
- } else
- cache = NULL;
- spin_unlock(&fs_info->block_group_cache_lock);
- return cache;
-}
-
-static int cache_save_setup(struct btrfs_block_group_cache *block_group,
- struct btrfs_trans_handle *trans,
- struct btrfs_path *path)
-{
- struct btrfs_fs_info *fs_info = block_group->fs_info;
- struct btrfs_root *root = fs_info->tree_root;
- struct inode *inode = NULL;
- struct extent_changeset *data_reserved = NULL;
- u64 alloc_hint = 0;
- int dcs = BTRFS_DC_ERROR;
- u64 num_pages = 0;
- int retries = 0;
- int ret = 0;
-
- /*
- * If this block group is smaller than 100 megs don't bother caching the
- * block group.
- */
- if (block_group->key.offset < (100 * SZ_1M)) {
- spin_lock(&block_group->lock);
- block_group->disk_cache_state = BTRFS_DC_WRITTEN;
- spin_unlock(&block_group->lock);
- return 0;
- }
-
- if (trans->aborted)
- return 0;
-again:
- inode = lookup_free_space_inode(block_group, path);
- if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {
- ret = PTR_ERR(inode);
- btrfs_release_path(path);
- goto out;
- }
-
- if (IS_ERR(inode)) {
- BUG_ON(retries);
- retries++;
-
- if (block_group->ro)
- goto out_free;
-
- ret = create_free_space_inode(trans, block_group, path);
- if (ret)
- goto out_free;
- goto again;
- }
-
- /*
- * We want to set the generation to 0, that way if anything goes wrong
- * from here on out we know not to trust this cache when we load up next
- * time.
- */
- BTRFS_I(inode)->generation = 0;
- ret = btrfs_update_inode(trans, root, inode);
- if (ret) {
- /*
- * So theoretically we could recover from this, simply set the
- * super cache generation to 0 so we know to invalidate the
- * cache, but then we'd have to keep track of the block groups
- * that fail this way so we know we _have_ to reset this cache
- * before the next commit or risk reading stale cache. So to
- * limit our exposure to horrible edge cases lets just abort the
- * transaction, this only happens in really bad situations
- * anyway.
- */
- btrfs_abort_transaction(trans, ret);
- goto out_put;
- }
- WARN_ON(ret);
-
- /* We've already setup this transaction, go ahead and exit */
- if (block_group->cache_generation == trans->transid &&
- i_size_read(inode)) {
- dcs = BTRFS_DC_SETUP;
- goto out_put;
- }
-
- if (i_size_read(inode) > 0) {
- ret = btrfs_check_trunc_cache_free_space(fs_info,
- &fs_info->global_block_rsv);
- if (ret)
- goto out_put;
-
- ret = btrfs_truncate_free_space_cache(trans, NULL, inode);
- if (ret)
- goto out_put;
- }
-
- spin_lock(&block_group->lock);
- if (block_group->cached != BTRFS_CACHE_FINISHED ||
- !btrfs_test_opt(fs_info, SPACE_CACHE)) {
- /*
- * don't bother trying to write stuff out _if_
- * a) we're not cached,
- * b) we're with nospace_cache mount option,
- * c) we're with v2 space_cache (FREE_SPACE_TREE).
- */
- dcs = BTRFS_DC_WRITTEN;
- spin_unlock(&block_group->lock);
- goto out_put;
- }
- spin_unlock(&block_group->lock);
-
- /*
- * We hit an ENOSPC when setting up the cache in this transaction, just
- * skip doing the setup, we've already cleared the cache so we're safe.
- */
- if (test_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags)) {
- ret = -ENOSPC;
- goto out_put;
- }
-
- /*
- * Try to preallocate enough space based on how big the block group is.
- * Keep in mind this has to include any pinned space which could end up
- * taking up quite a bit since it's not folded into the other space
- * cache.
- */
- num_pages = div_u64(block_group->key.offset, SZ_256M);
- if (!num_pages)
- num_pages = 1;
-
- num_pages *= 16;
- num_pages *= PAGE_SIZE;
-
- ret = btrfs_check_data_free_space(inode, &data_reserved, 0, num_pages);
- if (ret)
- goto out_put;
-
- ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
- num_pages, num_pages,
- &alloc_hint);
- /*
- * Our cache requires contiguous chunks so that we don't modify a bunch
- * of metadata or split extents when writing the cache out, which means
- * we can enospc if we are heavily fragmented in addition to just normal
- * out of space conditions. So if we hit this just skip setting up any
- * other block groups for this transaction, maybe we'll unpin enough
- * space the next time around.
- */
- if (!ret)
- dcs = BTRFS_DC_SETUP;
- else if (ret == -ENOSPC)
- set_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags);
-
-out_put:
- iput(inode);
-out_free:
- btrfs_release_path(path);
-out:
- spin_lock(&block_group->lock);
- if (!ret && dcs == BTRFS_DC_SETUP)
- block_group->cache_generation = trans->transid;
- block_group->disk_cache_state = dcs;
- spin_unlock(&block_group->lock);
-
- extent_changeset_free(data_reserved);
- return ret;
-}
-
-int btrfs_setup_space_cache(struct btrfs_trans_handle *trans)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_block_group_cache *cache, *tmp;
- struct btrfs_transaction *cur_trans = trans->transaction;
- struct btrfs_path *path;
-
- if (list_empty(&cur_trans->dirty_bgs) ||
- !btrfs_test_opt(fs_info, SPACE_CACHE))
- return 0;
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
-
- /* Could add new block groups, use _safe just in case */
- list_for_each_entry_safe(cache, tmp, &cur_trans->dirty_bgs,
- dirty_list) {
- if (cache->disk_cache_state == BTRFS_DC_CLEAR)
- cache_save_setup(cache, trans, path);
- }
-
- btrfs_free_path(path);
- return 0;
-}
-
-/*
- * transaction commit does final block group cache writeback during a
- * critical section where nothing is allowed to change the FS. This is
- * required in order for the cache to actually match the block group,
- * but can introduce a lot of latency into the commit.
- *
- * So, btrfs_start_dirty_block_groups is here to kick off block group
- * cache IO. There's a chance we'll have to redo some of it if the
- * block group changes again during the commit, but it greatly reduces
- * the commit latency by getting rid of the easy block groups while
- * we're still allowing others to join the commit.
- */
-int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_block_group_cache *cache;
- struct btrfs_transaction *cur_trans = trans->transaction;
- int ret = 0;
- int should_put;
- 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);
- if (list_empty(&cur_trans->dirty_bgs)) {
- spin_unlock(&cur_trans->dirty_bgs_lock);
- return 0;
- }
- list_splice_init(&cur_trans->dirty_bgs, &dirty);
- spin_unlock(&cur_trans->dirty_bgs_lock);
-
-again:
- /*
- * make sure all the block groups on our dirty list actually
- * exist
- */
- btrfs_create_pending_block_groups(trans);
-
- if (!path) {
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
- }
-
- /*
- * cache_write_mutex is here only to save us from balance or automatic
- * removal of empty block groups deleting this block group while we are
- * writing out the cache
- */
- mutex_lock(&trans->transaction->cache_write_mutex);
- while (!list_empty(&dirty)) {
- bool drop_reserve = true;
-
- cache = list_first_entry(&dirty,
- struct btrfs_block_group_cache,
- dirty_list);
- /*
- * this can happen if something re-dirties a block
- * group that is already under IO. Just wait for it to
- * finish and then do it all again
- */
- if (!list_empty(&cache->io_list)) {
- list_del_init(&cache->io_list);
- btrfs_wait_cache_io(trans, cache, path);
- btrfs_put_block_group(cache);
- }
-
-
- /*
- * btrfs_wait_cache_io uses the cache->dirty_list to decide
- * if it should update the cache_state. Don't delete
- * until after we wait.
- *
- * Since we're not running in the commit critical section
- * we need the dirty_bgs_lock to protect from update_block_group
- */
- spin_lock(&cur_trans->dirty_bgs_lock);
- list_del_init(&cache->dirty_list);
- spin_unlock(&cur_trans->dirty_bgs_lock);
-
- should_put = 1;
-
- cache_save_setup(cache, trans, path);
-
- if (cache->disk_cache_state == BTRFS_DC_SETUP) {
- 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;
-
- /*
- * The cache_write_mutex is protecting the
- * io_list, also refer to the definition of
- * btrfs_transaction::io_bgs for more details
- */
- list_add_tail(&cache->io_list, io);
- } else {
- /*
- * if we failed to write the cache, the
- * generation will be bad and life goes on
- */
- ret = 0;
- }
- }
- if (!ret) {
- ret = write_one_cache_group(trans, path, cache);
- /*
- * Our block group might still be attached to the list
- * of new block groups in the transaction handle of some
- * other task (struct btrfs_trans_handle->new_bgs). This
- * means its block group item isn't yet in the extent
- * tree. If this happens ignore the error, as we will
- * try again later in the critical section of the
- * transaction commit.
- */
- if (ret == -ENOENT) {
- ret = 0;
- spin_lock(&cur_trans->dirty_bgs_lock);
- if (list_empty(&cache->dirty_list)) {
- list_add_tail(&cache->dirty_list,
- &cur_trans->dirty_bgs);
- btrfs_get_block_group(cache);
- drop_reserve = false;
- }
- spin_unlock(&cur_trans->dirty_bgs_lock);
- } else if (ret) {
- btrfs_abort_transaction(trans, ret);
- }
- }
-
- /* if it's not on the io list, we need to put the block group */
- if (should_put)
- btrfs_put_block_group(cache);
- if (drop_reserve)
- btrfs_delayed_refs_rsv_release(fs_info, 1);
-
- if (ret)
- break;
+ ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+ flags = btrfs_extent_flags(leaf, ei);
- /*
- * 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);
- mutex_lock(&trans->transaction->cache_write_mutex);
- }
- mutex_unlock(&trans->transaction->cache_write_mutex);
+ ptr = (unsigned long)(ei + 1);
+ end = (unsigned long)ei + item_size;
- /*
- * 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 && loops == 0) {
- loops++;
- spin_lock(&cur_trans->dirty_bgs_lock);
- list_splice_init(&cur_trans->dirty_bgs, &dirty);
- /*
- * dirty_bgs_lock protects us from concurrent block group
- * deletes too (not just cache_write_mutex).
- */
- if (!list_empty(&dirty)) {
- spin_unlock(&cur_trans->dirty_bgs_lock);
- goto again;
- }
- spin_unlock(&cur_trans->dirty_bgs_lock);
- } else if (ret < 0) {
- btrfs_cleanup_dirty_bgs(cur_trans, fs_info);
+ if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
+ ptr += sizeof(struct btrfs_tree_block_info);
+ BUG_ON(ptr > end);
}
- btrfs_free_path(path);
- return ret;
-}
-
-int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans)
-{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_block_group_cache *cache;
- struct btrfs_transaction *cur_trans = trans->transaction;
- int ret = 0;
- 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)
- return -ENOMEM;
-
- /*
- * Even though we are in the critical section of the transaction commit,
- * we can still have concurrent tasks adding elements to this
- * transaction's list of dirty block groups. These tasks correspond to
- * endio free space workers started when writeback finishes for a
- * space cache, which run inode.c:btrfs_finish_ordered_io(), and can
- * allocate new block groups as a result of COWing nodes of the root
- * tree when updating the free space inode. The writeback for the space
- * caches is triggered by an earlier call to
- * btrfs_start_dirty_block_groups() and iterations of the following
- * loop.
- * Also we want to do the cache_save_setup first and then run the
- * delayed refs to make sure we have the best chance at doing this all
- * in one shot.
- */
- spin_lock(&cur_trans->dirty_bgs_lock);
- while (!list_empty(&cur_trans->dirty_bgs)) {
- cache = list_first_entry(&cur_trans->dirty_bgs,
- struct btrfs_block_group_cache,
- dirty_list);
+ if (owner >= BTRFS_FIRST_FREE_OBJECTID)
+ needed = BTRFS_REF_TYPE_DATA;
+ else
+ needed = BTRFS_REF_TYPE_BLOCK;
- /*
- * this can happen if cache_save_setup re-dirties a block
- * group that is already under IO. Just wait for it to
- * finish and then do it all again
- */
- if (!list_empty(&cache->io_list)) {
- spin_unlock(&cur_trans->dirty_bgs_lock);
- list_del_init(&cache->io_list);
- btrfs_wait_cache_io(trans, cache, path);
- btrfs_put_block_group(cache);
- spin_lock(&cur_trans->dirty_bgs_lock);
+ err = -ENOENT;
+ while (1) {
+ if (ptr >= end) {
+ WARN_ON(ptr > end);
+ break;
+ }
+ iref = (struct btrfs_extent_inline_ref *)ptr;
+ type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
+ if (type == BTRFS_REF_TYPE_INVALID) {
+ err = -EUCLEAN;
+ goto out;
}
- /*
- * don't remove from the dirty list until after we've waited
- * on any pending IO
- */
- list_del_init(&cache->dirty_list);
- spin_unlock(&cur_trans->dirty_bgs_lock);
- should_put = 1;
-
- cache_save_setup(cache, trans, path);
+ if (want < type)
+ break;
+ if (want > type) {
+ ptr += btrfs_extent_inline_ref_size(type);
+ continue;
+ }
- if (!ret)
- ret = btrfs_run_delayed_refs(trans,
- (unsigned long) -1);
-
- if (!ret && cache->disk_cache_state == BTRFS_DC_SETUP) {
- 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 {
- /*
- * if we failed to write the cache, the
- * generation will be bad and life goes on
- */
- ret = 0;
+ if (type == BTRFS_EXTENT_DATA_REF_KEY) {
+ struct btrfs_extent_data_ref *dref;
+ dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+ if (match_extent_data_ref(leaf, dref, root_objectid,
+ owner, offset)) {
+ err = 0;
+ break;
}
- }
- if (!ret) {
- ret = write_one_cache_group(trans, path, cache);
- /*
- * One of the free space endio workers might have
- * created a new block group while updating a free space
- * cache's inode (at inode.c:btrfs_finish_ordered_io())
- * and hasn't released its transaction handle yet, in
- * which case the new block group is still attached to
- * its transaction handle and its creation has not
- * 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
- * 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);
+ if (hash_extent_data_ref_item(leaf, dref) <
+ hash_extent_data_ref(root_objectid, owner, offset))
+ break;
+ } else {
+ u64 ref_offset;
+ ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
+ if (parent > 0) {
+ if (parent == ref_offset) {
+ err = 0;
+ break;
+ }
+ if (ref_offset < parent)
+ break;
+ } else {
+ if (root_objectid == ref_offset) {
+ err = 0;
+ break;
+ }
+ if (ref_offset < root_objectid)
+ break;
}
- if (ret)
- btrfs_abort_transaction(trans, ret);
}
-
- /* if its not on the io list, we need to put the block group */
- if (should_put)
- btrfs_put_block_group(cache);
- btrfs_delayed_refs_rsv_release(fs_info, 1);
- spin_lock(&cur_trans->dirty_bgs_lock);
+ ptr += btrfs_extent_inline_ref_size(type);
}
- spin_unlock(&cur_trans->dirty_bgs_lock);
-
- /*
- * Refer to the definition of io_bgs member for details why it's safe
- * to use it without any locking
- */
- while (!list_empty(io)) {
- cache = list_first_entry(io, struct btrfs_block_group_cache,
- io_list);
- list_del_init(&cache->io_list);
- btrfs_wait_cache_io(trans, cache, path);
- btrfs_put_block_group(cache);
+ if (err == -ENOENT && insert) {
+ if (item_size + extra_size >=
+ BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
+ err = -EAGAIN;
+ goto out;
+ }
+ /*
+ * To add new inline back ref, we have to make sure
+ * there is no corresponding back ref item.
+ * For simplicity, we just do not add new inline back
+ * ref if there is any kind of item for this block
+ */
+ if (find_next_key(path, 0, &key) == 0 &&
+ key.objectid == bytenr &&
+ key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
+ err = -EAGAIN;
+ goto out;
+ }
}
-
- btrfs_free_path(path);
- return ret;
-}
-
-int btrfs_extent_readonly(struct btrfs_fs_info *fs_info, u64 bytenr)
-{
- struct btrfs_block_group_cache *block_group;
- int readonly = 0;
-
- block_group = btrfs_lookup_block_group(fs_info, bytenr);
- if (!block_group || block_group->ro)
- readonly = 1;
- if (block_group)
- btrfs_put_block_group(block_group);
- return readonly;
-}
-
-bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr)
-{
- struct btrfs_block_group_cache *bg;
- bool ret = true;
-
- bg = btrfs_lookup_block_group(fs_info, bytenr);
- if (!bg)
- return false;
-
- spin_lock(&bg->lock);
- if (bg->ro)
- ret = false;
- else
- atomic_inc(&bg->nocow_writers);
- spin_unlock(&bg->lock);
-
- /* no put on block group, done by btrfs_dec_nocow_writers */
- if (!ret)
- btrfs_put_block_group(bg);
-
- return ret;
-
-}
-
-void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr)
-{
- struct btrfs_block_group_cache *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);
- btrfs_put_block_group(bg);
-}
-
-void btrfs_wait_nocow_writers(struct btrfs_block_group_cache *bg)
-{
- wait_var_event(&bg->nocow_writers, !atomic_read(&bg->nocow_writers));
-}
-
-static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
-{
- u64 extra_flags = chunk_to_extended(flags) &
- BTRFS_EXTENDED_PROFILE_MASK;
-
- write_seqlock(&fs_info->profiles_lock);
- if (flags & BTRFS_BLOCK_GROUP_DATA)
- fs_info->avail_data_alloc_bits |= extra_flags;
- if (flags & BTRFS_BLOCK_GROUP_METADATA)
- fs_info->avail_metadata_alloc_bits |= extra_flags;
- if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
- fs_info->avail_system_alloc_bits |= extra_flags;
- write_sequnlock(&fs_info->profiles_lock);
-}
-
-/*
- * returns target flags in extended format or 0 if restripe for this
- * chunk_type is not in progress
- *
- * should be called with balance_lock held
- */
-static u64 get_restripe_target(struct btrfs_fs_info *fs_info, u64 flags)
-{
- struct btrfs_balance_control *bctl = fs_info->balance_ctl;
- u64 target = 0;
-
- if (!bctl)
- return 0;
-
- if (flags & BTRFS_BLOCK_GROUP_DATA &&
- bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) {
- target = BTRFS_BLOCK_GROUP_DATA | bctl->data.target;
- } else if (flags & BTRFS_BLOCK_GROUP_SYSTEM &&
- bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {
- target = BTRFS_BLOCK_GROUP_SYSTEM | bctl->sys.target;
- } else if (flags & BTRFS_BLOCK_GROUP_METADATA &&
- bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) {
- target = BTRFS_BLOCK_GROUP_METADATA | bctl->meta.target;
+ *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
+out:
+ if (insert) {
+ path->keep_locks = 0;
+ btrfs_unlock_up_safe(path, 1);
}
-
- return target;
+ return err;
}
/*
- * @flags: available profiles in extended format (see ctree.h)
- *
- * Returns reduced profile in chunk format. If profile changing is in
- * progress (either running or paused) picks the target profile (if it's
- * already available), otherwise falls back to plain reducing.
+ * helper to add new inline back ref
*/
-static u64 btrfs_reduce_alloc_profile(struct btrfs_fs_info *fs_info, u64 flags)
-{
- u64 num_devices = fs_info->fs_devices->rw_devices;
- u64 target;
- u64 raid_type;
- u64 allowed = 0;
-
- /*
- * see if restripe for this chunk_type is in progress, if so
- * try to reduce to the target profile
- */
- 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);
-
- /* First, mask out the RAID levels which aren't possible */
- for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
- if (num_devices >= btrfs_raid_array[raid_type].devs_min)
- allowed |= btrfs_raid_array[raid_type].bg_flag;
- }
- allowed &= flags;
-
- if (allowed & BTRFS_BLOCK_GROUP_RAID6)
- allowed = BTRFS_BLOCK_GROUP_RAID6;
- else if (allowed & BTRFS_BLOCK_GROUP_RAID5)
- allowed = BTRFS_BLOCK_GROUP_RAID5;
- else if (allowed & BTRFS_BLOCK_GROUP_RAID10)
- allowed = BTRFS_BLOCK_GROUP_RAID10;
- else if (allowed & BTRFS_BLOCK_GROUP_RAID1)
- allowed = BTRFS_BLOCK_GROUP_RAID1;
- else if (allowed & BTRFS_BLOCK_GROUP_RAID0)
- allowed = BTRFS_BLOCK_GROUP_RAID0;
-
- flags &= ~BTRFS_BLOCK_GROUP_PROFILE_MASK;
-
- return extended_to_chunk(flags | allowed);
-}
-
-static u64 get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags)
-{
- unsigned seq;
- u64 flags;
-
- do {
- flags = orig_flags;
- seq = read_seqbegin(&fs_info->profiles_lock);
-
- if (flags & BTRFS_BLOCK_GROUP_DATA)
- flags |= fs_info->avail_data_alloc_bits;
- else if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
- flags |= fs_info->avail_system_alloc_bits;
- else if (flags & BTRFS_BLOCK_GROUP_METADATA)
- flags |= fs_info->avail_metadata_alloc_bits;
- } while (read_seqretry(&fs_info->profiles_lock, seq));
-
- return btrfs_reduce_alloc_profile(fs_info, flags);
-}
-
-static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data)
+static noinline_for_stack
+void setup_inline_extent_backref(struct btrfs_fs_info *fs_info,
+ struct btrfs_path *path,
+ struct btrfs_extent_inline_ref *iref,
+ u64 parent, u64 root_objectid,
+ u64 owner, u64 offset, int refs_to_add,
+ struct btrfs_delayed_extent_op *extent_op)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- u64 flags;
- u64 ret;
-
- if (data)
- flags = BTRFS_BLOCK_GROUP_DATA;
- else if (root == fs_info->chunk_root)
- flags = BTRFS_BLOCK_GROUP_SYSTEM;
- else
- flags = BTRFS_BLOCK_GROUP_METADATA;
+ struct extent_buffer *leaf;
+ struct btrfs_extent_item *ei;
+ unsigned long ptr;
+ unsigned long end;
+ unsigned long item_offset;
+ u64 refs;
+ int size;
+ int type;
- ret = get_alloc_profile(fs_info, flags);
- return ret;
-}
+ leaf = path->nodes[0];
+ ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+ item_offset = (unsigned long)iref - (unsigned long)ei;
-u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
-{
- return get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
-}
+ type = extent_ref_type(parent, owner);
+ size = btrfs_extent_inline_ref_size(type);
-u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
-{
- return get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
-}
+ btrfs_extend_item(path, size);
-u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
-{
- return get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
-}
+ ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+ refs = btrfs_extent_refs(leaf, ei);
+ refs += refs_to_add;
+ btrfs_set_extent_refs(leaf, ei, refs);
+ if (extent_op)
+ __run_delayed_extent_op(extent_op, leaf, ei);
-static void force_metadata_allocation(struct btrfs_fs_info *info)
-{
- struct list_head *head = &info->space_info;
- struct btrfs_space_info *found;
+ ptr = (unsigned long)ei + item_offset;
+ end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
+ if (ptr < end - size)
+ memmove_extent_buffer(leaf, ptr + size, ptr,
+ end - size - ptr);
- rcu_read_lock();
- list_for_each_entry_rcu(found, head, list) {
- if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
- found->force_alloc = CHUNK_ALLOC_FORCE;
+ iref = (struct btrfs_extent_inline_ref *)ptr;
+ btrfs_set_extent_inline_ref_type(leaf, iref, type);
+ if (type == BTRFS_EXTENT_DATA_REF_KEY) {
+ struct btrfs_extent_data_ref *dref;
+ dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+ btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
+ btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
+ btrfs_set_extent_data_ref_offset(leaf, dref, offset);
+ btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
+ } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
+ struct btrfs_shared_data_ref *sref;
+ sref = (struct btrfs_shared_data_ref *)(iref + 1);
+ btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
+ btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
+ } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
+ btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
+ } else {
+ btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
}
- rcu_read_unlock();
+ btrfs_mark_buffer_dirty(leaf);
}
-static int should_alloc_chunk(struct btrfs_fs_info *fs_info,
- struct btrfs_space_info *sinfo, int force)
+static int lookup_extent_backref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ struct btrfs_extent_inline_ref **ref_ret,
+ u64 bytenr, u64 num_bytes, u64 parent,
+ u64 root_objectid, u64 owner, u64 offset)
{
- u64 bytes_used = btrfs_space_info_used(sinfo, false);
- u64 thresh;
+ int ret;
- if (force == CHUNK_ALLOC_FORCE)
- return 1;
+ ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr,
+ num_bytes, parent, root_objectid,
+ owner, offset, 0);
+ if (ret != -ENOENT)
+ return ret;
- /*
- * in limited mode, we want to have some free space up to
- * about 1% of the FS size.
- */
- if (force == CHUNK_ALLOC_LIMITED) {
- thresh = btrfs_super_total_bytes(fs_info->super_copy);
- thresh = max_t(u64, SZ_64M, div_factor_fine(thresh, 1));
+ btrfs_release_path(path);
+ *ref_ret = NULL;
- if (sinfo->total_bytes - bytes_used < thresh)
- return 1;
+ if (owner < BTRFS_FIRST_FREE_OBJECTID) {
+ ret = lookup_tree_block_ref(trans, path, bytenr, parent,
+ root_objectid);
+ } else {
+ ret = lookup_extent_data_ref(trans, path, bytenr, parent,
+ root_objectid, owner, offset);
}
-
- if (bytes_used + SZ_2M < div_factor(sinfo->total_bytes, 8))
- return 0;
- return 1;
-}
-
-static u64 get_profile_num_devs(struct btrfs_fs_info *fs_info, u64 type)
-{
- u64 num_dev;
-
- num_dev = btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)].devs_max;
- if (!num_dev)
- num_dev = fs_info->fs_devices->rw_devices;
-
- return num_dev;
+ return ret;
}
/*
- * If @is_allocation is true, reserve space in the system space info necessary
- * for allocating a chunk, otherwise if it's false, reserve space necessary for
- * removing a chunk.
+ * helper to update/remove inline back ref
*/
-void check_system_chunk(struct btrfs_trans_handle *trans, u64 type)
+static noinline_for_stack
+void update_inline_extent_backref(struct btrfs_path *path,
+ struct btrfs_extent_inline_ref *iref,
+ int refs_to_mod,
+ struct btrfs_delayed_extent_op *extent_op,
+ int *last_ref)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_space_info *info;
- u64 left;
- u64 thresh;
- int ret = 0;
- u64 num_devs;
+ struct extent_buffer *leaf = path->nodes[0];
+ struct btrfs_extent_item *ei;
+ struct btrfs_extent_data_ref *dref = NULL;
+ struct btrfs_shared_data_ref *sref = NULL;
+ unsigned long ptr;
+ unsigned long end;
+ u32 item_size;
+ int size;
+ int type;
+ u64 refs;
+
+ ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+ refs = btrfs_extent_refs(leaf, ei);
+ WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
+ refs += refs_to_mod;
+ btrfs_set_extent_refs(leaf, ei, refs);
+ if (extent_op)
+ __run_delayed_extent_op(extent_op, leaf, ei);
/*
- * Needed because we can end up allocating a system chunk and for an
- * atomic and race free space reservation in the chunk block reserve.
+ * If type is invalid, we should have bailed out after
+ * lookup_inline_extent_backref().
*/
- lockdep_assert_held(&fs_info->chunk_mutex);
-
- info = btrfs_find_space_info(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
- spin_lock(&info->lock);
- left = info->total_bytes - btrfs_space_info_used(info, true);
- spin_unlock(&info->lock);
+ type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
+ ASSERT(type != BTRFS_REF_TYPE_INVALID);
- num_devs = get_profile_num_devs(fs_info, type);
+ if (type == BTRFS_EXTENT_DATA_REF_KEY) {
+ dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+ refs = btrfs_extent_data_ref_count(leaf, dref);
+ } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
+ sref = (struct btrfs_shared_data_ref *)(iref + 1);
+ refs = btrfs_shared_data_ref_count(leaf, sref);
+ } else {
+ refs = 1;
+ BUG_ON(refs_to_mod != -1);
+ }
- /* num_devs device items to update and 1 chunk item to add or remove */
- thresh = btrfs_calc_trunc_metadata_size(fs_info, num_devs) +
- btrfs_calc_trans_metadata_size(fs_info, 1);
+ BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
+ refs += refs_to_mod;
- if (left < thresh && btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
- btrfs_info(fs_info, "left=%llu, need=%llu, flags=%llu",
- left, thresh, type);
- btrfs_dump_space_info(fs_info, info, 0, 0);
+ if (refs > 0) {
+ if (type == BTRFS_EXTENT_DATA_REF_KEY)
+ btrfs_set_extent_data_ref_count(leaf, dref, refs);
+ else
+ btrfs_set_shared_data_ref_count(leaf, sref, refs);
+ } else {
+ *last_ref = 1;
+ size = btrfs_extent_inline_ref_size(type);
+ item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+ ptr = (unsigned long)iref;
+ end = (unsigned long)ei + item_size;
+ if (ptr + size < end)
+ memmove_extent_buffer(leaf, ptr, ptr + size,
+ end - ptr - size);
+ item_size -= size;
+ btrfs_truncate_item(path, item_size, 1);
}
+ btrfs_mark_buffer_dirty(leaf);
+}
- if (left < thresh) {
- u64 flags = btrfs_system_alloc_profile(fs_info);
+static noinline_for_stack
+int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ u64 bytenr, u64 num_bytes, u64 parent,
+ u64 root_objectid, u64 owner,
+ u64 offset, int refs_to_add,
+ struct btrfs_delayed_extent_op *extent_op)
+{
+ struct btrfs_extent_inline_ref *iref;
+ int ret;
- /*
- * Ignore failure to create system chunk. We might end up not
- * needing it, as we might not need to COW all nodes/leafs from
- * 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);
+ ret = lookup_inline_extent_backref(trans, path, &iref, bytenr,
+ num_bytes, parent, root_objectid,
+ owner, offset, 1);
+ if (ret == 0) {
+ BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
+ update_inline_extent_backref(path, iref, refs_to_add,
+ extent_op, NULL);
+ } else if (ret == -ENOENT) {
+ setup_inline_extent_backref(trans->fs_info, path, iref, parent,
+ root_objectid, owner, offset,
+ refs_to_add, extent_op);
+ ret = 0;
}
+ return ret;
+}
- if (!ret) {
- ret = btrfs_block_rsv_add(fs_info->chunk_root,
- &fs_info->chunk_block_rsv,
- thresh, BTRFS_RESERVE_NO_FLUSH);
- if (!ret)
- trans->chunk_bytes_reserved += thresh;
+static int insert_extent_backref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ u64 bytenr, u64 parent, u64 root_objectid,
+ u64 owner, u64 offset, int refs_to_add)
+{
+ int ret;
+ if (owner < BTRFS_FIRST_FREE_OBJECTID) {
+ BUG_ON(refs_to_add != 1);
+ ret = insert_tree_block_ref(trans, path, bytenr, parent,
+ root_objectid);
+ } else {
+ ret = insert_extent_data_ref(trans, path, bytenr, parent,
+ root_objectid, owner, offset,
+ refs_to_add);
}
+ return ret;
}
-/*
- * 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:
- * - 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.
- */
-int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
- enum btrfs_chunk_alloc_enum force)
+static int remove_extent_backref(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ struct btrfs_extent_inline_ref *iref,
+ int refs_to_drop, int is_data, int *last_ref)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_space_info *space_info;
- bool wait_for_alloc = false;
- bool should_alloc = false;
int ret = 0;
- /* Don't re-enter if we're already allocating a chunk */
- if (trans->allocating_chunk)
- return -ENOSPC;
-
- space_info = btrfs_find_space_info(fs_info, flags);
- ASSERT(space_info);
-
- do {
- spin_lock(&space_info->lock);
- if (force < space_info->force_alloc)
- force = space_info->force_alloc;
- should_alloc = should_alloc_chunk(fs_info, space_info, force);
- if (space_info->full) {
- /* No more free physical space */
- if (should_alloc)
- ret = -ENOSPC;
- else
- ret = 0;
- spin_unlock(&space_info->lock);
- return ret;
- } else if (!should_alloc) {
- spin_unlock(&space_info->lock);
- return 0;
- } else if (space_info->chunk_alloc) {
- /*
- * Someone is already allocating, so we need to block
- * until this someone is finished and then loop to
- * recheck if we should continue with our allocation
- * attempt.
- */
- wait_for_alloc = true;
- spin_unlock(&space_info->lock);
- mutex_lock(&fs_info->chunk_mutex);
- mutex_unlock(&fs_info->chunk_mutex);
- } else {
- /* Proceed with allocation */
- space_info->chunk_alloc = 1;
- wait_for_alloc = false;
- spin_unlock(&space_info->lock);
- }
-
- cond_resched();
- } while (wait_for_alloc);
-
- mutex_lock(&fs_info->chunk_mutex);
- trans->allocating_chunk = true;
+ BUG_ON(!is_data && refs_to_drop != 1);
+ if (iref) {
+ update_inline_extent_backref(path, iref, -refs_to_drop, NULL,
+ last_ref);
+ } else if (is_data) {
+ ret = remove_extent_data_ref(trans, path, refs_to_drop,
+ last_ref);
+ } else {
+ *last_ref = 1;
+ ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
+ }
+ return ret;
+}
- /*
- * If we have mixed data/metadata chunks we want to make sure we keep
- * allocating mixed chunks instead of individual chunks.
- */
- if (btrfs_mixed_space_info(space_info))
- flags |= (BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA);
+static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
+ u64 *discarded_bytes)
+{
+ int j, ret = 0;
+ u64 bytes_left, end;
+ u64 aligned_start = ALIGN(start, 1 << 9);
- /*
- * if we're doing a data chunk, go ahead and make sure that
- * we keep a reasonable number of metadata chunks allocated in the
- * FS as well.
- */
- if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
- fs_info->data_chunk_allocations++;
- if (!(fs_info->data_chunk_allocations %
- fs_info->metadata_ratio))
- force_metadata_allocation(fs_info);
+ if (WARN_ON(start != aligned_start)) {
+ len -= aligned_start - start;
+ len = round_down(len, 1 << 9);
+ start = aligned_start;
}
- /*
- * 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);
- trans->allocating_chunk = false;
+ *discarded_bytes = 0;
- spin_lock(&space_info->lock);
- if (ret < 0) {
- if (ret == -ENOSPC)
- space_info->full = 1;
- else
- goto out;
- } else {
- ret = 1;
- space_info->max_extent_size = 0;
- }
+ if (!len)
+ return 0;
- space_info->force_alloc = CHUNK_ALLOC_NO_FORCE;
-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);
+ end = start + len;
+ bytes_left = len;
- return ret;
-}
+ /* Skip any superblocks on this device. */
+ for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
+ u64 sb_start = btrfs_sb_offset(j);
+ u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
+ u64 size = sb_start - start;
-static int update_block_group(struct btrfs_trans_handle *trans,
- u64 bytenr, u64 num_bytes, int alloc)
-{
- struct btrfs_fs_info *info = trans->fs_info;
- struct btrfs_block_group_cache *cache = NULL;
- u64 total = num_bytes;
- u64 old_val;
- u64 byte_in_group;
- int factor;
- int ret = 0;
+ if (!in_range(sb_start, start, bytes_left) &&
+ !in_range(sb_end, start, bytes_left) &&
+ !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
+ continue;
- /* block accounting for super block */
- spin_lock(&info->delalloc_root_lock);
- old_val = btrfs_super_bytes_used(info->super_copy);
- if (alloc)
- old_val += num_bytes;
- else
- old_val -= num_bytes;
- btrfs_set_super_bytes_used(info->super_copy, old_val);
- spin_unlock(&info->delalloc_root_lock);
-
- while (total) {
- cache = btrfs_lookup_block_group(info, bytenr);
- if (!cache) {
- ret = -ENOENT;
- break;
+ /*
+ * Superblock spans beginning of range. Adjust start and
+ * try again.
+ */
+ if (sb_start <= start) {
+ start += sb_end - start;
+ if (start > end) {
+ bytes_left = 0;
+ break;
+ }
+ bytes_left = end - start;
+ continue;
}
- factor = btrfs_bg_type_to_factor(cache->flags);
-
- /*
- * If this block group has free space cache written out, we
- * need to make sure to load it if we are removing space. This
- * is because we need the unpinning stage to actually add the
- * space back to the block group, otherwise we will leak space.
- */
- if (!alloc && cache->cached == BTRFS_CACHE_NO)
- cache_block_group(cache, 1);
-
- byte_in_group = bytenr - cache->key.objectid;
- WARN_ON(byte_in_group > cache->key.offset);
-
- spin_lock(&cache->space_info->lock);
- spin_lock(&cache->lock);
- if (btrfs_test_opt(info, SPACE_CACHE) &&
- cache->disk_cache_state < BTRFS_DC_CLEAR)
- cache->disk_cache_state = BTRFS_DC_CLEAR;
-
- old_val = btrfs_block_group_used(&cache->item);
- num_bytes = min(total, cache->key.offset - byte_in_group);
- if (alloc) {
- old_val += num_bytes;
- btrfs_set_block_group_used(&cache->item, old_val);
- cache->reserved -= num_bytes;
- cache->space_info->bytes_reserved -= num_bytes;
- cache->space_info->bytes_used += num_bytes;
- cache->space_info->disk_used += num_bytes * factor;
- spin_unlock(&cache->lock);
- spin_unlock(&cache->space_info->lock);
- } else {
- old_val -= num_bytes;
- btrfs_set_block_group_used(&cache->item, old_val);
- cache->pinned += num_bytes;
- btrfs_space_info_update_bytes_pinned(info,
- cache->space_info, num_bytes);
- cache->space_info->bytes_used -= num_bytes;
- cache->space_info->disk_used -= num_bytes * factor;
- spin_unlock(&cache->lock);
- spin_unlock(&cache->space_info->lock);
-
- trace_btrfs_space_reservation(info, "pinned",
- cache->space_info->flags,
- num_bytes, 1);
- percpu_counter_add_batch(&cache->space_info->total_bytes_pinned,
- num_bytes,
- BTRFS_TOTAL_BYTES_PINNED_BATCH);
- set_extent_dirty(info->pinned_extents,
- bytenr, bytenr + num_bytes - 1,
- GFP_NOFS | __GFP_NOFAIL);
+ if (size) {
+ ret = blkdev_issue_discard(bdev, start >> 9, size >> 9,
+ GFP_NOFS, 0);
+ if (!ret)
+ *discarded_bytes += size;
+ else if (ret != -EOPNOTSUPP)
+ return ret;
}
- spin_lock(&trans->transaction->dirty_bgs_lock);
- if (list_empty(&cache->dirty_list)) {
- list_add_tail(&cache->dirty_list,
- &trans->transaction->dirty_bgs);
- trans->delayed_ref_updates++;
- btrfs_get_block_group(cache);
+ start = sb_end;
+ if (start > end) {
+ bytes_left = 0;
+ break;
}
- spin_unlock(&trans->transaction->dirty_bgs_lock);
-
- /*
- * No longer have used bytes in this block group, queue it for
- * deletion. We do this after adding the block group to the
- * dirty list to avoid races between cleaner kthread and space
- * cache writeout.
- */
- if (!alloc && old_val == 0)
- btrfs_mark_bg_unused(cache);
-
- btrfs_put_block_group(cache);
- total -= num_bytes;
- bytenr += num_bytes;
+ bytes_left = end - start;
}
- /* Modified block groups are accounted for in the delayed_refs_rsv. */
- btrfs_update_delayed_refs_rsv(trans);
+ if (bytes_left) {
+ ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9,
+ GFP_NOFS, 0);
+ if (!ret)
+ *discarded_bytes += bytes_left;
+ }
return ret;
}
-static u64 first_logical_byte(struct btrfs_fs_info *fs_info, u64 search_start)
+int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
+ u64 num_bytes, u64 *actual_bytes)
{
- struct btrfs_block_group_cache *cache;
- u64 bytenr;
+ int ret;
+ u64 discarded_bytes = 0;
+ struct btrfs_bio *bbio = NULL;
- spin_lock(&fs_info->block_group_cache_lock);
- bytenr = fs_info->first_logical_byte;
- spin_unlock(&fs_info->block_group_cache_lock);
- if (bytenr < (u64)-1)
- return bytenr;
+ /*
+ * Avoid races with device replace and make sure our bbio has devices
+ * associated to its stripes that don't go away while we are discarding.
+ */
+ btrfs_bio_counter_inc_blocked(fs_info);
+ /* Tell the block device(s) that the sectors can be discarded */
+ ret = btrfs_map_block(fs_info, BTRFS_MAP_DISCARD, bytenr, &num_bytes,
+ &bbio, 0);
+ /* Error condition is -ENOMEM */
+ if (!ret) {
+ struct btrfs_bio_stripe *stripe = bbio->stripes;
+ int i;
- cache = btrfs_lookup_first_block_group(fs_info, search_start);
- if (!cache)
- return 0;
- bytenr = cache->key.objectid;
- btrfs_put_block_group(cache);
+ for (i = 0; i < bbio->num_stripes; i++, stripe++) {
+ u64 bytes;
+ struct request_queue *req_q;
- return bytenr;
-}
+ if (!stripe->dev->bdev) {
+ ASSERT(btrfs_test_opt(fs_info, DEGRADED));
+ continue;
+ }
+ req_q = bdev_get_queue(stripe->dev->bdev);
+ if (!blk_queue_discard(req_q))
+ continue;
-static int pin_down_extent(struct btrfs_block_group_cache *cache,
- u64 bytenr, u64 num_bytes, int reserved)
-{
- struct btrfs_fs_info *fs_info = cache->fs_info;
+ ret = btrfs_issue_discard(stripe->dev->bdev,
+ stripe->physical,
+ stripe->length,
+ &bytes);
+ if (!ret)
+ discarded_bytes += bytes;
+ else if (ret != -EOPNOTSUPP)
+ break; /* Logic errors or -ENOMEM, or -EIO but I don't know how that could happen JDM */
- spin_lock(&cache->space_info->lock);
- spin_lock(&cache->lock);
- cache->pinned += num_bytes;
- btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info,
- num_bytes);
- if (reserved) {
- cache->reserved -= num_bytes;
- cache->space_info->bytes_reserved -= num_bytes;
+ /*
+ * Just in case we get back EOPNOTSUPP for some reason,
+ * just ignore the return value so we don't screw up
+ * people calling discard_extent.
+ */
+ ret = 0;
+ }
+ btrfs_put_bbio(bbio);
}
- spin_unlock(&cache->lock);
- spin_unlock(&cache->space_info->lock);
+ btrfs_bio_counter_dec(fs_info);
- trace_btrfs_space_reservation(fs_info, "pinned",
- cache->space_info->flags, num_bytes, 1);
- percpu_counter_add_batch(&cache->space_info->total_bytes_pinned,
- num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH);
- set_extent_dirty(fs_info->pinned_extents, bytenr,
- bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
- return 0;
+ if (actual_bytes)
+ *actual_bytes = discarded_bytes;
+
+
+ if (ret == -EOPNOTSUPP)
+ ret = 0;
+ return ret;
}
-/*
- * this function must be called within transaction
- */
-int btrfs_pin_extent(struct btrfs_fs_info *fs_info,
- u64 bytenr, u64 num_bytes, int reserved)
+/* Can return -ENOMEM */
+int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_ref *generic_ref)
{
- struct btrfs_block_group_cache *cache;
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ int old_ref_mod, new_ref_mod;
+ int ret;
- cache = btrfs_lookup_block_group(fs_info, bytenr);
- BUG_ON(!cache); /* Logic error */
+ ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
+ generic_ref->action);
+ BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
+ generic_ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID);
- pin_down_extent(cache, bytenr, num_bytes, reserved);
+ if (generic_ref->type == BTRFS_REF_METADATA)
+ ret = btrfs_add_delayed_tree_ref(trans, generic_ref,
+ NULL, &old_ref_mod, &new_ref_mod);
+ else
+ ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0,
+ &old_ref_mod, &new_ref_mod);
- btrfs_put_block_group(cache);
- return 0;
+ btrfs_ref_tree_mod(fs_info, generic_ref);
+
+ if (ret == 0 && old_ref_mod < 0 && new_ref_mod >= 0)
+ sub_pinned_bytes(fs_info, generic_ref);
+
+ return ret;
}
/*
- * this function must be called within transaction
+ * __btrfs_inc_extent_ref - insert backreference for a given extent
+ *
+ * @trans: Handle of transaction
+ *
+ * @node: The delayed ref node used to get the bytenr/length for
+ * extent whose references are incremented.
+ *
+ * @parent: If this is a shared extent (BTRFS_SHARED_DATA_REF_KEY/
+ * BTRFS_SHARED_BLOCK_REF_KEY) then it holds the logical
+ * bytenr of the parent block. Since new extents are always
+ * created with indirect references, this will only be the case
+ * when relocating a shared extent. In that case, root_objectid
+ * will be BTRFS_TREE_RELOC_OBJECTID. Otheriwse, parent must
+ * be 0
+ *
+ * @root_objectid: The id of the root where this modification has originated,
+ * this can be either one of the well-known metadata trees or
+ * the subvolume id which references this extent.
+ *
+ * @owner: For data extents it is the inode number of the owning file.
+ * For metadata extents this parameter holds the level in the
+ * tree of the extent.
+ *
+ * @offset: For metadata extents the offset is ignored and is currently
+ * always passed as 0. For data extents it is the fileoffset
+ * this extent belongs to.
+ *
+ * @refs_to_add Number of references to add
+ *
+ * @extent_op Pointer to a structure, holding information necessary when
+ * updating a tree block's flags
+ *
*/
-int btrfs_pin_extent_for_log_replay(struct btrfs_fs_info *fs_info,
- u64 bytenr, u64 num_bytes)
+static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_node *node,
+ u64 parent, u64 root_objectid,
+ u64 owner, u64 offset, int refs_to_add,
+ struct btrfs_delayed_extent_op *extent_op)
{
- struct btrfs_block_group_cache *cache;
+ struct btrfs_path *path;
+ struct extent_buffer *leaf;
+ struct btrfs_extent_item *item;
+ struct btrfs_key key;
+ u64 bytenr = node->bytenr;
+ u64 num_bytes = node->num_bytes;
+ u64 refs;
int ret;
- cache = btrfs_lookup_block_group(fs_info, bytenr);
- if (!cache)
- return -EINVAL;
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ path->reada = READA_FORWARD;
+ path->leave_spinning = 1;
+ /* this will setup the path even if it fails to insert the back ref */
+ ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes,
+ parent, root_objectid, owner,
+ offset, refs_to_add, extent_op);
+ if ((ret < 0 && ret != -EAGAIN) || !ret)
+ goto out;
/*
- * pull in the free space cache (if any) so that our pin
- * removes the free space from the cache. We have load_only set
- * to one because the slow code to read in the free extents does check
- * the pinned extents.
+ * Ok we had -EAGAIN which means we didn't have space to insert and
+ * inline extent ref, so just update the reference count and add a
+ * normal backref.
*/
- cache_block_group(cache, 1);
+ leaf = path->nodes[0];
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+ item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+ refs = btrfs_extent_refs(leaf, item);
+ btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
+ if (extent_op)
+ __run_delayed_extent_op(extent_op, leaf, item);
- pin_down_extent(cache, bytenr, num_bytes, 0);
+ btrfs_mark_buffer_dirty(leaf);
+ btrfs_release_path(path);
- /* remove us from the free space cache (if we're there at all) */
- ret = btrfs_remove_free_space(cache, bytenr, num_bytes);
- btrfs_put_block_group(cache);
+ path->reada = READA_FORWARD;
+ path->leave_spinning = 1;
+ /* now insert the actual backref */
+ ret = insert_extent_backref(trans, path, bytenr, parent, root_objectid,
+ owner, offset, refs_to_add);
+ if (ret)
+ btrfs_abort_transaction(trans, ret);
+out:
+ btrfs_free_path(path);
return ret;
}
-static int __exclude_logged_extent(struct btrfs_fs_info *fs_info,
- u64 start, u64 num_bytes)
-{
- int ret;
- struct btrfs_block_group_cache *block_group;
- struct btrfs_caching_control *caching_ctl;
+static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_node *node,
+ struct btrfs_delayed_extent_op *extent_op,
+ int insert_reserved)
+{
+ int ret = 0;
+ struct btrfs_delayed_data_ref *ref;
+ struct btrfs_key ins;
+ u64 parent = 0;
+ u64 ref_root = 0;
+ u64 flags = 0;
+
+ ins.objectid = node->bytenr;
+ ins.offset = node->num_bytes;
+ ins.type = BTRFS_EXTENT_ITEM_KEY;
- block_group = btrfs_lookup_block_group(fs_info, start);
- if (!block_group)
- return -EINVAL;
+ ref = btrfs_delayed_node_to_data_ref(node);
+ trace_run_delayed_data_ref(trans->fs_info, node, ref, node->action);
- cache_block_group(block_group, 0);
- caching_ctl = get_caching_control(block_group);
+ if (node->type == BTRFS_SHARED_DATA_REF_KEY)
+ parent = ref->parent;
+ ref_root = ref->root;
- if (!caching_ctl) {
- /* Logic error */
- BUG_ON(!block_group_cache_done(block_group));
- ret = btrfs_remove_free_space(block_group, start, num_bytes);
+ if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
+ if (extent_op)
+ flags |= extent_op->flags_to_set;
+ ret = alloc_reserved_file_extent(trans, parent, ref_root,
+ flags, ref->objectid,
+ ref->offset, &ins,
+ node->ref_mod);
+ } else if (node->action == BTRFS_ADD_DELAYED_REF) {
+ ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
+ ref->objectid, ref->offset,
+ node->ref_mod, extent_op);
+ } else if (node->action == BTRFS_DROP_DELAYED_REF) {
+ ret = __btrfs_free_extent(trans, node, parent,
+ ref_root, ref->objectid,
+ ref->offset, node->ref_mod,
+ extent_op);
} else {
- mutex_lock(&caching_ctl->mutex);
+ BUG();
+ }
+ return ret;
+}
- if (start >= caching_ctl->progress) {
- ret = add_excluded_extent(fs_info, start, num_bytes);
- } else if (start + num_bytes <= caching_ctl->progress) {
- ret = btrfs_remove_free_space(block_group,
- start, num_bytes);
- } else {
- num_bytes = caching_ctl->progress - start;
- ret = btrfs_remove_free_space(block_group,
- start, num_bytes);
- if (ret)
- goto out_lock;
+static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
+ struct extent_buffer *leaf,
+ struct btrfs_extent_item *ei)
+{
+ u64 flags = btrfs_extent_flags(leaf, ei);
+ if (extent_op->update_flags) {
+ flags |= extent_op->flags_to_set;
+ btrfs_set_extent_flags(leaf, ei, flags);
+ }
- num_bytes = (start + num_bytes) -
- caching_ctl->progress;
- start = caching_ctl->progress;
- ret = add_excluded_extent(fs_info, start, num_bytes);
- }
-out_lock:
- mutex_unlock(&caching_ctl->mutex);
- put_caching_control(caching_ctl);
+ if (extent_op->update_key) {
+ struct btrfs_tree_block_info *bi;
+ BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
+ bi = (struct btrfs_tree_block_info *)(ei + 1);
+ btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
}
- btrfs_put_block_group(block_group);
- return ret;
}
-int btrfs_exclude_logged_extents(struct extent_buffer *eb)
+static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_head *head,
+ struct btrfs_delayed_extent_op *extent_op)
{
- struct btrfs_fs_info *fs_info = eb->fs_info;
- struct btrfs_file_extent_item *item;
+ struct btrfs_fs_info *fs_info = trans->fs_info;
struct btrfs_key key;
- int found_type;
- int i;
- int ret = 0;
+ struct btrfs_path *path;
+ struct btrfs_extent_item *ei;
+ struct extent_buffer *leaf;
+ u32 item_size;
+ int ret;
+ int err = 0;
+ int metadata = !extent_op->is_data;
- if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS))
+ if (trans->aborted)
return 0;
- for (i = 0; i < btrfs_header_nritems(eb); i++) {
- btrfs_item_key_to_cpu(eb, &key, i);
- if (key.type != BTRFS_EXTENT_DATA_KEY)
- continue;
- item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
- found_type = btrfs_file_extent_type(eb, item);
- if (found_type == BTRFS_FILE_EXTENT_INLINE)
- continue;
- if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
- continue;
- key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
- key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
- ret = __exclude_logged_extent(fs_info, key.objectid, key.offset);
- if (ret)
- break;
- }
+ if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+ metadata = 0;
- return ret;
-}
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
-static void
-btrfs_inc_block_group_reservations(struct btrfs_block_group_cache *bg)
-{
- atomic_inc(&bg->reservations);
-}
+ key.objectid = head->bytenr;
-void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
- const u64 start)
-{
- struct btrfs_block_group_cache *bg;
+ if (metadata) {
+ key.type = BTRFS_METADATA_ITEM_KEY;
+ key.offset = extent_op->level;
+ } else {
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = head->num_bytes;
+ }
- bg = btrfs_lookup_block_group(fs_info, start);
- ASSERT(bg);
- if (atomic_dec_and_test(&bg->reservations))
- wake_up_var(&bg->reservations);
- btrfs_put_block_group(bg);
-}
+again:
+ path->reada = READA_FORWARD;
+ path->leave_spinning = 1;
+ ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 1);
+ if (ret < 0) {
+ err = ret;
+ goto out;
+ }
+ if (ret > 0) {
+ if (metadata) {
+ if (path->slots[0] > 0) {
+ path->slots[0]--;
+ btrfs_item_key_to_cpu(path->nodes[0], &key,
+ path->slots[0]);
+ if (key.objectid == head->bytenr &&
+ key.type == BTRFS_EXTENT_ITEM_KEY &&
+ key.offset == head->num_bytes)
+ ret = 0;
+ }
+ if (ret > 0) {
+ btrfs_release_path(path);
+ metadata = 0;
-void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
-{
- struct btrfs_space_info *space_info = bg->space_info;
+ key.objectid = head->bytenr;
+ key.offset = head->num_bytes;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ goto again;
+ }
+ } else {
+ err = -EIO;
+ goto out;
+ }
+ }
- ASSERT(bg->ro);
+ leaf = path->nodes[0];
+ item_size = btrfs_item_size_nr(leaf, path->slots[0]);
- if (!(bg->flags & BTRFS_BLOCK_GROUP_DATA))
- return;
+ if (unlikely(item_size < sizeof(*ei))) {
+ err = -EINVAL;
+ btrfs_print_v0_err(fs_info);
+ btrfs_abort_transaction(trans, err);
+ goto out;
+ }
- /*
- * Our block group is read only but before we set it to read only,
- * some task might have had allocated an extent from it already, but it
- * has not yet created a respective ordered extent (and added it to a
- * root's list of ordered extents).
- * Therefore wait for any task currently allocating extents, since the
- * block group's reservations counter is incremented while a read lock
- * on the groups' semaphore is held and decremented after releasing
- * the read access on that semaphore and creating the ordered extent.
- */
- down_write(&space_info->groups_sem);
- up_write(&space_info->groups_sem);
+ ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+ __run_delayed_extent_op(extent_op, leaf, ei);
- wait_var_event(&bg->reservations, !atomic_read(&bg->reservations));
+ btrfs_mark_buffer_dirty(leaf);
+out:
+ btrfs_free_path(path);
+ return err;
}
-/**
- * btrfs_add_reserved_bytes - update the block_group and space info counters
- * @cache: The cache we are manipulating
- * @ram_bytes: The number of bytes of file content, and will be same to
- * @num_bytes except for the compress path.
- * @num_bytes: The number of bytes in question
- * @delalloc: The blocks are allocated for the delalloc write
- *
- * This is called by the allocator when it reserves space. If this is a
- * reservation and the block group has become read only we cannot make the
- * reservation and return -EAGAIN, otherwise this function always succeeds.
- */
-static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
- u64 ram_bytes, u64 num_bytes, int delalloc)
+static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_node *node,
+ struct btrfs_delayed_extent_op *extent_op,
+ int insert_reserved)
{
- struct btrfs_space_info *space_info = cache->space_info;
int ret = 0;
+ struct btrfs_delayed_tree_ref *ref;
+ u64 parent = 0;
+ u64 ref_root = 0;
- spin_lock(&space_info->lock);
- spin_lock(&cache->lock);
- if (cache->ro) {
- ret = -EAGAIN;
+ ref = btrfs_delayed_node_to_tree_ref(node);
+ trace_run_delayed_tree_ref(trans->fs_info, node, ref, node->action);
+
+ if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
+ parent = ref->parent;
+ ref_root = ref->root;
+
+ if (node->ref_mod != 1) {
+ btrfs_err(trans->fs_info,
+ "btree block(%llu) has %d references rather than 1: action %d ref_root %llu parent %llu",
+ node->bytenr, node->ref_mod, node->action, ref_root,
+ parent);
+ return -EIO;
+ }
+ if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
+ BUG_ON(!extent_op || !extent_op->update_flags);
+ ret = alloc_reserved_tree_block(trans, node, extent_op);
+ } else if (node->action == BTRFS_ADD_DELAYED_REF) {
+ ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
+ ref->level, 0, 1, extent_op);
+ } else if (node->action == BTRFS_DROP_DELAYED_REF) {
+ ret = __btrfs_free_extent(trans, node, parent, ref_root,
+ ref->level, 0, 1, extent_op);
} else {
- cache->reserved += num_bytes;
- space_info->bytes_reserved += num_bytes;
- btrfs_space_info_update_bytes_may_use(cache->fs_info,
- space_info, -ram_bytes);
- if (delalloc)
- cache->delalloc_bytes += num_bytes;
+ BUG();
}
- spin_unlock(&cache->lock);
- spin_unlock(&space_info->lock);
return ret;
}
-/**
- * btrfs_free_reserved_bytes - update the block_group and space info counters
- * @cache: The cache we are manipulating
- * @num_bytes: The number of bytes in question
- * @delalloc: The blocks are allocated for the delalloc write
- *
- * This is called by somebody who is freeing space that was never actually used
- * on disk. For example if you reserve some space for a new leaf in transaction
- * A and before transaction A commits you free that leaf, you call this with
- * reserve set to 0 in order to clear the reservation.
- */
-
-static void btrfs_free_reserved_bytes(struct btrfs_block_group_cache *cache,
- u64 num_bytes, int delalloc)
+/* helper function to actually process a single delayed ref entry */
+static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_node *node,
+ struct btrfs_delayed_extent_op *extent_op,
+ int insert_reserved)
{
- struct btrfs_space_info *space_info = cache->space_info;
+ int ret = 0;
- spin_lock(&space_info->lock);
- spin_lock(&cache->lock);
- if (cache->ro)
- space_info->bytes_readonly += num_bytes;
- cache->reserved -= num_bytes;
- space_info->bytes_reserved -= num_bytes;
- space_info->max_extent_size = 0;
+ if (trans->aborted) {
+ if (insert_reserved)
+ btrfs_pin_extent(trans->fs_info, node->bytenr,
+ node->num_bytes, 1);
+ return 0;
+ }
- if (delalloc)
- cache->delalloc_bytes -= num_bytes;
- spin_unlock(&cache->lock);
- spin_unlock(&space_info->lock);
-}
-void btrfs_prepare_extent_commit(struct btrfs_fs_info *fs_info)
-{
- struct btrfs_caching_control *next;
- struct btrfs_caching_control *caching_ctl;
- struct btrfs_block_group_cache *cache;
+ if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
+ node->type == BTRFS_SHARED_BLOCK_REF_KEY)
+ ret = run_delayed_tree_ref(trans, node, extent_op,
+ insert_reserved);
+ else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
+ node->type == BTRFS_SHARED_DATA_REF_KEY)
+ ret = run_delayed_data_ref(trans, node, extent_op,
+ insert_reserved);
+ else
+ BUG();
+ if (ret && insert_reserved)
+ btrfs_pin_extent(trans->fs_info, node->bytenr,
+ node->num_bytes, 1);
+ return ret;
+}
- down_write(&fs_info->commit_root_sem);
+static inline struct btrfs_delayed_ref_node *
+select_delayed_ref(struct btrfs_delayed_ref_head *head)
+{
+ struct btrfs_delayed_ref_node *ref;
- list_for_each_entry_safe(caching_ctl, next,
- &fs_info->caching_block_groups, list) {
- cache = caching_ctl->block_group;
- if (block_group_cache_done(cache)) {
- cache->last_byte_to_unpin = (u64)-1;
- list_del_init(&caching_ctl->list);
- put_caching_control(caching_ctl);
- } else {
- cache->last_byte_to_unpin = caching_ctl->progress;
- }
- }
+ if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
+ return NULL;
- if (fs_info->pinned_extents == &fs_info->freed_extents[0])
- fs_info->pinned_extents = &fs_info->freed_extents[1];
- else
- fs_info->pinned_extents = &fs_info->freed_extents[0];
+ /*
+ * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
+ * This is to prevent a ref count from going down to zero, which deletes
+ * the extent item from the extent tree, when there still are references
+ * to add, which would fail because they would not find the extent item.
+ */
+ if (!list_empty(&head->ref_add_list))
+ return list_first_entry(&head->ref_add_list,
+ struct btrfs_delayed_ref_node, add_list);
- up_write(&fs_info->commit_root_sem);
+ ref = rb_entry(rb_first_cached(&head->ref_tree),
+ struct btrfs_delayed_ref_node, ref_node);
+ ASSERT(list_empty(&ref->add_list));
+ return ref;
+}
- btrfs_update_global_block_rsv(fs_info);
+static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
+ struct btrfs_delayed_ref_head *head)
+{
+ spin_lock(&delayed_refs->lock);
+ head->processing = 0;
+ delayed_refs->num_heads_ready++;
+ spin_unlock(&delayed_refs->lock);
+ btrfs_delayed_ref_unlock(head);
}
-/*
- * Returns the free cluster for the given space info and sets empty_cluster to
- * what it should be based on the mount options.
- */
-static struct btrfs_free_cluster *
-fetch_cluster_info(struct btrfs_fs_info *fs_info,
- struct btrfs_space_info *space_info, u64 *empty_cluster)
+static struct btrfs_delayed_extent_op *cleanup_extent_op(
+ struct btrfs_delayed_ref_head *head)
{
- struct btrfs_free_cluster *ret = NULL;
+ struct btrfs_delayed_extent_op *extent_op = head->extent_op;
- *empty_cluster = 0;
- if (btrfs_mixed_space_info(space_info))
- return ret;
+ if (!extent_op)
+ return NULL;
- if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
- ret = &fs_info->meta_alloc_cluster;
- if (btrfs_test_opt(fs_info, SSD))
- *empty_cluster = SZ_2M;
- else
- *empty_cluster = SZ_64K;
- } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
- btrfs_test_opt(fs_info, SSD_SPREAD)) {
- *empty_cluster = SZ_2M;
- ret = &fs_info->data_alloc_cluster;
+ if (head->must_insert_reserved) {
+ head->extent_op = NULL;
+ btrfs_free_delayed_extent_op(extent_op);
+ return NULL;
}
-
- return ret;
+ return extent_op;
}
-static int unpin_extent_range(struct btrfs_fs_info *fs_info,
- u64 start, u64 end,
- const bool return_free_space)
+static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_head *head)
{
- struct btrfs_block_group_cache *cache = NULL;
- struct btrfs_space_info *space_info;
- struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
- struct btrfs_free_cluster *cluster = NULL;
- u64 len;
- u64 total_unpinned = 0;
- u64 empty_cluster = 0;
- bool readonly;
-
- while (start <= end) {
- readonly = false;
- if (!cache ||
- start >= cache->key.objectid + cache->key.offset) {
- if (cache)
- btrfs_put_block_group(cache);
- total_unpinned = 0;
- cache = btrfs_lookup_block_group(fs_info, start);
- BUG_ON(!cache); /* Logic error */
+ struct btrfs_delayed_extent_op *extent_op;
+ int ret;
- cluster = fetch_cluster_info(fs_info,
- cache->space_info,
- &empty_cluster);
- empty_cluster <<= 1;
- }
+ extent_op = cleanup_extent_op(head);
+ if (!extent_op)
+ return 0;
+ head->extent_op = NULL;
+ spin_unlock(&head->lock);
+ ret = run_delayed_extent_op(trans, head, extent_op);
+ btrfs_free_delayed_extent_op(extent_op);
+ return ret ? ret : 1;
+}
- len = cache->key.objectid + cache->key.offset - start;
- len = min(len, end + 1 - start);
+void btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info,
+ struct btrfs_delayed_ref_root *delayed_refs,
+ struct btrfs_delayed_ref_head *head)
+{
+ int nr_items = 1; /* Dropping this ref head update. */
- if (start < cache->last_byte_to_unpin) {
- len = min(len, cache->last_byte_to_unpin - start);
- if (return_free_space)
- btrfs_add_free_space(cache, start, len);
- }
+ if (head->total_ref_mod < 0) {
+ struct btrfs_space_info *space_info;
+ u64 flags;
- start += len;
- total_unpinned += len;
- space_info = cache->space_info;
+ if (head->is_data)
+ flags = BTRFS_BLOCK_GROUP_DATA;
+ else if (head->is_system)
+ flags = BTRFS_BLOCK_GROUP_SYSTEM;
+ else
+ flags = BTRFS_BLOCK_GROUP_METADATA;
+ space_info = btrfs_find_space_info(fs_info, flags);
+ ASSERT(space_info);
+ percpu_counter_add_batch(&space_info->total_bytes_pinned,
+ -head->num_bytes,
+ BTRFS_TOTAL_BYTES_PINNED_BATCH);
/*
- * If this space cluster has been marked as fragmented and we've
- * unpinned enough in this block group to potentially allow a
- * cluster to be created inside of it go ahead and clear the
- * fragmented check.
+ * We had csum deletions accounted for in our delayed refs rsv,
+ * we need to drop the csum leaves for this update from our
+ * delayed_refs_rsv.
*/
- if (cluster && cluster->fragmented &&
- total_unpinned > empty_cluster) {
- spin_lock(&cluster->lock);
- cluster->fragmented = 0;
- spin_unlock(&cluster->lock);
- }
-
- spin_lock(&space_info->lock);
- spin_lock(&cache->lock);
- cache->pinned -= len;
- btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len);
-
- trace_btrfs_space_reservation(fs_info, "pinned",
- space_info->flags, len, 0);
- space_info->max_extent_size = 0;
- percpu_counter_add_batch(&space_info->total_bytes_pinned,
- -len, BTRFS_TOTAL_BYTES_PINNED_BATCH);
- if (cache->ro) {
- space_info->bytes_readonly += len;
- readonly = true;
- }
- spin_unlock(&cache->lock);
- if (!readonly && return_free_space &&
- global_rsv->space_info == space_info) {
- u64 to_add = len;
-
- spin_lock(&global_rsv->lock);
- if (!global_rsv->full) {
- to_add = min(len, global_rsv->size -
- global_rsv->reserved);
- global_rsv->reserved += to_add;
- btrfs_space_info_update_bytes_may_use(fs_info,
- space_info, to_add);
- if (global_rsv->reserved >= global_rsv->size)
- global_rsv->full = 1;
- trace_btrfs_space_reservation(fs_info,
- "space_info",
- space_info->flags,
- to_add, 1);
- len -= to_add;
- }
- spin_unlock(&global_rsv->lock);
- /* Add to any tickets we may have */
- if (len)
- btrfs_space_info_add_new_bytes(fs_info,
- space_info, len);
+ if (head->is_data) {
+ spin_lock(&delayed_refs->lock);
+ delayed_refs->pending_csums -= head->num_bytes;
+ spin_unlock(&delayed_refs->lock);
+ nr_items += btrfs_csum_bytes_to_leaves(fs_info,
+ head->num_bytes);
}
- spin_unlock(&space_info->lock);
}
- if (cache)
- btrfs_put_block_group(cache);
- return 0;
+ btrfs_delayed_refs_rsv_release(fs_info, nr_items);
}
-int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
+static int cleanup_ref_head(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_head *head)
{
+
struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_block_group_cache *block_group, *tmp;
- struct list_head *deleted_bgs;
- struct extent_io_tree *unpin;
- u64 start;
- u64 end;
+ struct btrfs_delayed_ref_root *delayed_refs;
int ret;
- if (fs_info->pinned_extents == &fs_info->freed_extents[0])
- unpin = &fs_info->freed_extents[1];
- else
- unpin = &fs_info->freed_extents[0];
-
- while (!trans->aborted) {
- struct extent_state *cached_state = NULL;
-
- mutex_lock(&fs_info->unused_bg_unpin_mutex);
- ret = find_first_extent_bit(unpin, 0, &start, &end,
- EXTENT_DIRTY, &cached_state);
- if (ret) {
- mutex_unlock(&fs_info->unused_bg_unpin_mutex);
- break;
- }
-
- if (btrfs_test_opt(fs_info, DISCARD))
- ret = btrfs_discard_extent(fs_info, start,
- end + 1 - start, NULL);
-
- clear_extent_dirty(unpin, start, end, &cached_state);
- unpin_extent_range(fs_info, start, end, true);
- mutex_unlock(&fs_info->unused_bg_unpin_mutex);
- free_extent_state(cached_state);
- cond_resched();
- }
-
- /*
- * Transaction is finished. We don't need the lock anymore. We
- * do need to clean up the block groups in case of a transaction
- * abort.
- */
- deleted_bgs = &trans->transaction->deleted_bgs;
- list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) {
- u64 trimmed = 0;
+ delayed_refs = &trans->transaction->delayed_refs;
- ret = -EROFS;
- if (!trans->aborted)
- ret = btrfs_discard_extent(fs_info,
- block_group->key.objectid,
- block_group->key.offset,
- &trimmed);
+ ret = run_and_cleanup_extent_op(trans, head);
+ if (ret < 0) {
+ unselect_delayed_ref_head(delayed_refs, head);
+ btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
+ return ret;
+ } else if (ret) {
+ return ret;
+ }
- list_del_init(&block_group->bg_list);
- btrfs_put_block_group_trimming(block_group);
- btrfs_put_block_group(block_group);
+ /*
+ * Need to drop our head ref lock and re-acquire the delayed ref lock
+ * and then re-check to make sure nobody got added.
+ */
+ spin_unlock(&head->lock);
+ spin_lock(&delayed_refs->lock);
+ spin_lock(&head->lock);
+ if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) {
+ spin_unlock(&head->lock);
+ spin_unlock(&delayed_refs->lock);
+ return 1;
+ }
+ btrfs_delete_ref_head(delayed_refs, head);
+ spin_unlock(&head->lock);
+ spin_unlock(&delayed_refs->lock);
- if (ret) {
- const char *errstr = btrfs_decode_error(ret);
- btrfs_warn(fs_info,
- "discard failed while removing blockgroup: errno=%d %s",
- ret, errstr);
+ if (head->must_insert_reserved) {
+ btrfs_pin_extent(fs_info, head->bytenr,
+ head->num_bytes, 1);
+ if (head->is_data) {
+ ret = btrfs_del_csums(trans, fs_info, head->bytenr,
+ head->num_bytes);
}
}
+ btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
+
+ trace_run_delayed_ref_head(fs_info, head, 0);
+ btrfs_delayed_ref_unlock(head);
+ btrfs_put_delayed_ref_head(head);
return 0;
}
-static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_node *node, u64 parent,
- u64 root_objectid, u64 owner_objectid,
- u64 owner_offset, int refs_to_drop,
- struct btrfs_delayed_extent_op *extent_op)
+static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head(
+ struct btrfs_trans_handle *trans)
{
- struct btrfs_fs_info *info = trans->fs_info;
- struct btrfs_key key;
- struct btrfs_path *path;
- struct btrfs_root *extent_root = info->extent_root;
- struct extent_buffer *leaf;
- struct btrfs_extent_item *ei;
- struct btrfs_extent_inline_ref *iref;
+ struct btrfs_delayed_ref_root *delayed_refs =
+ &trans->transaction->delayed_refs;
+ struct btrfs_delayed_ref_head *head = NULL;
int ret;
- int is_data;
- int extent_slot = 0;
- int found_extent = 0;
- int num_to_del = 1;
- u32 item_size;
- u64 refs;
- u64 bytenr = node->bytenr;
- u64 num_bytes = node->num_bytes;
- int last_ref = 0;
- bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA);
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
-
- path->reada = READA_FORWARD;
- path->leave_spinning = 1;
-
- is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
- BUG_ON(!is_data && refs_to_drop != 1);
- if (is_data)
- skinny_metadata = false;
+ spin_lock(&delayed_refs->lock);
+ head = btrfs_select_ref_head(delayed_refs);
+ if (!head) {
+ spin_unlock(&delayed_refs->lock);
+ return head;
+ }
- ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes,
- parent, root_objectid, owner_objectid,
- owner_offset);
- if (ret == 0) {
- extent_slot = path->slots[0];
- while (extent_slot >= 0) {
- btrfs_item_key_to_cpu(path->nodes[0], &key,
- extent_slot);
- if (key.objectid != bytenr)
- break;
- if (key.type == BTRFS_EXTENT_ITEM_KEY &&
- key.offset == num_bytes) {
- found_extent = 1;
- break;
- }
- if (key.type == BTRFS_METADATA_ITEM_KEY &&
- key.offset == owner_objectid) {
- found_extent = 1;
- break;
- }
- if (path->slots[0] - extent_slot > 5)
- break;
- extent_slot--;
- }
+ /*
+ * Grab the lock that says we are going to process all the refs for
+ * this head
+ */
+ ret = btrfs_delayed_ref_lock(delayed_refs, head);
+ spin_unlock(&delayed_refs->lock);
- if (!found_extent) {
- BUG_ON(iref);
- ret = remove_extent_backref(trans, path, NULL,
- refs_to_drop,
- is_data, &last_ref);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- btrfs_release_path(path);
- path->leave_spinning = 1;
+ /*
+ * We may have dropped the spin lock to get the head mutex lock, and
+ * that might have given someone else time to free the head. If that's
+ * true, it has been removed from our list and we can move on.
+ */
+ if (ret == -EAGAIN)
+ head = ERR_PTR(-EAGAIN);
- key.objectid = bytenr;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- key.offset = num_bytes;
+ return head;
+}
- if (!is_data && skinny_metadata) {
- key.type = BTRFS_METADATA_ITEM_KEY;
- key.offset = owner_objectid;
- }
+static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_head *locked_ref,
+ unsigned long *run_refs)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_delayed_ref_root *delayed_refs;
+ struct btrfs_delayed_extent_op *extent_op;
+ struct btrfs_delayed_ref_node *ref;
+ int must_insert_reserved = 0;
+ int ret;
- ret = btrfs_search_slot(trans, extent_root,
- &key, path, -1, 1);
- if (ret > 0 && skinny_metadata && path->slots[0]) {
- /*
- * Couldn't find our skinny metadata item,
- * see if we have ye olde extent item.
- */
- path->slots[0]--;
- btrfs_item_key_to_cpu(path->nodes[0], &key,
- path->slots[0]);
- if (key.objectid == bytenr &&
- key.type == BTRFS_EXTENT_ITEM_KEY &&
- key.offset == num_bytes)
- ret = 0;
- }
+ delayed_refs = &trans->transaction->delayed_refs;
- if (ret > 0 && skinny_metadata) {
- skinny_metadata = false;
- key.objectid = bytenr;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- key.offset = num_bytes;
- btrfs_release_path(path);
- ret = btrfs_search_slot(trans, extent_root,
- &key, path, -1, 1);
- }
+ lockdep_assert_held(&locked_ref->mutex);
+ lockdep_assert_held(&locked_ref->lock);
- if (ret) {
- btrfs_err(info,
- "umm, got %d back from search, was looking for %llu",
- ret, bytenr);
- if (ret > 0)
- btrfs_print_leaf(path->nodes[0]);
- }
- if (ret < 0) {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- extent_slot = path->slots[0];
+ while ((ref = select_delayed_ref(locked_ref))) {
+ if (ref->seq &&
+ btrfs_check_delayed_seq(fs_info, ref->seq)) {
+ spin_unlock(&locked_ref->lock);
+ unselect_delayed_ref_head(delayed_refs, locked_ref);
+ return -EAGAIN;
}
- } else if (WARN_ON(ret == -ENOENT)) {
- btrfs_print_leaf(path->nodes[0]);
- btrfs_err(info,
- "unable to find ref byte nr %llu parent %llu root %llu owner %llu offset %llu",
- bytenr, parent, root_objectid, owner_objectid,
- owner_offset);
- btrfs_abort_transaction(trans, ret);
- goto out;
- } else {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
-
- leaf = path->nodes[0];
- item_size = btrfs_item_size_nr(leaf, extent_slot);
- if (unlikely(item_size < sizeof(*ei))) {
- ret = -EINVAL;
- btrfs_print_v0_err(info);
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- ei = btrfs_item_ptr(leaf, extent_slot,
- struct btrfs_extent_item);
- if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
- key.type == BTRFS_EXTENT_ITEM_KEY) {
- struct btrfs_tree_block_info *bi;
- BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
- bi = (struct btrfs_tree_block_info *)(ei + 1);
- WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
- }
- refs = btrfs_extent_refs(leaf, ei);
- if (refs < refs_to_drop) {
- btrfs_err(info,
- "trying to drop %d refs but we only have %Lu for bytenr %Lu",
- refs_to_drop, refs, bytenr);
- ret = -EINVAL;
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- refs -= refs_to_drop;
+ (*run_refs)++;
+ ref->in_tree = 0;
+ rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree);
+ RB_CLEAR_NODE(&ref->ref_node);
+ if (!list_empty(&ref->add_list))
+ list_del(&ref->add_list);
+ /*
+ * When we play the delayed ref, also correct the ref_mod on
+ * head
+ */
+ switch (ref->action) {
+ case BTRFS_ADD_DELAYED_REF:
+ case BTRFS_ADD_DELAYED_EXTENT:
+ locked_ref->ref_mod -= ref->ref_mod;
+ break;
+ case BTRFS_DROP_DELAYED_REF:
+ locked_ref->ref_mod += ref->ref_mod;
+ break;
+ default:
+ WARN_ON(1);
+ }
+ atomic_dec(&delayed_refs->num_entries);
- if (refs > 0) {
- if (extent_op)
- __run_delayed_extent_op(extent_op, leaf, ei);
/*
- * In the case of inline back ref, reference count will
- * be updated by remove_extent_backref
+ * Record the must_insert_reserved flag before we drop the
+ * spin lock.
*/
- if (iref) {
- BUG_ON(!found_extent);
- } else {
- btrfs_set_extent_refs(leaf, ei, refs);
- btrfs_mark_buffer_dirty(leaf);
- }
- if (found_extent) {
- ret = remove_extent_backref(trans, path, iref,
- refs_to_drop, is_data,
- &last_ref);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- }
- } else {
- if (found_extent) {
- BUG_ON(is_data && refs_to_drop !=
- extent_data_ref_count(path, iref));
- if (iref) {
- BUG_ON(path->slots[0] != extent_slot);
- } else {
- BUG_ON(path->slots[0] != extent_slot + 1);
- path->slots[0] = extent_slot;
- num_to_del = 2;
- }
- }
+ must_insert_reserved = locked_ref->must_insert_reserved;
+ locked_ref->must_insert_reserved = 0;
- last_ref = 1;
- ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
- num_to_del);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- btrfs_release_path(path);
+ extent_op = locked_ref->extent_op;
+ locked_ref->extent_op = NULL;
+ spin_unlock(&locked_ref->lock);
- if (is_data) {
- ret = btrfs_del_csums(trans, info, bytenr, num_bytes);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
- }
+ ret = run_one_delayed_ref(trans, ref, extent_op,
+ must_insert_reserved);
- ret = add_to_free_space_tree(trans, bytenr, num_bytes);
+ btrfs_free_delayed_extent_op(extent_op);
if (ret) {
- btrfs_abort_transaction(trans, ret);
- goto out;
+ unselect_delayed_ref_head(delayed_refs, locked_ref);
+ btrfs_put_delayed_ref(ref);
+ btrfs_debug(fs_info, "run_one_delayed_ref returned %d",
+ ret);
+ return ret;
}
- ret = update_block_group(trans, bytenr, num_bytes, 0);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- goto out;
- }
+ btrfs_put_delayed_ref(ref);
+ cond_resched();
+
+ spin_lock(&locked_ref->lock);
+ btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
}
- btrfs_release_path(path);
-out:
- btrfs_free_path(path);
- return ret;
+ return 0;
}
/*
- * when we free an block, it is possible (and likely) that we free the last
- * delayed ref for that extent as well. This searches the delayed ref tree for
- * a given extent, and if there are no other delayed refs to be processed, it
- * removes it from the tree.
+ * Returns 0 on success or if called with an already aborted transaction.
+ * Returns -ENOMEM or -EIO on failure and will abort the transaction.
*/
-static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
- u64 bytenr)
+static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
+ unsigned long nr)
{
- struct btrfs_delayed_ref_head *head;
+ struct btrfs_fs_info *fs_info = trans->fs_info;
struct btrfs_delayed_ref_root *delayed_refs;
- int ret = 0;
+ struct btrfs_delayed_ref_head *locked_ref = NULL;
+ ktime_t start = ktime_get();
+ int ret;
+ unsigned long count = 0;
+ unsigned long actual_count = 0;
delayed_refs = &trans->transaction->delayed_refs;
- spin_lock(&delayed_refs->lock);
- head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
- if (!head)
- goto out_delayed_unlock;
+ do {
+ if (!locked_ref) {
+ locked_ref = btrfs_obtain_ref_head(trans);
+ if (IS_ERR_OR_NULL(locked_ref)) {
+ if (PTR_ERR(locked_ref) == -EAGAIN) {
+ continue;
+ } else {
+ break;
+ }
+ }
+ count++;
+ }
+ /*
+ * We need to try and merge add/drops of the same ref since we
+ * can run into issues with relocate dropping the implicit ref
+ * and then it being added back again before the drop can
+ * finish. If we merged anything we need to re-loop so we can
+ * get a good ref.
+ * Or we can get node references of the same type that weren't
+ * merged when created due to bumps in the tree mod seq, and
+ * we need to merge them to prevent adding an inline extent
+ * backref before dropping it (triggering a BUG_ON at
+ * insert_inline_extent_backref()).
+ */
+ spin_lock(&locked_ref->lock);
+ btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
- spin_lock(&head->lock);
- if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root))
- goto out;
+ ret = btrfs_run_delayed_refs_for_head(trans, locked_ref,
+ &actual_count);
+ if (ret < 0 && ret != -EAGAIN) {
+ /*
+ * Error, btrfs_run_delayed_refs_for_head already
+ * unlocked everything so just bail out
+ */
+ return ret;
+ } else if (!ret) {
+ /*
+ * Success, perform the usual cleanup of a processed
+ * head
+ */
+ ret = cleanup_ref_head(trans, locked_ref);
+ if (ret > 0 ) {
+ /* We dropped our lock, we need to loop. */
+ ret = 0;
+ continue;
+ } else if (ret) {
+ return ret;
+ }
+ }
- if (cleanup_extent_op(head) != NULL)
- goto out;
+ /*
+ * Either success case or btrfs_run_delayed_refs_for_head
+ * returned -EAGAIN, meaning we need to select another head
+ */
+
+ locked_ref = NULL;
+ cond_resched();
+ } while ((nr != -1 && count < nr) || locked_ref);
/*
- * waiting for the lock here would deadlock. If someone else has it
- * locked they are already in the process of dropping it anyway
+ * We don't want to include ref heads since we can have empty ref heads
+ * and those will drastically skew our runtime down since we just do
+ * accounting, no actual extent tree updates.
*/
- if (!mutex_trylock(&head->mutex))
- goto out;
-
- btrfs_delete_ref_head(delayed_refs, head);
- head->processing = 0;
-
- spin_unlock(&head->lock);
- spin_unlock(&delayed_refs->lock);
-
- BUG_ON(head->extent_op);
- if (head->must_insert_reserved)
- ret = 1;
-
- btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head);
- mutex_unlock(&head->mutex);
- btrfs_put_delayed_ref_head(head);
- return ret;
-out:
- spin_unlock(&head->lock);
+ if (actual_count > 0) {
+ u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
+ u64 avg;
-out_delayed_unlock:
- spin_unlock(&delayed_refs->lock);
+ /*
+ * We weigh the current average higher than our current runtime
+ * to avoid large swings in the average.
+ */
+ spin_lock(&delayed_refs->lock);
+ avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
+ fs_info->avg_delayed_ref_runtime = avg >> 2; /* div by 4 */
+ spin_unlock(&delayed_refs->lock);
+ }
return 0;
}
-void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct extent_buffer *buf,
- u64 parent, int last_ref)
+#ifdef SCRAMBLE_DELAYED_REFS
+/*
+ * Normally delayed refs get processed in ascending bytenr order. This
+ * correlates in most cases to the order added. To expose dependencies on this
+ * order, we start to process the tree in the middle instead of the beginning
+ */
+static u64 find_middle(struct rb_root *root)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- struct btrfs_ref generic_ref = { 0 };
- int pin = 1;
- int ret;
-
- btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF,
- buf->start, buf->len, parent);
- btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf),
- root->root_key.objectid);
-
- if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
- int old_ref_mod, new_ref_mod;
+ struct rb_node *n = root->rb_node;
+ struct btrfs_delayed_ref_node *entry;
+ int alt = 1;
+ u64 middle;
+ u64 first = 0, last = 0;
- btrfs_ref_tree_mod(fs_info, &generic_ref);
- ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL,
- &old_ref_mod, &new_ref_mod);
- BUG_ON(ret); /* -ENOMEM */
- pin = old_ref_mod >= 0 && new_ref_mod < 0;
+ n = rb_first(root);
+ if (n) {
+ entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
+ first = entry->bytenr;
}
+ n = rb_last(root);
+ if (n) {
+ entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
+ last = entry->bytenr;
+ }
+ n = root->rb_node;
- if (last_ref && btrfs_header_generation(buf) == trans->transid) {
- struct btrfs_block_group_cache *cache;
-
- if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
- ret = check_ref_cleanup(trans, buf->start);
- if (!ret)
- goto out;
- }
-
- pin = 0;
- cache = btrfs_lookup_block_group(fs_info, buf->start);
-
- if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
- pin_down_extent(cache, buf->start, buf->len, 1);
- btrfs_put_block_group(cache);
- goto out;
- }
+ while (n) {
+ entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
+ WARN_ON(!entry->in_tree);
- WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
+ middle = entry->bytenr;
- btrfs_add_free_space(cache, buf->start, buf->len);
- btrfs_free_reserved_bytes(cache, buf->len, 0);
- btrfs_put_block_group(cache);
- trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len);
- }
-out:
- if (pin)
- add_pinned_bytes(fs_info, &generic_ref);
+ if (alt)
+ n = n->rb_left;
+ else
+ n = n->rb_right;
- if (last_ref) {
- /*
- * Deleting the buffer, clear the corrupt flag since it doesn't
- * matter anymore.
- */
- clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
+ alt = 1 - alt;
}
+ return middle;
}
+#endif
-/* Can return -ENOMEM */
-int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
+static inline u64 heads_to_leaves(struct btrfs_fs_info *fs_info, u64 heads)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- int old_ref_mod, new_ref_mod;
- int ret;
+ u64 num_bytes;
- if (btrfs_is_testing(fs_info))
- return 0;
+ num_bytes = heads * (sizeof(struct btrfs_extent_item) +
+ sizeof(struct btrfs_extent_inline_ref));
+ if (!btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+ num_bytes += heads * sizeof(struct btrfs_tree_block_info);
/*
- * tree log blocks never actually go into the extent allocation
- * tree, just update pinning info and exit early.
+ * We don't ever fill up leaves all the way so multiply by 2 just to be
+ * closer to what we're really going to want to use.
*/
- if ((ref->type == BTRFS_REF_METADATA &&
- ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
- (ref->type == BTRFS_REF_DATA &&
- ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)) {
- /* unlocks the pinned mutex */
- btrfs_pin_extent(fs_info, ref->bytenr, ref->len, 1);
- old_ref_mod = new_ref_mod = 0;
- ret = 0;
- } else if (ref->type == BTRFS_REF_METADATA) {
- ret = btrfs_add_delayed_tree_ref(trans, ref, NULL,
- &old_ref_mod, &new_ref_mod);
- } else {
- ret = btrfs_add_delayed_data_ref(trans, ref, 0,
- &old_ref_mod, &new_ref_mod);
- }
-
- if (!((ref->type == BTRFS_REF_METADATA &&
- ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
- (ref->type == BTRFS_REF_DATA &&
- ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)))
- btrfs_ref_tree_mod(fs_info, ref);
+ return div_u64(num_bytes, BTRFS_LEAF_DATA_SIZE(fs_info));
+}
- if (ret == 0 && old_ref_mod >= 0 && new_ref_mod < 0)
- add_pinned_bytes(fs_info, ref);
+/*
+ * Takes the number of bytes to be csumm'ed and figures out how many leaves it
+ * would require to store the csums for that many bytes.
+ */
+u64 btrfs_csum_bytes_to_leaves(struct btrfs_fs_info *fs_info, u64 csum_bytes)
+{
+ u64 csum_size;
+ u64 num_csums_per_leaf;
+ u64 num_csums;
- return ret;
+ csum_size = BTRFS_MAX_ITEM_SIZE(fs_info);
+ num_csums_per_leaf = div64_u64(csum_size,
+ (u64)btrfs_super_csum_size(fs_info->super_copy));
+ num_csums = div64_u64(csum_bytes, fs_info->sectorsize);
+ num_csums += num_csums_per_leaf - 1;
+ num_csums = div64_u64(num_csums, num_csums_per_leaf);
+ return num_csums;
}
/*
- * when we wait for progress in the block group caching, its because
- * our allocation attempt failed at least once. So, we must sleep
- * and let some progress happen before we try again.
- *
- * This function will sleep at least once waiting for new free space to
- * show up, and then it will check the block group free space numbers
- * for our min num_bytes. Another option is to have it go ahead
- * and look in the rbtree for a free extent of a given size, but this
- * is a good start.
+ * this starts processing the delayed reference count updates and
+ * extent insertions we have queued up so far. count can be
+ * 0, which means to process everything in the tree at the start
+ * of the run (but not newly added entries), or it can be some target
+ * number you'd like to process.
*
- * Callers of this must check if cache->cached == BTRFS_CACHE_ERROR before using
- * any of the information in this block group.
+ * Returns 0 on success or if called with an aborted transaction
+ * Returns <0 on error and aborts the transaction
*/
-static noinline void
-wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
- u64 num_bytes)
+int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
+ unsigned long count)
{
- struct btrfs_caching_control *caching_ctl;
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct rb_node *node;
+ struct btrfs_delayed_ref_root *delayed_refs;
+ struct btrfs_delayed_ref_head *head;
+ int ret;
+ int run_all = count == (unsigned long)-1;
- caching_ctl = get_caching_control(cache);
- if (!caching_ctl)
- return;
+ /* We'll clean this up in btrfs_cleanup_transaction */
+ if (trans->aborted)
+ return 0;
- wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
- (cache->free_space_ctl->free_space >= num_bytes));
+ if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
+ return 0;
- put_caching_control(caching_ctl);
-}
+ delayed_refs = &trans->transaction->delayed_refs;
+ if (count == 0)
+ count = atomic_read(&delayed_refs->num_entries) * 2;
-static noinline int
-wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
-{
- struct btrfs_caching_control *caching_ctl;
- int ret = 0;
+again:
+#ifdef SCRAMBLE_DELAYED_REFS
+ delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
+#endif
+ ret = __btrfs_run_delayed_refs(trans, count);
+ if (ret < 0) {
+ btrfs_abort_transaction(trans, ret);
+ return ret;
+ }
- caching_ctl = get_caching_control(cache);
- if (!caching_ctl)
- return (cache->cached == BTRFS_CACHE_ERROR) ? -EIO : 0;
+ if (run_all) {
+ btrfs_create_pending_block_groups(trans);
- wait_event(caching_ctl->wait, block_group_cache_done(cache));
- if (cache->cached == BTRFS_CACHE_ERROR)
- ret = -EIO;
- put_caching_control(caching_ctl);
- return ret;
-}
+ spin_lock(&delayed_refs->lock);
+ node = rb_first_cached(&delayed_refs->href_root);
+ if (!node) {
+ spin_unlock(&delayed_refs->lock);
+ goto out;
+ }
+ head = rb_entry(node, struct btrfs_delayed_ref_head,
+ href_node);
+ refcount_inc(&head->refs);
+ spin_unlock(&delayed_refs->lock);
-enum btrfs_loop_type {
- LOOP_CACHING_NOWAIT,
- LOOP_CACHING_WAIT,
- LOOP_ALLOC_CHUNK,
- LOOP_NO_EMPTY_SIZE,
-};
+ /* Mutex was contended, block until it's released and retry. */
+ mutex_lock(&head->mutex);
+ mutex_unlock(&head->mutex);
-static inline void
-btrfs_lock_block_group(struct btrfs_block_group_cache *cache,
- int delalloc)
-{
- if (delalloc)
- down_read(&cache->data_rwsem);
+ btrfs_put_delayed_ref_head(head);
+ cond_resched();
+ goto again;
+ }
+out:
+ return 0;
}
-static inline void
-btrfs_grab_block_group(struct btrfs_block_group_cache *cache,
- int delalloc)
+int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
+ u64 bytenr, u64 num_bytes, u64 flags,
+ int level, int is_data)
{
- btrfs_get_block_group(cache);
- if (delalloc)
- down_read(&cache->data_rwsem);
+ struct btrfs_delayed_extent_op *extent_op;
+ int ret;
+
+ extent_op = btrfs_alloc_delayed_extent_op();
+ if (!extent_op)
+ return -ENOMEM;
+
+ extent_op->flags_to_set = flags;
+ extent_op->update_flags = true;
+ extent_op->update_key = false;
+ extent_op->is_data = is_data ? true : false;
+ extent_op->level = level;
+
+ ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
+ if (ret)
+ btrfs_free_delayed_extent_op(extent_op);
+ return ret;
}
-static struct btrfs_block_group_cache *
-btrfs_lock_cluster(struct btrfs_block_group_cache *block_group,
- struct btrfs_free_cluster *cluster,
- int delalloc)
+static noinline int check_delayed_ref(struct btrfs_root *root,
+ struct btrfs_path *path,
+ u64 objectid, u64 offset, u64 bytenr)
{
- struct btrfs_block_group_cache *used_bg = NULL;
-
- spin_lock(&cluster->refill_lock);
- while (1) {
- used_bg = cluster->block_group;
- if (!used_bg)
- return NULL;
+ struct btrfs_delayed_ref_head *head;
+ struct btrfs_delayed_ref_node *ref;
+ struct btrfs_delayed_data_ref *data_ref;
+ struct btrfs_delayed_ref_root *delayed_refs;
+ struct btrfs_transaction *cur_trans;
+ struct rb_node *node;
+ int ret = 0;
- if (used_bg == block_group)
- return used_bg;
+ spin_lock(&root->fs_info->trans_lock);
+ cur_trans = root->fs_info->running_transaction;
+ if (cur_trans)
+ refcount_inc(&cur_trans->use_count);
+ spin_unlock(&root->fs_info->trans_lock);
+ if (!cur_trans)
+ return 0;
- btrfs_get_block_group(used_bg);
+ delayed_refs = &cur_trans->delayed_refs;
+ spin_lock(&delayed_refs->lock);
+ head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
+ if (!head) {
+ spin_unlock(&delayed_refs->lock);
+ btrfs_put_transaction(cur_trans);
+ return 0;
+ }
- if (!delalloc)
- return used_bg;
+ if (!mutex_trylock(&head->mutex)) {
+ refcount_inc(&head->refs);
+ spin_unlock(&delayed_refs->lock);
- if (down_read_trylock(&used_bg->data_rwsem))
- return used_bg;
+ btrfs_release_path(path);
- spin_unlock(&cluster->refill_lock);
+ /*
+ * Mutex was contended, block until it's released and let
+ * caller try again
+ */
+ mutex_lock(&head->mutex);
+ mutex_unlock(&head->mutex);
+ btrfs_put_delayed_ref_head(head);
+ btrfs_put_transaction(cur_trans);
+ return -EAGAIN;
+ }
+ spin_unlock(&delayed_refs->lock);
- /* We should only have one-level nested. */
- down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING);
+ spin_lock(&head->lock);
+ /*
+ * XXX: We should replace this with a proper search function in the
+ * future.
+ */
+ for (node = rb_first_cached(&head->ref_tree); node;
+ node = rb_next(node)) {
+ ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
+ /* If it's a shared ref we know a cross reference exists */
+ if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
+ ret = 1;
+ break;
+ }
- spin_lock(&cluster->refill_lock);
- if (used_bg == cluster->block_group)
- return used_bg;
+ data_ref = btrfs_delayed_node_to_data_ref(ref);
- up_read(&used_bg->data_rwsem);
- btrfs_put_block_group(used_bg);
+ /*
+ * If our ref doesn't match the one we're currently looking at
+ * then we have a cross reference.
+ */
+ if (data_ref->root != root->root_key.objectid ||
+ data_ref->objectid != objectid ||
+ data_ref->offset != offset) {
+ ret = 1;
+ break;
+ }
}
+ spin_unlock(&head->lock);
+ mutex_unlock(&head->mutex);
+ btrfs_put_transaction(cur_trans);
+ return ret;
}
-static inline void
-btrfs_release_block_group(struct btrfs_block_group_cache *cache,
- int delalloc)
+static noinline int check_committed_ref(struct btrfs_root *root,
+ struct btrfs_path *path,
+ u64 objectid, u64 offset, u64 bytenr)
{
- if (delalloc)
- up_read(&cache->data_rwsem);
- btrfs_put_block_group(cache);
-}
-
-/*
- * Structure used internally for find_free_extent() function. Wraps needed
- * parameters.
- */
-struct find_free_extent_ctl {
- /* Basic allocation info */
- u64 ram_bytes;
- u64 num_bytes;
- u64 empty_size;
- u64 flags;
- int delalloc;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ struct btrfs_root *extent_root = fs_info->extent_root;
+ struct extent_buffer *leaf;
+ struct btrfs_extent_data_ref *ref;
+ struct btrfs_extent_inline_ref *iref;
+ struct btrfs_extent_item *ei;
+ struct btrfs_key key;
+ u32 item_size;
+ int type;
+ int ret;
- /* Where to start the search inside the bg */
- u64 search_start;
+ key.objectid = bytenr;
+ key.offset = (u64)-1;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
- /* For clustered allocation */
- u64 empty_cluster;
+ ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
+ if (ret < 0)
+ goto out;
+ BUG_ON(ret == 0); /* Corruption */
- bool have_caching_bg;
- bool orig_have_caching_bg;
+ ret = -ENOENT;
+ if (path->slots[0] == 0)
+ goto out;
- /* RAID index, converted from flags */
- int index;
+ path->slots[0]--;
+ leaf = path->nodes[0];
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- /*
- * Current loop number, check find_free_extent_update_loop() for details
- */
- int loop;
+ if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
+ goto out;
- /*
- * Whether we're refilling a cluster, if true we need to re-search
- * current block group but don't try to refill the cluster again.
- */
- bool retry_clustered;
+ ret = 1;
+ item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+ ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
- /*
- * Whether we're updating free space cache, if true we need to re-search
- * current block group but don't try updating free space cache again.
- */
- bool retry_unclustered;
+ /* If extent item has more than 1 inline ref then it's shared */
+ if (item_size != sizeof(*ei) +
+ btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
+ goto out;
- /* If current block group is cached */
- int cached;
+ /* If extent created before last snapshot => it's definitely shared */
+ if (btrfs_extent_generation(leaf, ei) <=
+ btrfs_root_last_snapshot(&root->root_item))
+ goto out;
- /* Max contiguous hole found */
- u64 max_extent_size;
+ iref = (struct btrfs_extent_inline_ref *)(ei + 1);
- /* Total free space from free space cache, not always contiguous */
- u64 total_free_space;
+ /* If this extent has SHARED_DATA_REF then it's shared */
+ type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
+ if (type != BTRFS_EXTENT_DATA_REF_KEY)
+ goto out;
- /* Found result */
- u64 found_offset;
-};
+ ref = (struct btrfs_extent_data_ref *)(&iref->offset);
+ if (btrfs_extent_refs(leaf, ei) !=
+ btrfs_extent_data_ref_count(leaf, ref) ||
+ btrfs_extent_data_ref_root(leaf, ref) !=
+ root->root_key.objectid ||
+ btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
+ btrfs_extent_data_ref_offset(leaf, ref) != offset)
+ goto out;
+ ret = 0;
+out:
+ return ret;
+}
-/*
- * Helper function for find_free_extent().
- *
- * Return -ENOENT to inform caller that we need fallback to unclustered mode.
- * Return -EAGAIN to inform caller that we need to re-search this block group
- * Return >0 to inform caller that we find nothing
- * Return 0 means we have found a location and set ffe_ctl->found_offset.
- */
-static int find_free_extent_clustered(struct btrfs_block_group_cache *bg,
- struct btrfs_free_cluster *last_ptr,
- struct find_free_extent_ctl *ffe_ctl,
- struct btrfs_block_group_cache **cluster_bg_ret)
+int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset,
+ u64 bytenr)
{
- struct btrfs_block_group_cache *cluster_bg;
- u64 aligned_cluster;
- u64 offset;
+ struct btrfs_path *path;
int ret;
- cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc);
- if (!cluster_bg)
- goto refill_cluster;
- if (cluster_bg != bg && (cluster_bg->ro ||
- !block_group_bits(cluster_bg, ffe_ctl->flags)))
- goto release_cluster;
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
- offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr,
- ffe_ctl->num_bytes, cluster_bg->key.objectid,
- &ffe_ctl->max_extent_size);
- if (offset) {
- /* We have a block, we're done */
- spin_unlock(&last_ptr->refill_lock);
- trace_btrfs_reserve_extent_cluster(cluster_bg,
- ffe_ctl->search_start, ffe_ctl->num_bytes);
- *cluster_bg_ret = cluster_bg;
- ffe_ctl->found_offset = offset;
- return 0;
- }
- WARN_ON(last_ptr->block_group != cluster_bg);
+ do {
+ ret = check_committed_ref(root, path, objectid,
+ offset, bytenr);
+ if (ret && ret != -ENOENT)
+ goto out;
-release_cluster:
- /*
- * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so
- * lets just skip it and let the allocator find whatever block it can
- * find. If we reach this point, we will have tried the cluster
- * allocator plenty of times and not have found anything, so we are
- * likely way too fragmented for the clustering stuff to find anything.
- *
- * However, if the cluster is taken from the current block group,
- * release the cluster first, so that we stand a better chance of
- * succeeding in the unclustered allocation.
- */
- if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) {
- spin_unlock(&last_ptr->refill_lock);
- btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
- return -ENOENT;
- }
+ ret = check_delayed_ref(root, path, objectid, offset, bytenr);
+ } while (ret == -EAGAIN);
- /* This cluster didn't work out, free it and start over */
- btrfs_return_cluster_to_free_space(NULL, last_ptr);
+out:
+ btrfs_free_path(path);
+ if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
+ WARN_ON(ret > 0);
+ return ret;
+}
- if (cluster_bg != bg)
- btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
+static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct extent_buffer *buf,
+ int full_backref, int inc)
+{
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ u64 bytenr;
+ u64 num_bytes;
+ u64 parent;
+ u64 ref_root;
+ u32 nritems;
+ struct btrfs_key key;
+ struct btrfs_file_extent_item *fi;
+ struct btrfs_ref generic_ref = { 0 };
+ bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC);
+ int i;
+ int action;
+ int level;
+ int ret = 0;
-refill_cluster:
- if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) {
- spin_unlock(&last_ptr->refill_lock);
- return -ENOENT;
- }
+ if (btrfs_is_testing(fs_info))
+ return 0;
- aligned_cluster = max_t(u64,
- ffe_ctl->empty_cluster + ffe_ctl->empty_size,
- bg->full_stripe_len);
- ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start,
- ffe_ctl->num_bytes, aligned_cluster);
- if (ret == 0) {
- /* Now pull our allocation out of this cluster */
- offset = btrfs_alloc_from_cluster(bg, last_ptr,
- ffe_ctl->num_bytes, ffe_ctl->search_start,
- &ffe_ctl->max_extent_size);
- if (offset) {
- /* We found one, proceed */
- spin_unlock(&last_ptr->refill_lock);
- trace_btrfs_reserve_extent_cluster(bg,
- ffe_ctl->search_start,
- ffe_ctl->num_bytes);
- ffe_ctl->found_offset = offset;
- return 0;
- }
- } else if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT &&
- !ffe_ctl->retry_clustered) {
- spin_unlock(&last_ptr->refill_lock);
+ ref_root = btrfs_header_owner(buf);
+ nritems = btrfs_header_nritems(buf);
+ level = btrfs_header_level(buf);
- ffe_ctl->retry_clustered = true;
- wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
- ffe_ctl->empty_cluster + ffe_ctl->empty_size);
- return -EAGAIN;
- }
- /*
- * At this point we either didn't find a cluster or we weren't able to
- * allocate a block from our cluster. Free the cluster we've been
- * trying to use, and go to the next block group.
- */
- btrfs_return_cluster_to_free_space(NULL, last_ptr);
- spin_unlock(&last_ptr->refill_lock);
- return 1;
-}
+ if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state) && level == 0)
+ return 0;
-/*
- * Return >0 to inform caller that we find nothing
- * Return 0 when we found an free extent and set ffe_ctrl->found_offset
- * Return -EAGAIN to inform caller that we need to re-search this block group
- */
-static int find_free_extent_unclustered(struct btrfs_block_group_cache *bg,
- struct btrfs_free_cluster *last_ptr,
- struct find_free_extent_ctl *ffe_ctl)
-{
- u64 offset;
+ if (full_backref)
+ parent = buf->start;
+ else
+ parent = 0;
+ if (inc)
+ action = BTRFS_ADD_DELAYED_REF;
+ else
+ action = BTRFS_DROP_DELAYED_REF;
- /*
- * We are doing an unclustered allocation, set the fragmented flag so
- * we don't bother trying to setup a cluster again until we get more
- * space.
- */
- if (unlikely(last_ptr)) {
- spin_lock(&last_ptr->lock);
- last_ptr->fragmented = 1;
- spin_unlock(&last_ptr->lock);
- }
- if (ffe_ctl->cached) {
- struct btrfs_free_space_ctl *free_space_ctl;
+ for (i = 0; i < nritems; i++) {
+ if (level == 0) {
+ btrfs_item_key_to_cpu(buf, &key, i);
+ if (key.type != BTRFS_EXTENT_DATA_KEY)
+ continue;
+ fi = btrfs_item_ptr(buf, i,
+ struct btrfs_file_extent_item);
+ if (btrfs_file_extent_type(buf, fi) ==
+ BTRFS_FILE_EXTENT_INLINE)
+ continue;
+ bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+ if (bytenr == 0)
+ continue;
- free_space_ctl = bg->free_space_ctl;
- spin_lock(&free_space_ctl->tree_lock);
- if (free_space_ctl->free_space <
- ffe_ctl->num_bytes + ffe_ctl->empty_cluster +
- ffe_ctl->empty_size) {
- ffe_ctl->total_free_space = max_t(u64,
- ffe_ctl->total_free_space,
- free_space_ctl->free_space);
- spin_unlock(&free_space_ctl->tree_lock);
- return 1;
+ num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
+ key.offset -= btrfs_file_extent_offset(buf, fi);
+ btrfs_init_generic_ref(&generic_ref, action, bytenr,
+ num_bytes, parent);
+ generic_ref.real_root = root->root_key.objectid;
+ btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
+ key.offset);
+ generic_ref.skip_qgroup = for_reloc;
+ if (inc)
+ ret = btrfs_inc_extent_ref(trans, &generic_ref);
+ else
+ ret = btrfs_free_extent(trans, &generic_ref);
+ if (ret)
+ goto fail;
+ } else {
+ bytenr = btrfs_node_blockptr(buf, i);
+ num_bytes = fs_info->nodesize;
+ btrfs_init_generic_ref(&generic_ref, action, bytenr,
+ num_bytes, parent);
+ generic_ref.real_root = root->root_key.objectid;
+ btrfs_init_tree_ref(&generic_ref, level - 1, ref_root);
+ generic_ref.skip_qgroup = for_reloc;
+ if (inc)
+ ret = btrfs_inc_extent_ref(trans, &generic_ref);
+ else
+ ret = btrfs_free_extent(trans, &generic_ref);
+ if (ret)
+ goto fail;
}
- spin_unlock(&free_space_ctl->tree_lock);
}
+ return 0;
+fail:
+ return ret;
+}
- offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start,
- ffe_ctl->num_bytes, ffe_ctl->empty_size,
- &ffe_ctl->max_extent_size);
+int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+ struct extent_buffer *buf, int full_backref)
+{
+ return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
+}
- /*
- * If we didn't find a chunk, and we haven't failed on this block group
- * before, and this block group is in the middle of caching and we are
- * ok with waiting, then go ahead and wait for progress to be made, and
- * set @retry_unclustered to true.
- *
- * If @retry_unclustered is true then we've already waited on this
- * block group once and should move on to the next block group.
- */
- if (!offset && !ffe_ctl->retry_unclustered && !ffe_ctl->cached &&
- ffe_ctl->loop > LOOP_CACHING_NOWAIT) {
- wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
- ffe_ctl->empty_size);
- ffe_ctl->retry_unclustered = true;
- return -EAGAIN;
- } else if (!offset) {
- return 1;
- }
- ffe_ctl->found_offset = offset;
- return 0;
+int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+ struct extent_buffer *buf, int full_backref)
+{
+ return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
}
-/*
- * Return >0 means caller needs to re-search for free extent
- * Return 0 means we have the needed free extent.
- * Return <0 means we failed to locate any free extent.
- */
-static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info,
- struct btrfs_free_cluster *last_ptr,
- struct btrfs_key *ins,
- struct find_free_extent_ctl *ffe_ctl,
- int full_search, bool use_cluster)
+int btrfs_extent_readonly(struct btrfs_fs_info *fs_info, u64 bytenr)
{
- struct btrfs_root *root = fs_info->extent_root;
- int ret;
+ struct btrfs_block_group_cache *block_group;
+ int readonly = 0;
- if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) &&
- ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg)
- ffe_ctl->orig_have_caching_bg = true;
+ block_group = btrfs_lookup_block_group(fs_info, bytenr);
+ if (!block_group || block_group->ro)
+ readonly = 1;
+ if (block_group)
+ btrfs_put_block_group(block_group);
+ return readonly;
+}
- if (!ins->objectid && ffe_ctl->loop >= LOOP_CACHING_WAIT &&
- ffe_ctl->have_caching_bg)
- return 1;
+static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data)
+{
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ u64 flags;
+ u64 ret;
- if (!ins->objectid && ++(ffe_ctl->index) < BTRFS_NR_RAID_TYPES)
- return 1;
+ if (data)
+ flags = BTRFS_BLOCK_GROUP_DATA;
+ else if (root == fs_info->chunk_root)
+ flags = BTRFS_BLOCK_GROUP_SYSTEM;
+ else
+ flags = BTRFS_BLOCK_GROUP_METADATA;
- if (ins->objectid) {
- if (!use_cluster && last_ptr) {
- spin_lock(&last_ptr->lock);
- last_ptr->window_start = ins->objectid;
- spin_unlock(&last_ptr->lock);
- }
- return 0;
- }
+ ret = btrfs_get_alloc_profile(fs_info, flags);
+ return ret;
+}
- /*
- * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
- * caching kthreads as we move along
- * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
- * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
- * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
- * again
- */
- if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) {
- ffe_ctl->index = 0;
- if (ffe_ctl->loop == LOOP_CACHING_NOWAIT) {
- /*
- * We want to skip the LOOP_CACHING_WAIT step if we
- * don't have any uncached bgs and we've already done a
- * full search through.
- */
- if (ffe_ctl->orig_have_caching_bg || !full_search)
- ffe_ctl->loop = LOOP_CACHING_WAIT;
- else
- ffe_ctl->loop = LOOP_ALLOC_CHUNK;
- } else {
- ffe_ctl->loop++;
- }
+static u64 first_logical_byte(struct btrfs_fs_info *fs_info, u64 search_start)
+{
+ struct btrfs_block_group_cache *cache;
+ u64 bytenr;
- if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) {
- struct btrfs_trans_handle *trans;
- int exist = 0;
+ spin_lock(&fs_info->block_group_cache_lock);
+ bytenr = fs_info->first_logical_byte;
+ spin_unlock(&fs_info->block_group_cache_lock);
- trans = current->journal_info;
- if (trans)
- exist = 1;
- else
- trans = btrfs_join_transaction(root);
+ if (bytenr < (u64)-1)
+ return bytenr;
- if (IS_ERR(trans)) {
- ret = PTR_ERR(trans);
- return ret;
- }
+ cache = btrfs_lookup_first_block_group(fs_info, search_start);
+ if (!cache)
+ return 0;
- ret = btrfs_chunk_alloc(trans, ffe_ctl->flags,
- CHUNK_ALLOC_FORCE);
+ bytenr = cache->key.objectid;
+ btrfs_put_block_group(cache);
- /*
- * If we can't allocate a new chunk we've already looped
- * through at least once, move on to the NO_EMPTY_SIZE
- * case.
- */
- if (ret == -ENOSPC)
- ffe_ctl->loop = LOOP_NO_EMPTY_SIZE;
+ return bytenr;
+}
- /* Do not bail out on ENOSPC since we can do more. */
- if (ret < 0 && ret != -ENOSPC)
- btrfs_abort_transaction(trans, ret);
- else
- ret = 0;
- if (!exist)
- btrfs_end_transaction(trans);
- if (ret)
- return ret;
- }
+static int pin_down_extent(struct btrfs_block_group_cache *cache,
+ u64 bytenr, u64 num_bytes, int reserved)
+{
+ struct btrfs_fs_info *fs_info = cache->fs_info;
- if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) {
- /*
- * Don't loop again if we already have no empty_size and
- * no empty_cluster.
- */
- if (ffe_ctl->empty_size == 0 &&
- ffe_ctl->empty_cluster == 0)
- return -ENOSPC;
- ffe_ctl->empty_size = 0;
- ffe_ctl->empty_cluster = 0;
- }
- return 1;
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ cache->pinned += num_bytes;
+ btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info,
+ num_bytes);
+ if (reserved) {
+ cache->reserved -= num_bytes;
+ cache->space_info->bytes_reserved -= num_bytes;
}
- return -ENOSPC;
+ 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(fs_info->pinned_extents, bytenr,
+ bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
+ return 0;
}
/*
- * walks the btree of allocated extents and find a hole of a given size.
- * The key ins is changed to record the hole:
- * ins->objectid == start position
- * ins->flags = BTRFS_EXTENT_ITEM_KEY
- * ins->offset == the size of the hole.
- * Any available blocks before search_start are skipped.
- *
- * If there is no suitable free space, we will record the max size of
- * the free space extent currently.
- *
- * The overall logic and call chain:
- *
- * find_free_extent()
- * |- Iterate through all block groups
- * | |- Get a valid block group
- * | |- Try to do clustered allocation in that block group
- * | |- Try to do unclustered allocation in that block group
- * | |- Check if the result is valid
- * | | |- If valid, then exit
- * | |- Jump to next block group
- * |
- * |- Push harder to find free extents
- * |- If not found, re-iterate all block groups
+ * this function must be called within transaction
*/
-static noinline int find_free_extent(struct btrfs_fs_info *fs_info,
- u64 ram_bytes, u64 num_bytes, u64 empty_size,
- u64 hint_byte, struct btrfs_key *ins,
- u64 flags, int delalloc)
+int btrfs_pin_extent(struct btrfs_fs_info *fs_info,
+ u64 bytenr, u64 num_bytes, int reserved)
{
- int ret = 0;
- struct btrfs_free_cluster *last_ptr = NULL;
- struct btrfs_block_group_cache *block_group = NULL;
- struct find_free_extent_ctl ffe_ctl = {0};
- struct btrfs_space_info *space_info;
- bool use_cluster = true;
- bool full_search = false;
+ struct btrfs_block_group_cache *cache;
- WARN_ON(num_bytes < fs_info->sectorsize);
+ cache = btrfs_lookup_block_group(fs_info, bytenr);
+ BUG_ON(!cache); /* Logic error */
- ffe_ctl.ram_bytes = ram_bytes;
- ffe_ctl.num_bytes = num_bytes;
- ffe_ctl.empty_size = empty_size;
- ffe_ctl.flags = flags;
- ffe_ctl.search_start = 0;
- ffe_ctl.retry_clustered = false;
- ffe_ctl.retry_unclustered = false;
- ffe_ctl.delalloc = delalloc;
- ffe_ctl.index = btrfs_bg_flags_to_raid_index(flags);
- ffe_ctl.have_caching_bg = false;
- ffe_ctl.orig_have_caching_bg = false;
- ffe_ctl.found_offset = 0;
+ pin_down_extent(cache, bytenr, num_bytes, reserved);
- ins->type = BTRFS_EXTENT_ITEM_KEY;
- ins->objectid = 0;
- ins->offset = 0;
+ btrfs_put_block_group(cache);
+ return 0;
+}
- trace_find_free_extent(fs_info, num_bytes, empty_size, flags);
+/*
+ * this function must be called within transaction
+ */
+int btrfs_pin_extent_for_log_replay(struct btrfs_fs_info *fs_info,
+ u64 bytenr, u64 num_bytes)
+{
+ struct btrfs_block_group_cache *cache;
+ int ret;
- space_info = btrfs_find_space_info(fs_info, flags);
- if (!space_info) {
- btrfs_err(fs_info, "No space info for %llu", flags);
- return -ENOSPC;
- }
+ cache = btrfs_lookup_block_group(fs_info, bytenr);
+ if (!cache)
+ return -EINVAL;
/*
- * If our free space is heavily fragmented we may not be able to make
- * big contiguous allocations, so instead of doing the expensive search
- * for free space, simply return ENOSPC with our max_extent_size so we
- * can go ahead and search for a more manageable chunk.
- *
- * If our max_extent_size is large enough for our allocation simply
- * disable clustering since we will likely not be able to find enough
- * space to create a cluster and induce latency trying.
+ * pull in the free space cache (if any) so that our pin
+ * removes the free space from the cache. We have load_only set
+ * to one because the slow code to read in the free extents does check
+ * the pinned extents.
*/
- if (unlikely(space_info->max_extent_size)) {
- spin_lock(&space_info->lock);
- if (space_info->max_extent_size &&
- num_bytes > space_info->max_extent_size) {
- ins->offset = space_info->max_extent_size;
- spin_unlock(&space_info->lock);
- return -ENOSPC;
- } else if (space_info->max_extent_size) {
- use_cluster = false;
- }
- spin_unlock(&space_info->lock);
- }
+ btrfs_cache_block_group(cache, 1);
- last_ptr = fetch_cluster_info(fs_info, space_info,
- &ffe_ctl.empty_cluster);
- if (last_ptr) {
- spin_lock(&last_ptr->lock);
- if (last_ptr->block_group)
- hint_byte = last_ptr->window_start;
- if (last_ptr->fragmented) {
- /*
- * We still set window_start so we can keep track of the
- * last place we found an allocation to try and save
- * some time.
- */
- hint_byte = last_ptr->window_start;
- use_cluster = false;
- }
- spin_unlock(&last_ptr->lock);
- }
+ pin_down_extent(cache, bytenr, num_bytes, 0);
- ffe_ctl.search_start = max(ffe_ctl.search_start,
- first_logical_byte(fs_info, 0));
- ffe_ctl.search_start = max(ffe_ctl.search_start, hint_byte);
- if (ffe_ctl.search_start == hint_byte) {
- block_group = btrfs_lookup_block_group(fs_info,
- ffe_ctl.search_start);
- /*
- * we don't want to use the block group if it doesn't match our
- * allocation bits, or if its not cached.
- *
- * However if we are re-searching with an ideal block group
- * picked out then we don't care that the block group is cached.
- */
- if (block_group && block_group_bits(block_group, flags) &&
- block_group->cached != BTRFS_CACHE_NO) {
- down_read(&space_info->groups_sem);
- if (list_empty(&block_group->list) ||
- block_group->ro) {
- /*
- * someone is removing this block group,
- * we can't jump into the have_block_group
- * target because our list pointers are not
- * valid
- */
- btrfs_put_block_group(block_group);
- up_read(&space_info->groups_sem);
- } else {
- ffe_ctl.index = btrfs_bg_flags_to_raid_index(
- block_group->flags);
- btrfs_lock_block_group(block_group, delalloc);
- goto have_block_group;
- }
- } else if (block_group) {
- btrfs_put_block_group(block_group);
- }
- }
-search:
- ffe_ctl.have_caching_bg = false;
- if (ffe_ctl.index == btrfs_bg_flags_to_raid_index(flags) ||
- ffe_ctl.index == 0)
- full_search = true;
- down_read(&space_info->groups_sem);
- list_for_each_entry(block_group,
- &space_info->block_groups[ffe_ctl.index], list) {
- /* If the block group is read-only, we can skip it entirely. */
- if (unlikely(block_group->ro))
- continue;
+ /* remove us from the free space cache (if we're there at all) */
+ ret = btrfs_remove_free_space(cache, bytenr, num_bytes);
+ btrfs_put_block_group(cache);
+ return ret;
+}
- btrfs_grab_block_group(block_group, delalloc);
- ffe_ctl.search_start = block_group->key.objectid;
+static int __exclude_logged_extent(struct btrfs_fs_info *fs_info,
+ u64 start, u64 num_bytes)
+{
+ int ret;
+ struct btrfs_block_group_cache *block_group;
+ struct btrfs_caching_control *caching_ctl;
- /*
- * this can happen if we end up cycling through all the
- * raid types, but we want to make sure we only allocate
- * for the proper type.
- */
- if (!block_group_bits(block_group, flags)) {
- u64 extra = BTRFS_BLOCK_GROUP_DUP |
- BTRFS_BLOCK_GROUP_RAID1_MASK |
- BTRFS_BLOCK_GROUP_RAID56_MASK |
- BTRFS_BLOCK_GROUP_RAID10;
+ block_group = btrfs_lookup_block_group(fs_info, start);
+ if (!block_group)
+ return -EINVAL;
- /*
- * if they asked for extra copies and this block group
- * doesn't provide them, bail. This does allow us to
- * fill raid0 from raid1.
- */
- if ((flags & extra) && !(block_group->flags & extra))
- goto loop;
- }
+ btrfs_cache_block_group(block_group, 0);
+ caching_ctl = btrfs_get_caching_control(block_group);
-have_block_group:
- ffe_ctl.cached = block_group_cache_done(block_group);
- if (unlikely(!ffe_ctl.cached)) {
- ffe_ctl.have_caching_bg = true;
- ret = cache_block_group(block_group, 0);
- BUG_ON(ret < 0);
- ret = 0;
+ if (!caching_ctl) {
+ /* Logic error */
+ BUG_ON(!btrfs_block_group_cache_done(block_group));
+ ret = btrfs_remove_free_space(block_group, start, num_bytes);
+ } else {
+ mutex_lock(&caching_ctl->mutex);
+
+ if (start >= caching_ctl->progress) {
+ ret = btrfs_add_excluded_extent(fs_info, start,
+ num_bytes);
+ } else if (start + num_bytes <= caching_ctl->progress) {
+ ret = btrfs_remove_free_space(block_group,
+ start, num_bytes);
+ } else {
+ num_bytes = caching_ctl->progress - start;
+ ret = btrfs_remove_free_space(block_group,
+ start, num_bytes);
+ if (ret)
+ goto out_lock;
+
+ num_bytes = (start + num_bytes) -
+ caching_ctl->progress;
+ start = caching_ctl->progress;
+ ret = btrfs_add_excluded_extent(fs_info, start,
+ num_bytes);
}
+out_lock:
+ mutex_unlock(&caching_ctl->mutex);
+ btrfs_put_caching_control(caching_ctl);
+ }
+ btrfs_put_block_group(block_group);
+ return ret;
+}
- if (unlikely(block_group->cached == BTRFS_CACHE_ERROR))
- goto loop;
+int btrfs_exclude_logged_extents(struct extent_buffer *eb)
+{
+ struct btrfs_fs_info *fs_info = eb->fs_info;
+ struct btrfs_file_extent_item *item;
+ struct btrfs_key key;
+ int found_type;
+ int i;
+ int ret = 0;
- /*
- * Ok we want to try and use the cluster allocator, so
- * lets look there
- */
- if (last_ptr && use_cluster) {
- struct btrfs_block_group_cache *cluster_bg = NULL;
+ if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS))
+ return 0;
- ret = find_free_extent_clustered(block_group, last_ptr,
- &ffe_ctl, &cluster_bg);
+ for (i = 0; i < btrfs_header_nritems(eb); i++) {
+ btrfs_item_key_to_cpu(eb, &key, i);
+ if (key.type != BTRFS_EXTENT_DATA_KEY)
+ continue;
+ item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
+ found_type = btrfs_file_extent_type(eb, item);
+ if (found_type == BTRFS_FILE_EXTENT_INLINE)
+ continue;
+ if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
+ continue;
+ key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
+ key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
+ ret = __exclude_logged_extent(fs_info, key.objectid, key.offset);
+ if (ret)
+ break;
+ }
- if (ret == 0) {
- if (cluster_bg && cluster_bg != block_group) {
- btrfs_release_block_group(block_group,
- delalloc);
- block_group = cluster_bg;
- }
- goto checks;
- } else if (ret == -EAGAIN) {
- goto have_block_group;
- } else if (ret > 0) {
- goto loop;
- }
- /* ret == -ENOENT case falls through */
- }
+ return ret;
+}
- ret = find_free_extent_unclustered(block_group, last_ptr,
- &ffe_ctl);
- if (ret == -EAGAIN)
- goto have_block_group;
- else if (ret > 0)
- goto loop;
- /* ret == 0 case falls through */
-checks:
- ffe_ctl.search_start = round_up(ffe_ctl.found_offset,
- fs_info->stripesize);
+static void
+btrfs_inc_block_group_reservations(struct btrfs_block_group_cache *bg)
+{
+ atomic_inc(&bg->reservations);
+}
- /* move on to the next group */
- if (ffe_ctl.search_start + num_bytes >
- block_group->key.objectid + block_group->key.offset) {
- btrfs_add_free_space(block_group, ffe_ctl.found_offset,
- num_bytes);
- goto loop;
- }
+void btrfs_prepare_extent_commit(struct btrfs_fs_info *fs_info)
+{
+ struct btrfs_caching_control *next;
+ struct btrfs_caching_control *caching_ctl;
+ struct btrfs_block_group_cache *cache;
- if (ffe_ctl.found_offset < ffe_ctl.search_start)
- btrfs_add_free_space(block_group, ffe_ctl.found_offset,
- ffe_ctl.search_start - ffe_ctl.found_offset);
+ down_write(&fs_info->commit_root_sem);
- ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
- num_bytes, delalloc);
- if (ret == -EAGAIN) {
- btrfs_add_free_space(block_group, ffe_ctl.found_offset,
- num_bytes);
- goto loop;
+ list_for_each_entry_safe(caching_ctl, next,
+ &fs_info->caching_block_groups, list) {
+ cache = caching_ctl->block_group;
+ if (btrfs_block_group_cache_done(cache)) {
+ cache->last_byte_to_unpin = (u64)-1;
+ list_del_init(&caching_ctl->list);
+ btrfs_put_caching_control(caching_ctl);
+ } else {
+ cache->last_byte_to_unpin = caching_ctl->progress;
}
- btrfs_inc_block_group_reservations(block_group);
+ }
- /* we are all good, lets return */
- ins->objectid = ffe_ctl.search_start;
- ins->offset = num_bytes;
+ if (fs_info->pinned_extents == &fs_info->freed_extents[0])
+ fs_info->pinned_extents = &fs_info->freed_extents[1];
+ else
+ fs_info->pinned_extents = &fs_info->freed_extents[0];
- trace_btrfs_reserve_extent(block_group, ffe_ctl.search_start,
- num_bytes);
- btrfs_release_block_group(block_group, delalloc);
- break;
-loop:
- ffe_ctl.retry_clustered = false;
- ffe_ctl.retry_unclustered = false;
- BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) !=
- ffe_ctl.index);
- btrfs_release_block_group(block_group, delalloc);
- cond_resched();
- }
- up_read(&space_info->groups_sem);
+ up_write(&fs_info->commit_root_sem);
- ret = find_free_extent_update_loop(fs_info, last_ptr, ins, &ffe_ctl,
- full_search, use_cluster);
- if (ret > 0)
- goto search;
+ btrfs_update_global_block_rsv(fs_info);
+}
- if (ret == -ENOSPC) {
- /*
- * Use ffe_ctl->total_free_space as fallback if we can't find
- * any contiguous hole.
- */
- if (!ffe_ctl.max_extent_size)
- ffe_ctl.max_extent_size = ffe_ctl.total_free_space;
- spin_lock(&space_info->lock);
- space_info->max_extent_size = ffe_ctl.max_extent_size;
- spin_unlock(&space_info->lock);
- ins->offset = ffe_ctl.max_extent_size;
+/*
+ * Returns the free cluster for the given space info and sets empty_cluster to
+ * what it should be based on the mount options.
+ */
+static struct btrfs_free_cluster *
+fetch_cluster_info(struct btrfs_fs_info *fs_info,
+ struct btrfs_space_info *space_info, u64 *empty_cluster)
+{
+ struct btrfs_free_cluster *ret = NULL;
+
+ *empty_cluster = 0;
+ if (btrfs_mixed_space_info(space_info))
+ return ret;
+
+ if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
+ ret = &fs_info->meta_alloc_cluster;
+ if (btrfs_test_opt(fs_info, SSD))
+ *empty_cluster = SZ_2M;
+ else
+ *empty_cluster = SZ_64K;
+ } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
+ btrfs_test_opt(fs_info, SSD_SPREAD)) {
+ *empty_cluster = SZ_2M;
+ ret = &fs_info->data_alloc_cluster;
}
+
return ret;
}
-/*
- * btrfs_reserve_extent - entry point to the extent allocator. Tries to find a
- * hole that is at least as big as @num_bytes.
- *
- * @root - The root that will contain this extent
- *
- * @ram_bytes - The amount of space in ram that @num_bytes take. This
- * is used for accounting purposes. This value differs
- * from @num_bytes only in the case of compressed extents.
- *
- * @num_bytes - Number of bytes to allocate on-disk.
- *
- * @min_alloc_size - Indicates the minimum amount of space that the
- * allocator should try to satisfy. In some cases
- * @num_bytes may be larger than what is required and if
- * the filesystem is fragmented then allocation fails.
- * However, the presence of @min_alloc_size gives a
- * chance to try and satisfy the smaller allocation.
- *
- * @empty_size - A hint that you plan on doing more COW. This is the
- * size in bytes the allocator should try to find free
- * next to the block it returns. This is just a hint and
- * may be ignored by the allocator.
- *
- * @hint_byte - Hint to the allocator to start searching above the byte
- * address passed. It might be ignored.
- *
- * @ins - This key is modified to record the found hole. It will
- * have the following values:
- * ins->objectid == start position
- * ins->flags = BTRFS_EXTENT_ITEM_KEY
- * ins->offset == the size of the hole.
- *
- * @is_data - Boolean flag indicating whether an extent is
- * allocated for data (true) or metadata (false)
- *
- * @delalloc - Boolean flag indicating whether this allocation is for
- * delalloc or not. If 'true' data_rwsem of block groups
- * is going to be acquired.
- *
- *
- * Returns 0 when an allocation succeeded or < 0 when an error occurred. In
- * case -ENOSPC is returned then @ins->offset will contain the size of the
- * largest available hole the allocator managed to find.
- */
-int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
- u64 num_bytes, u64 min_alloc_size,
- u64 empty_size, u64 hint_byte,
- struct btrfs_key *ins, int is_data, int delalloc)
+static int unpin_extent_range(struct btrfs_fs_info *fs_info,
+ u64 start, u64 end,
+ const bool return_free_space)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- bool final_tried = num_bytes == min_alloc_size;
- u64 flags;
- int ret;
+ struct btrfs_block_group_cache *cache = NULL;
+ struct btrfs_space_info *space_info;
+ struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
+ struct btrfs_free_cluster *cluster = NULL;
+ u64 len;
+ u64 total_unpinned = 0;
+ u64 empty_cluster = 0;
+ bool readonly;
+
+ while (start <= end) {
+ readonly = false;
+ if (!cache ||
+ start >= cache->key.objectid + cache->key.offset) {
+ if (cache)
+ btrfs_put_block_group(cache);
+ total_unpinned = 0;
+ cache = btrfs_lookup_block_group(fs_info, start);
+ BUG_ON(!cache); /* Logic error */
+
+ cluster = fetch_cluster_info(fs_info,
+ cache->space_info,
+ &empty_cluster);
+ empty_cluster <<= 1;
+ }
+
+ len = cache->key.objectid + cache->key.offset - start;
+ len = min(len, end + 1 - start);
+
+ if (start < cache->last_byte_to_unpin) {
+ len = min(len, cache->last_byte_to_unpin - start);
+ if (return_free_space)
+ btrfs_add_free_space(cache, start, len);
+ }
- flags = get_alloc_profile_by_root(root, is_data);
-again:
- WARN_ON(num_bytes < fs_info->sectorsize);
- ret = find_free_extent(fs_info, ram_bytes, num_bytes, empty_size,
- hint_byte, ins, flags, delalloc);
- if (!ret && !is_data) {
- btrfs_dec_block_group_reservations(fs_info, ins->objectid);
- } else if (ret == -ENOSPC) {
- if (!final_tried && ins->offset) {
- num_bytes = min(num_bytes >> 1, ins->offset);
- num_bytes = round_down(num_bytes,
- fs_info->sectorsize);
- num_bytes = max(num_bytes, min_alloc_size);
- ram_bytes = num_bytes;
- if (num_bytes == min_alloc_size)
- final_tried = true;
- goto again;
- } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
- struct btrfs_space_info *sinfo;
+ start += len;
+ total_unpinned += len;
+ space_info = cache->space_info;
- sinfo = btrfs_find_space_info(fs_info, flags);
- btrfs_err(fs_info,
- "allocation failed flags %llu, wanted %llu",
- flags, num_bytes);
- if (sinfo)
- btrfs_dump_space_info(fs_info, sinfo,
- num_bytes, 1);
+ /*
+ * If this space cluster has been marked as fragmented and we've
+ * unpinned enough in this block group to potentially allow a
+ * cluster to be created inside of it go ahead and clear the
+ * fragmented check.
+ */
+ if (cluster && cluster->fragmented &&
+ total_unpinned > empty_cluster) {
+ spin_lock(&cluster->lock);
+ cluster->fragmented = 0;
+ spin_unlock(&cluster->lock);
+ }
+
+ spin_lock(&space_info->lock);
+ spin_lock(&cache->lock);
+ cache->pinned -= len;
+ btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len);
+ space_info->max_extent_size = 0;
+ percpu_counter_add_batch(&space_info->total_bytes_pinned,
+ -len, BTRFS_TOTAL_BYTES_PINNED_BATCH);
+ if (cache->ro) {
+ space_info->bytes_readonly += len;
+ readonly = true;
+ }
+ spin_unlock(&cache->lock);
+ if (!readonly && return_free_space &&
+ global_rsv->space_info == space_info) {
+ u64 to_add = len;
+
+ spin_lock(&global_rsv->lock);
+ if (!global_rsv->full) {
+ to_add = min(len, global_rsv->size -
+ global_rsv->reserved);
+ global_rsv->reserved += to_add;
+ btrfs_space_info_update_bytes_may_use(fs_info,
+ space_info, to_add);
+ if (global_rsv->reserved >= global_rsv->size)
+ global_rsv->full = 1;
+ len -= to_add;
+ }
+ spin_unlock(&global_rsv->lock);
+ /* Add to any tickets we may have */
+ if (len)
+ btrfs_try_granting_tickets(fs_info,
+ space_info);
}
+ spin_unlock(&space_info->lock);
}
- return ret;
+ if (cache)
+ btrfs_put_block_group(cache);
+ return 0;
}
-static int __btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
- u64 start, u64 len,
- int pin, int delalloc)
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
{
- struct btrfs_block_group_cache *cache;
- int ret = 0;
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ struct btrfs_block_group_cache *block_group, *tmp;
+ struct list_head *deleted_bgs;
+ struct extent_io_tree *unpin;
+ u64 start;
+ u64 end;
+ int ret;
- cache = btrfs_lookup_block_group(fs_info, start);
- if (!cache) {
- btrfs_err(fs_info, "Unable to find block group for %llu",
- start);
- return -ENOSPC;
- }
+ if (fs_info->pinned_extents == &fs_info->freed_extents[0])
+ unpin = &fs_info->freed_extents[1];
+ else
+ unpin = &fs_info->freed_extents[0];
+
+ while (!trans->aborted) {
+ struct extent_state *cached_state = NULL;
+
+ mutex_lock(&fs_info->unused_bg_unpin_mutex);
+ ret = find_first_extent_bit(unpin, 0, &start, &end,
+ EXTENT_DIRTY, &cached_state);
+ if (ret) {
+ mutex_unlock(&fs_info->unused_bg_unpin_mutex);
+ break;
+ }
- if (pin)
- pin_down_extent(cache, start, len, 1);
- else {
if (btrfs_test_opt(fs_info, DISCARD))
- ret = btrfs_discard_extent(fs_info, start, len, NULL);
- btrfs_add_free_space(cache, start, len);
- btrfs_free_reserved_bytes(cache, len, delalloc);
- trace_btrfs_reserved_extent_free(fs_info, start, len);
+ ret = btrfs_discard_extent(fs_info, start,
+ end + 1 - start, NULL);
+
+ clear_extent_dirty(unpin, start, end, &cached_state);
+ unpin_extent_range(fs_info, start, end, true);
+ mutex_unlock(&fs_info->unused_bg_unpin_mutex);
+ free_extent_state(cached_state);
+ cond_resched();
}
- btrfs_put_block_group(cache);
- return ret;
-}
+ /*
+ * Transaction is finished. We don't need the lock anymore. We
+ * do need to clean up the block groups in case of a transaction
+ * abort.
+ */
+ deleted_bgs = &trans->transaction->deleted_bgs;
+ list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) {
+ u64 trimmed = 0;
-int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
- u64 start, u64 len, int delalloc)
-{
- return __btrfs_free_reserved_extent(fs_info, start, len, 0, delalloc);
-}
+ ret = -EROFS;
+ if (!trans->aborted)
+ ret = btrfs_discard_extent(fs_info,
+ block_group->key.objectid,
+ block_group->key.offset,
+ &trimmed);
-int btrfs_free_and_pin_reserved_extent(struct btrfs_fs_info *fs_info,
- u64 start, u64 len)
-{
- return __btrfs_free_reserved_extent(fs_info, start, len, 1, 0);
+ list_del_init(&block_group->bg_list);
+ btrfs_put_block_group_trimming(block_group);
+ btrfs_put_block_group(block_group);
+
+ if (ret) {
+ const char *errstr = btrfs_decode_error(ret);
+ btrfs_warn(fs_info,
+ "discard failed while removing blockgroup: errno=%d %s",
+ ret, errstr);
+ }
+ }
+
+ return 0;
}
-static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
- u64 parent, u64 root_objectid,
- u64 flags, u64 owner, u64 offset,
- struct btrfs_key *ins, int ref_mod)
+static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_node *node, u64 parent,
+ u64 root_objectid, u64 owner_objectid,
+ u64 owner_offset, int refs_to_drop,
+ struct btrfs_delayed_extent_op *extent_op)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- int ret;
- struct btrfs_extent_item *extent_item;
- struct btrfs_extent_inline_ref *iref;
+ struct btrfs_fs_info *info = trans->fs_info;
+ struct btrfs_key key;
struct btrfs_path *path;
+ struct btrfs_root *extent_root = info->extent_root;
struct extent_buffer *leaf;
- int type;
- u32 size;
+ struct btrfs_extent_item *ei;
+ struct btrfs_extent_inline_ref *iref;
+ int ret;
+ int is_data;
+ int extent_slot = 0;
+ int found_extent = 0;
+ int num_to_del = 1;
+ u32 item_size;
+ u64 refs;
+ u64 bytenr = node->bytenr;
+ u64 num_bytes = node->num_bytes;
+ int last_ref = 0;
+ bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA);
- if (parent > 0)
- type = BTRFS_SHARED_DATA_REF_KEY;
- else
- type = BTRFS_EXTENT_DATA_REF_KEY;
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ path->reada = READA_FORWARD;
+ path->leave_spinning = 1;
+
+ is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
+ BUG_ON(!is_data && refs_to_drop != 1);
+
+ if (is_data)
+ skinny_metadata = false;
+
+ ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes,
+ parent, root_objectid, owner_objectid,
+ owner_offset);
+ if (ret == 0) {
+ extent_slot = path->slots[0];
+ while (extent_slot >= 0) {
+ btrfs_item_key_to_cpu(path->nodes[0], &key,
+ extent_slot);
+ if (key.objectid != bytenr)
+ break;
+ if (key.type == BTRFS_EXTENT_ITEM_KEY &&
+ key.offset == num_bytes) {
+ found_extent = 1;
+ break;
+ }
+ if (key.type == BTRFS_METADATA_ITEM_KEY &&
+ key.offset == owner_objectid) {
+ found_extent = 1;
+ break;
+ }
+ if (path->slots[0] - extent_slot > 5)
+ break;
+ extent_slot--;
+ }
+
+ if (!found_extent) {
+ BUG_ON(iref);
+ ret = remove_extent_backref(trans, path, NULL,
+ refs_to_drop,
+ is_data, &last_ref);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+ btrfs_release_path(path);
+ path->leave_spinning = 1;
+
+ key.objectid = bytenr;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = num_bytes;
+
+ if (!is_data && skinny_metadata) {
+ key.type = BTRFS_METADATA_ITEM_KEY;
+ key.offset = owner_objectid;
+ }
- size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
+ ret = btrfs_search_slot(trans, extent_root,
+ &key, path, -1, 1);
+ if (ret > 0 && skinny_metadata && path->slots[0]) {
+ /*
+ * Couldn't find our skinny metadata item,
+ * see if we have ye olde extent item.
+ */
+ path->slots[0]--;
+ btrfs_item_key_to_cpu(path->nodes[0], &key,
+ path->slots[0]);
+ if (key.objectid == bytenr &&
+ key.type == BTRFS_EXTENT_ITEM_KEY &&
+ key.offset == num_bytes)
+ ret = 0;
+ }
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
+ if (ret > 0 && skinny_metadata) {
+ skinny_metadata = false;
+ key.objectid = bytenr;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = num_bytes;
+ btrfs_release_path(path);
+ ret = btrfs_search_slot(trans, extent_root,
+ &key, path, -1, 1);
+ }
- path->leave_spinning = 1;
- ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
- ins, size);
- if (ret) {
- btrfs_free_path(path);
- return ret;
+ if (ret) {
+ btrfs_err(info,
+ "umm, got %d back from search, was looking for %llu",
+ ret, bytenr);
+ if (ret > 0)
+ btrfs_print_leaf(path->nodes[0]);
+ }
+ if (ret < 0) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+ extent_slot = path->slots[0];
+ }
+ } else if (WARN_ON(ret == -ENOENT)) {
+ btrfs_print_leaf(path->nodes[0]);
+ btrfs_err(info,
+ "unable to find ref byte nr %llu parent %llu root %llu owner %llu offset %llu",
+ bytenr, parent, root_objectid, owner_objectid,
+ owner_offset);
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ } else {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
}
leaf = path->nodes[0];
- extent_item = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_item);
- btrfs_set_extent_refs(leaf, extent_item, ref_mod);
- btrfs_set_extent_generation(leaf, extent_item, trans->transid);
- btrfs_set_extent_flags(leaf, extent_item,
- flags | BTRFS_EXTENT_FLAG_DATA);
+ item_size = btrfs_item_size_nr(leaf, extent_slot);
+ if (unlikely(item_size < sizeof(*ei))) {
+ ret = -EINVAL;
+ btrfs_print_v0_err(info);
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+ ei = btrfs_item_ptr(leaf, extent_slot,
+ struct btrfs_extent_item);
+ if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
+ key.type == BTRFS_EXTENT_ITEM_KEY) {
+ struct btrfs_tree_block_info *bi;
+ BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
+ bi = (struct btrfs_tree_block_info *)(ei + 1);
+ WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
+ }
- iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
- btrfs_set_extent_inline_ref_type(leaf, iref, type);
- if (parent > 0) {
- struct btrfs_shared_data_ref *ref;
- ref = (struct btrfs_shared_data_ref *)(iref + 1);
- btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
- btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
- } else {
- struct btrfs_extent_data_ref *ref;
- ref = (struct btrfs_extent_data_ref *)(&iref->offset);
- btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
- btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
- btrfs_set_extent_data_ref_offset(leaf, ref, offset);
- btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
+ refs = btrfs_extent_refs(leaf, ei);
+ if (refs < refs_to_drop) {
+ btrfs_err(info,
+ "trying to drop %d refs but we only have %Lu for bytenr %Lu",
+ refs_to_drop, refs, bytenr);
+ ret = -EINVAL;
+ btrfs_abort_transaction(trans, ret);
+ goto out;
}
+ refs -= refs_to_drop;
- btrfs_mark_buffer_dirty(path->nodes[0]);
- btrfs_free_path(path);
+ if (refs > 0) {
+ if (extent_op)
+ __run_delayed_extent_op(extent_op, leaf, ei);
+ /*
+ * In the case of inline back ref, reference count will
+ * be updated by remove_extent_backref
+ */
+ if (iref) {
+ BUG_ON(!found_extent);
+ } else {
+ btrfs_set_extent_refs(leaf, ei, refs);
+ btrfs_mark_buffer_dirty(leaf);
+ }
+ if (found_extent) {
+ ret = remove_extent_backref(trans, path, iref,
+ refs_to_drop, is_data,
+ &last_ref);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+ }
+ } else {
+ if (found_extent) {
+ BUG_ON(is_data && refs_to_drop !=
+ extent_data_ref_count(path, iref));
+ if (iref) {
+ BUG_ON(path->slots[0] != extent_slot);
+ } else {
+ BUG_ON(path->slots[0] != extent_slot + 1);
+ path->slots[0] = extent_slot;
+ num_to_del = 2;
+ }
+ }
- ret = remove_from_free_space_tree(trans, ins->objectid, ins->offset);
- if (ret)
- return ret;
+ last_ref = 1;
+ ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
+ num_to_del);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+ btrfs_release_path(path);
- ret = update_block_group(trans, ins->objectid, ins->offset, 1);
- if (ret) { /* -ENOENT, logic error */
- btrfs_err(fs_info, "update block group failed for %llu %llu",
- ins->objectid, ins->offset);
- BUG();
+ if (is_data) {
+ ret = btrfs_del_csums(trans, info, bytenr, num_bytes);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+ }
+
+ ret = add_to_free_space_tree(trans, bytenr, num_bytes);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
+
+ ret = btrfs_update_block_group(trans, bytenr, num_bytes, 0);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ goto out;
+ }
}
- trace_btrfs_reserved_extent_alloc(fs_info, ins->objectid, ins->offset);
+ btrfs_release_path(path);
+
+out:
+ btrfs_free_path(path);
return ret;
}
-static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_node *node,
- struct btrfs_delayed_extent_op *extent_op)
+/*
+ * when we free an block, it is possible (and likely) that we free the last
+ * delayed ref for that extent as well. This searches the delayed ref tree for
+ * a given extent, and if there are no other delayed refs to be processed, it
+ * removes it from the tree.
+ */
+static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
+ u64 bytenr)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- int ret;
- struct btrfs_extent_item *extent_item;
- struct btrfs_key extent_key;
- struct btrfs_tree_block_info *block_info;
- struct btrfs_extent_inline_ref *iref;
- struct btrfs_path *path;
- struct extent_buffer *leaf;
- struct btrfs_delayed_tree_ref *ref;
- u32 size = sizeof(*extent_item) + sizeof(*iref);
- u64 num_bytes;
- u64 flags = extent_op->flags_to_set;
- bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
-
- ref = btrfs_delayed_node_to_tree_ref(node);
-
- extent_key.objectid = node->bytenr;
- if (skinny_metadata) {
- extent_key.offset = ref->level;
- extent_key.type = BTRFS_METADATA_ITEM_KEY;
- num_bytes = fs_info->nodesize;
- } else {
- extent_key.offset = node->num_bytes;
- extent_key.type = BTRFS_EXTENT_ITEM_KEY;
- size += sizeof(*block_info);
- num_bytes = node->num_bytes;
- }
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
+ struct btrfs_delayed_ref_head *head;
+ struct btrfs_delayed_ref_root *delayed_refs;
+ int ret = 0;
- path->leave_spinning = 1;
- ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
- &extent_key, size);
- if (ret) {
- btrfs_free_path(path);
- return ret;
- }
+ delayed_refs = &trans->transaction->delayed_refs;
+ spin_lock(&delayed_refs->lock);
+ head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
+ if (!head)
+ goto out_delayed_unlock;
- leaf = path->nodes[0];
- extent_item = btrfs_item_ptr(leaf, path->slots[0],
- struct btrfs_extent_item);
- btrfs_set_extent_refs(leaf, extent_item, 1);
- btrfs_set_extent_generation(leaf, extent_item, trans->transid);
- btrfs_set_extent_flags(leaf, extent_item,
- flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
+ spin_lock(&head->lock);
+ if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root))
+ goto out;
- if (skinny_metadata) {
- iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
- } else {
- block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
- btrfs_set_tree_block_key(leaf, block_info, &extent_op->key);
- btrfs_set_tree_block_level(leaf, block_info, ref->level);
- iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
- }
+ if (cleanup_extent_op(head) != NULL)
+ goto out;
- if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) {
- BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
- btrfs_set_extent_inline_ref_type(leaf, iref,
- BTRFS_SHARED_BLOCK_REF_KEY);
- btrfs_set_extent_inline_ref_offset(leaf, iref, ref->parent);
- } else {
- btrfs_set_extent_inline_ref_type(leaf, iref,
- BTRFS_TREE_BLOCK_REF_KEY);
- btrfs_set_extent_inline_ref_offset(leaf, iref, ref->root);
- }
+ /*
+ * waiting for the lock here would deadlock. If someone else has it
+ * locked they are already in the process of dropping it anyway
+ */
+ if (!mutex_trylock(&head->mutex))
+ goto out;
- btrfs_mark_buffer_dirty(leaf);
- btrfs_free_path(path);
+ btrfs_delete_ref_head(delayed_refs, head);
+ head->processing = 0;
- ret = remove_from_free_space_tree(trans, extent_key.objectid,
- num_bytes);
- if (ret)
- return ret;
+ spin_unlock(&head->lock);
+ spin_unlock(&delayed_refs->lock);
- ret = update_block_group(trans, extent_key.objectid,
- fs_info->nodesize, 1);
- if (ret) { /* -ENOENT, logic error */
- btrfs_err(fs_info, "update block group failed for %llu %llu",
- extent_key.objectid, extent_key.offset);
- BUG();
- }
+ BUG_ON(head->extent_op);
+ if (head->must_insert_reserved)
+ ret = 1;
- trace_btrfs_reserved_extent_alloc(fs_info, extent_key.objectid,
- fs_info->nodesize);
+ btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head);
+ mutex_unlock(&head->mutex);
+ btrfs_put_delayed_ref_head(head);
return ret;
+out:
+ spin_unlock(&head->lock);
+
+out_delayed_unlock:
+ spin_unlock(&delayed_refs->lock);
+ return 0;
}
-int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
- struct btrfs_root *root, u64 owner,
- u64 offset, u64 ram_bytes,
- struct btrfs_key *ins)
+void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct extent_buffer *buf,
+ u64 parent, int last_ref)
{
+ struct btrfs_fs_info *fs_info = root->fs_info;
struct btrfs_ref generic_ref = { 0 };
+ int pin = 1;
int ret;
- BUG_ON(root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID);
+ btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF,
+ buf->start, buf->len, parent);
+ btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf),
+ root->root_key.objectid);
- btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
- ins->objectid, ins->offset, 0);
- btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, offset);
- btrfs_ref_tree_mod(root->fs_info, &generic_ref);
- ret = btrfs_add_delayed_data_ref(trans, &generic_ref,
- ram_bytes, NULL, NULL);
- return ret;
+ if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+ int old_ref_mod, new_ref_mod;
+
+ btrfs_ref_tree_mod(fs_info, &generic_ref);
+ ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL,
+ &old_ref_mod, &new_ref_mod);
+ BUG_ON(ret); /* -ENOMEM */
+ pin = old_ref_mod >= 0 && new_ref_mod < 0;
+ }
+
+ if (last_ref && btrfs_header_generation(buf) == trans->transid) {
+ struct btrfs_block_group_cache *cache;
+
+ if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+ ret = check_ref_cleanup(trans, buf->start);
+ if (!ret)
+ goto out;
+ }
+
+ pin = 0;
+ cache = btrfs_lookup_block_group(fs_info, buf->start);
+
+ if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
+ pin_down_extent(cache, buf->start, buf->len, 1);
+ btrfs_put_block_group(cache);
+ goto out;
+ }
+
+ WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
+
+ btrfs_add_free_space(cache, buf->start, buf->len);
+ btrfs_free_reserved_bytes(cache, buf->len, 0);
+ btrfs_put_block_group(cache);
+ trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len);
+ }
+out:
+ if (pin)
+ add_pinned_bytes(fs_info, &generic_ref);
+
+ if (last_ref) {
+ /*
+ * Deleting the buffer, clear the corrupt flag since it doesn't
+ * matter anymore.
+ */
+ clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
+ }
}
-/*
- * this is used by the tree logging recovery code. It records that
- * an extent has been allocated and makes sure to clear the free
- * space cache bits as well
- */
-int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
- u64 root_objectid, u64 owner, u64 offset,
- struct btrfs_key *ins)
+/* Can return -ENOMEM */
+int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
{
struct btrfs_fs_info *fs_info = trans->fs_info;
+ int old_ref_mod, new_ref_mod;
int ret;
- struct btrfs_block_group_cache *block_group;
- struct btrfs_space_info *space_info;
+
+ if (btrfs_is_testing(fs_info))
+ return 0;
/*
- * Mixed block groups will exclude before processing the log so we only
- * need to do the exclude dance if this fs isn't mixed.
+ * tree log blocks never actually go into the extent allocation
+ * tree, just update pinning info and exit early.
*/
- if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
- ret = __exclude_logged_extent(fs_info, ins->objectid,
- ins->offset);
- if (ret)
- return ret;
+ if ((ref->type == BTRFS_REF_METADATA &&
+ ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
+ (ref->type == BTRFS_REF_DATA &&
+ ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)) {
+ /* unlocks the pinned mutex */
+ btrfs_pin_extent(fs_info, ref->bytenr, ref->len, 1);
+ old_ref_mod = new_ref_mod = 0;
+ ret = 0;
+ } else if (ref->type == BTRFS_REF_METADATA) {
+ ret = btrfs_add_delayed_tree_ref(trans, ref, NULL,
+ &old_ref_mod, &new_ref_mod);
+ } else {
+ ret = btrfs_add_delayed_data_ref(trans, ref, 0,
+ &old_ref_mod, &new_ref_mod);
}
- block_group = btrfs_lookup_block_group(fs_info, ins->objectid);
- if (!block_group)
- return -EINVAL;
+ if (!((ref->type == BTRFS_REF_METADATA &&
+ ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
+ (ref->type == BTRFS_REF_DATA &&
+ ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)))
+ btrfs_ref_tree_mod(fs_info, ref);
- space_info = block_group->space_info;
- spin_lock(&space_info->lock);
- spin_lock(&block_group->lock);
- space_info->bytes_reserved += ins->offset;
- block_group->reserved += ins->offset;
- spin_unlock(&block_group->lock);
- spin_unlock(&space_info->lock);
+ if (ret == 0 && old_ref_mod >= 0 && new_ref_mod < 0)
+ add_pinned_bytes(fs_info, ref);
- ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner,
- offset, ins, 1);
- btrfs_put_block_group(block_group);
return ret;
}
-static struct extent_buffer *
-btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
- u64 bytenr, int level, u64 owner)
-{
- struct btrfs_fs_info *fs_info = root->fs_info;
- struct extent_buffer *buf;
-
- buf = btrfs_find_create_tree_block(fs_info, bytenr);
- if (IS_ERR(buf))
- return buf;
+enum btrfs_loop_type {
+ LOOP_CACHING_NOWAIT,
+ LOOP_CACHING_WAIT,
+ LOOP_ALLOC_CHUNK,
+ LOOP_NO_EMPTY_SIZE,
+};
- /*
- * Extra safety check in case the extent tree is corrupted and extent
- * allocator chooses to use a tree block which is already used and
- * locked.
- */
- if (buf->lock_owner == current->pid) {
- btrfs_err_rl(fs_info,
-"tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected",
- buf->start, btrfs_header_owner(buf), current->pid);
- free_extent_buffer(buf);
- return ERR_PTR(-EUCLEAN);
- }
+static inline void
+btrfs_lock_block_group(struct btrfs_block_group_cache *cache,
+ int delalloc)
+{
+ if (delalloc)
+ down_read(&cache->data_rwsem);
+}
- btrfs_set_buffer_lockdep_class(root->root_key.objectid, buf, level);
- btrfs_tree_lock(buf);
- btrfs_clean_tree_block(buf);
- clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
+static inline void
+btrfs_grab_block_group(struct btrfs_block_group_cache *cache,
+ int delalloc)
+{
+ btrfs_get_block_group(cache);
+ if (delalloc)
+ down_read(&cache->data_rwsem);
+}
- btrfs_set_lock_blocking_write(buf);
- set_extent_buffer_uptodate(buf);
+static struct btrfs_block_group_cache *
+btrfs_lock_cluster(struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster,
+ int delalloc)
+{
+ struct btrfs_block_group_cache *used_bg = NULL;
- memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header));
- btrfs_set_header_level(buf, level);
- btrfs_set_header_bytenr(buf, buf->start);
- btrfs_set_header_generation(buf, trans->transid);
- btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
- btrfs_set_header_owner(buf, owner);
- write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid);
- write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid);
- if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
- buf->log_index = root->log_transid % 2;
- /*
- * we allow two log transactions at a time, use different
- * EXTENT bit to differentiate dirty pages.
- */
- if (buf->log_index == 0)
- set_extent_dirty(&root->dirty_log_pages, buf->start,
- buf->start + buf->len - 1, GFP_NOFS);
- else
- set_extent_new(&root->dirty_log_pages, buf->start,
- buf->start + buf->len - 1);
- } else {
- buf->log_index = -1;
- set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
- buf->start + buf->len - 1, GFP_NOFS);
+ spin_lock(&cluster->refill_lock);
+ while (1) {
+ used_bg = cluster->block_group;
+ if (!used_bg)
+ return NULL;
+
+ if (used_bg == block_group)
+ return used_bg;
+
+ btrfs_get_block_group(used_bg);
+
+ if (!delalloc)
+ return used_bg;
+
+ if (down_read_trylock(&used_bg->data_rwsem))
+ return used_bg;
+
+ spin_unlock(&cluster->refill_lock);
+
+ /* We should only have one-level nested. */
+ down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING);
+
+ spin_lock(&cluster->refill_lock);
+ if (used_bg == cluster->block_group)
+ return used_bg;
+
+ up_read(&used_bg->data_rwsem);
+ btrfs_put_block_group(used_bg);
}
- trans->dirty = true;
- /* this returns a buffer locked for blocking */
- return buf;
+}
+
+static inline void
+btrfs_release_block_group(struct btrfs_block_group_cache *cache,
+ int delalloc)
+{
+ if (delalloc)
+ up_read(&cache->data_rwsem);
+ btrfs_put_block_group(cache);
}
/*
- * finds a free extent and does all the dirty work required for allocation
- * returns the tree buffer or an ERR_PTR on error.
+ * Structure used internally for find_free_extent() function. Wraps needed
+ * parameters.
*/
-struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- u64 parent, u64 root_objectid,
- const struct btrfs_disk_key *key,
- int level, u64 hint,
- u64 empty_size)
-{
- struct btrfs_fs_info *fs_info = root->fs_info;
- struct btrfs_key ins;
- struct btrfs_block_rsv *block_rsv;
- struct extent_buffer *buf;
- struct btrfs_delayed_extent_op *extent_op;
- struct btrfs_ref generic_ref = { 0 };
- u64 flags = 0;
- int ret;
- u32 blocksize = fs_info->nodesize;
- bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
+struct find_free_extent_ctl {
+ /* Basic allocation info */
+ u64 ram_bytes;
+ u64 num_bytes;
+ u64 empty_size;
+ u64 flags;
+ int delalloc;
-#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
- if (btrfs_is_testing(fs_info)) {
- buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
- level, root_objectid);
- if (!IS_ERR(buf))
- root->alloc_bytenr += blocksize;
- return buf;
- }
-#endif
+ /* Where to start the search inside the bg */
+ u64 search_start;
- block_rsv = btrfs_use_block_rsv(trans, root, blocksize);
- if (IS_ERR(block_rsv))
- return ERR_CAST(block_rsv);
+ /* For clustered allocation */
+ u64 empty_cluster;
- ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
- empty_size, hint, &ins, 0, 0);
- if (ret)
- goto out_unuse;
+ bool have_caching_bg;
+ bool orig_have_caching_bg;
- buf = btrfs_init_new_buffer(trans, root, ins.objectid, level,
- root_objectid);
- if (IS_ERR(buf)) {
- ret = PTR_ERR(buf);
- goto out_free_reserved;
- }
+ /* RAID index, converted from flags */
+ int index;
- if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
- if (parent == 0)
- parent = ins.objectid;
- flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
- } else
- BUG_ON(parent > 0);
+ /*
+ * Current loop number, check find_free_extent_update_loop() for details
+ */
+ int loop;
- if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
- extent_op = btrfs_alloc_delayed_extent_op();
- if (!extent_op) {
- ret = -ENOMEM;
- goto out_free_buf;
- }
- if (key)
- memcpy(&extent_op->key, key, sizeof(extent_op->key));
- else
- memset(&extent_op->key, 0, sizeof(extent_op->key));
- extent_op->flags_to_set = flags;
- extent_op->update_key = skinny_metadata ? false : true;
- extent_op->update_flags = true;
- extent_op->is_data = false;
- extent_op->level = level;
+ /*
+ * Whether we're refilling a cluster, if true we need to re-search
+ * current block group but don't try to refill the cluster again.
+ */
+ bool retry_clustered;
- btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
- ins.objectid, ins.offset, parent);
- generic_ref.real_root = root->root_key.objectid;
- btrfs_init_tree_ref(&generic_ref, level, root_objectid);
- btrfs_ref_tree_mod(fs_info, &generic_ref);
- ret = btrfs_add_delayed_tree_ref(trans, &generic_ref,
- extent_op, NULL, NULL);
- if (ret)
- goto out_free_delayed;
- }
- return buf;
+ /*
+ * Whether we're updating free space cache, if true we need to re-search
+ * current block group but don't try updating free space cache again.
+ */
+ bool retry_unclustered;
-out_free_delayed:
- btrfs_free_delayed_extent_op(extent_op);
-out_free_buf:
- free_extent_buffer(buf);
-out_free_reserved:
- btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0);
-out_unuse:
- btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize);
- return ERR_PTR(ret);
-}
+ /* If current block group is cached */
+ int cached;
-struct walk_control {
- u64 refs[BTRFS_MAX_LEVEL];
- u64 flags[BTRFS_MAX_LEVEL];
- struct btrfs_key update_progress;
- struct btrfs_key drop_progress;
- int drop_level;
- int stage;
- int level;
- int shared_level;
- int update_ref;
- int keep_locks;
- int reada_slot;
- int reada_count;
- int restarted;
+ /* Max contiguous hole found */
+ u64 max_extent_size;
+
+ /* Total free space from free space cache, not always contiguous */
+ u64 total_free_space;
+
+ /* Found result */
+ u64 found_offset;
};
-#define DROP_REFERENCE 1
-#define UPDATE_BACKREF 2
-static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct walk_control *wc,
- struct btrfs_path *path)
+/*
+ * Helper function for find_free_extent().
+ *
+ * Return -ENOENT to inform caller that we need fallback to unclustered mode.
+ * Return -EAGAIN to inform caller that we need to re-search this block group
+ * Return >0 to inform caller that we find nothing
+ * Return 0 means we have found a location and set ffe_ctl->found_offset.
+ */
+static int find_free_extent_clustered(struct btrfs_block_group_cache *bg,
+ struct btrfs_free_cluster *last_ptr,
+ struct find_free_extent_ctl *ffe_ctl,
+ struct btrfs_block_group_cache **cluster_bg_ret)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- u64 bytenr;
- u64 generation;
- u64 refs;
- u64 flags;
- u32 nritems;
- struct btrfs_key key;
- struct extent_buffer *eb;
+ struct btrfs_block_group_cache *cluster_bg;
+ u64 aligned_cluster;
+ u64 offset;
int ret;
- int slot;
- int nread = 0;
-
- if (path->slots[wc->level] < wc->reada_slot) {
- wc->reada_count = wc->reada_count * 2 / 3;
- wc->reada_count = max(wc->reada_count, 2);
- } else {
- wc->reada_count = wc->reada_count * 3 / 2;
- wc->reada_count = min_t(int, wc->reada_count,
- BTRFS_NODEPTRS_PER_BLOCK(fs_info));
- }
-
- eb = path->nodes[wc->level];
- nritems = btrfs_header_nritems(eb);
- for (slot = path->slots[wc->level]; slot < nritems; slot++) {
- if (nread >= wc->reada_count)
- break;
+ cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc);
+ if (!cluster_bg)
+ goto refill_cluster;
+ if (cluster_bg != bg && (cluster_bg->ro ||
+ !block_group_bits(cluster_bg, ffe_ctl->flags)))
+ goto release_cluster;
- cond_resched();
- bytenr = btrfs_node_blockptr(eb, slot);
- generation = btrfs_node_ptr_generation(eb, slot);
+ offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr,
+ ffe_ctl->num_bytes, cluster_bg->key.objectid,
+ &ffe_ctl->max_extent_size);
+ if (offset) {
+ /* We have a block, we're done */
+ spin_unlock(&last_ptr->refill_lock);
+ trace_btrfs_reserve_extent_cluster(cluster_bg,
+ ffe_ctl->search_start, ffe_ctl->num_bytes);
+ *cluster_bg_ret = cluster_bg;
+ ffe_ctl->found_offset = offset;
+ return 0;
+ }
+ WARN_ON(last_ptr->block_group != cluster_bg);
- if (slot == path->slots[wc->level])
- goto reada;
+release_cluster:
+ /*
+ * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so
+ * lets just skip it and let the allocator find whatever block it can
+ * find. If we reach this point, we will have tried the cluster
+ * allocator plenty of times and not have found anything, so we are
+ * likely way too fragmented for the clustering stuff to find anything.
+ *
+ * However, if the cluster is taken from the current block group,
+ * release the cluster first, so that we stand a better chance of
+ * succeeding in the unclustered allocation.
+ */
+ if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) {
+ spin_unlock(&last_ptr->refill_lock);
+ btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
+ return -ENOENT;
+ }
- if (wc->stage == UPDATE_BACKREF &&
- generation <= root->root_key.offset)
- continue;
+ /* This cluster didn't work out, free it and start over */
+ btrfs_return_cluster_to_free_space(NULL, last_ptr);
- /* We don't lock the tree block, it's OK to be racy here */
- ret = btrfs_lookup_extent_info(trans, fs_info, bytenr,
- wc->level - 1, 1, &refs,
- &flags);
- /* We don't care about errors in readahead. */
- if (ret < 0)
- continue;
- BUG_ON(refs == 0);
+ if (cluster_bg != bg)
+ btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
- if (wc->stage == DROP_REFERENCE) {
- if (refs == 1)
- goto reada;
+refill_cluster:
+ if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) {
+ spin_unlock(&last_ptr->refill_lock);
+ return -ENOENT;
+ }
- if (wc->level == 1 &&
- (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
- continue;
- if (!wc->update_ref ||
- generation <= root->root_key.offset)
- continue;
- btrfs_node_key_to_cpu(eb, &key, slot);
- ret = btrfs_comp_cpu_keys(&key,
- &wc->update_progress);
- if (ret < 0)
- continue;
- } else {
- if (wc->level == 1 &&
- (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
- continue;
+ aligned_cluster = max_t(u64,
+ ffe_ctl->empty_cluster + ffe_ctl->empty_size,
+ bg->full_stripe_len);
+ ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start,
+ ffe_ctl->num_bytes, aligned_cluster);
+ if (ret == 0) {
+ /* Now pull our allocation out of this cluster */
+ offset = btrfs_alloc_from_cluster(bg, last_ptr,
+ ffe_ctl->num_bytes, ffe_ctl->search_start,
+ &ffe_ctl->max_extent_size);
+ if (offset) {
+ /* We found one, proceed */
+ spin_unlock(&last_ptr->refill_lock);
+ trace_btrfs_reserve_extent_cluster(bg,
+ ffe_ctl->search_start,
+ ffe_ctl->num_bytes);
+ ffe_ctl->found_offset = offset;
+ return 0;
}
-reada:
- readahead_tree_block(fs_info, bytenr);
- nread++;
+ } else if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT &&
+ !ffe_ctl->retry_clustered) {
+ spin_unlock(&last_ptr->refill_lock);
+
+ ffe_ctl->retry_clustered = true;
+ btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
+ ffe_ctl->empty_cluster + ffe_ctl->empty_size);
+ return -EAGAIN;
}
- wc->reada_slot = slot;
+ /*
+ * At this point we either didn't find a cluster or we weren't able to
+ * allocate a block from our cluster. Free the cluster we've been
+ * trying to use, and go to the next block group.
+ */
+ btrfs_return_cluster_to_free_space(NULL, last_ptr);
+ spin_unlock(&last_ptr->refill_lock);
+ return 1;
}
/*
- * helper to process tree block while walking down the tree.
- *
- * when wc->stage == UPDATE_BACKREF, this function updates
- * back refs for pointers in the block.
- *
- * NOTE: return value 1 means we should stop walking down.
+ * Return >0 to inform caller that we find nothing
+ * Return 0 when we found an free extent and set ffe_ctrl->found_offset
+ * Return -EAGAIN to inform caller that we need to re-search this block group
*/
-static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- struct walk_control *wc, int lookup_info)
+static int find_free_extent_unclustered(struct btrfs_block_group_cache *bg,
+ struct btrfs_free_cluster *last_ptr,
+ struct find_free_extent_ctl *ffe_ctl)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- int level = wc->level;
- struct extent_buffer *eb = path->nodes[level];
- u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
- int ret;
-
- if (wc->stage == UPDATE_BACKREF &&
- btrfs_header_owner(eb) != root->root_key.objectid)
- return 1;
+ u64 offset;
/*
- * when reference count of tree block is 1, it won't increase
- * again. once full backref flag is set, we never clear it.
+ * We are doing an unclustered allocation, set the fragmented flag so
+ * we don't bother trying to setup a cluster again until we get more
+ * space.
*/
- if (lookup_info &&
- ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
- (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
- BUG_ON(!path->locks[level]);
- ret = btrfs_lookup_extent_info(trans, fs_info,
- eb->start, level, 1,
- &wc->refs[level],
- &wc->flags[level]);
- BUG_ON(ret == -ENOMEM);
- if (ret)
- return ret;
- BUG_ON(wc->refs[level] == 0);
+ if (unlikely(last_ptr)) {
+ spin_lock(&last_ptr->lock);
+ last_ptr->fragmented = 1;
+ spin_unlock(&last_ptr->lock);
}
+ if (ffe_ctl->cached) {
+ struct btrfs_free_space_ctl *free_space_ctl;
- if (wc->stage == DROP_REFERENCE) {
- if (wc->refs[level] > 1)
+ free_space_ctl = bg->free_space_ctl;
+ spin_lock(&free_space_ctl->tree_lock);
+ if (free_space_ctl->free_space <
+ ffe_ctl->num_bytes + ffe_ctl->empty_cluster +
+ ffe_ctl->empty_size) {
+ ffe_ctl->total_free_space = max_t(u64,
+ ffe_ctl->total_free_space,
+ free_space_ctl->free_space);
+ spin_unlock(&free_space_ctl->tree_lock);
return 1;
-
- if (path->locks[level] && !wc->keep_locks) {
- btrfs_tree_unlock_rw(eb, path->locks[level]);
- path->locks[level] = 0;
}
- return 0;
+ spin_unlock(&free_space_ctl->tree_lock);
}
- /* wc->stage == UPDATE_BACKREF */
- if (!(wc->flags[level] & flag)) {
- BUG_ON(!path->locks[level]);
- ret = btrfs_inc_ref(trans, root, eb, 1);
- BUG_ON(ret); /* -ENOMEM */
- ret = btrfs_dec_ref(trans, root, eb, 0);
- BUG_ON(ret); /* -ENOMEM */
- ret = btrfs_set_disk_extent_flags(trans, eb->start,
- eb->len, flag,
- btrfs_header_level(eb), 0);
- BUG_ON(ret); /* -ENOMEM */
- wc->flags[level] |= flag;
- }
+ offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start,
+ ffe_ctl->num_bytes, ffe_ctl->empty_size,
+ &ffe_ctl->max_extent_size);
/*
- * the block is shared by multiple trees, so it's not good to
- * keep the tree lock
+ * If we didn't find a chunk, and we haven't failed on this block group
+ * before, and this block group is in the middle of caching and we are
+ * ok with waiting, then go ahead and wait for progress to be made, and
+ * set @retry_unclustered to true.
+ *
+ * If @retry_unclustered is true then we've already waited on this
+ * block group once and should move on to the next block group.
*/
- if (path->locks[level] && level > 0) {
- btrfs_tree_unlock_rw(eb, path->locks[level]);
- path->locks[level] = 0;
+ if (!offset && !ffe_ctl->retry_unclustered && !ffe_ctl->cached &&
+ ffe_ctl->loop > LOOP_CACHING_NOWAIT) {
+ btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
+ ffe_ctl->empty_size);
+ ffe_ctl->retry_unclustered = true;
+ return -EAGAIN;
+ } else if (!offset) {
+ return 1;
}
+ ffe_ctl->found_offset = offset;
return 0;
}
/*
- * This is used to verify a ref exists for this root to deal with a bug where we
- * would have a drop_progress key that hadn't been updated properly.
+ * Return >0 means caller needs to re-search for free extent
+ * Return 0 means we have the needed free extent.
+ * Return <0 means we failed to locate any free extent.
*/
-static int check_ref_exists(struct btrfs_trans_handle *trans,
- struct btrfs_root *root, u64 bytenr, u64 parent,
- int level)
+static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info,
+ struct btrfs_free_cluster *last_ptr,
+ struct btrfs_key *ins,
+ struct find_free_extent_ctl *ffe_ctl,
+ int full_search, bool use_cluster)
{
- struct btrfs_path *path;
- struct btrfs_extent_inline_ref *iref;
+ struct btrfs_root *root = fs_info->extent_root;
int ret;
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
+ if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) &&
+ ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg)
+ ffe_ctl->orig_have_caching_bg = true;
+
+ if (!ins->objectid && ffe_ctl->loop >= LOOP_CACHING_WAIT &&
+ ffe_ctl->have_caching_bg)
+ return 1;
+
+ if (!ins->objectid && ++(ffe_ctl->index) < BTRFS_NR_RAID_TYPES)
+ return 1;
+
+ if (ins->objectid) {
+ if (!use_cluster && last_ptr) {
+ spin_lock(&last_ptr->lock);
+ last_ptr->window_start = ins->objectid;
+ spin_unlock(&last_ptr->lock);
+ }
+ return 0;
+ }
+
+ /*
+ * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
+ * caching kthreads as we move along
+ * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
+ * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
+ * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
+ * again
+ */
+ if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) {
+ ffe_ctl->index = 0;
+ if (ffe_ctl->loop == LOOP_CACHING_NOWAIT) {
+ /*
+ * We want to skip the LOOP_CACHING_WAIT step if we
+ * don't have any uncached bgs and we've already done a
+ * full search through.
+ */
+ if (ffe_ctl->orig_have_caching_bg || !full_search)
+ ffe_ctl->loop = LOOP_CACHING_WAIT;
+ else
+ ffe_ctl->loop = LOOP_ALLOC_CHUNK;
+ } else {
+ ffe_ctl->loop++;
+ }
+
+ if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) {
+ struct btrfs_trans_handle *trans;
+ int exist = 0;
+
+ trans = current->journal_info;
+ if (trans)
+ exist = 1;
+ else
+ trans = btrfs_join_transaction(root);
+
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ return ret;
+ }
+
+ ret = btrfs_chunk_alloc(trans, ffe_ctl->flags,
+ CHUNK_ALLOC_FORCE);
+
+ /*
+ * If we can't allocate a new chunk we've already looped
+ * through at least once, move on to the NO_EMPTY_SIZE
+ * case.
+ */
+ if (ret == -ENOSPC)
+ ffe_ctl->loop = LOOP_NO_EMPTY_SIZE;
- ret = lookup_extent_backref(trans, path, &iref, bytenr,
- root->fs_info->nodesize, parent,
- root->root_key.objectid, level, 0);
- btrfs_free_path(path);
- if (ret == -ENOENT)
- return 0;
- if (ret < 0)
- return ret;
- return 1;
+ /* Do not bail out on ENOSPC since we can do more. */
+ if (ret < 0 && ret != -ENOSPC)
+ btrfs_abort_transaction(trans, ret);
+ else
+ ret = 0;
+ if (!exist)
+ btrfs_end_transaction(trans);
+ if (ret)
+ return ret;
+ }
+
+ if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) {
+ /*
+ * Don't loop again if we already have no empty_size and
+ * no empty_cluster.
+ */
+ if (ffe_ctl->empty_size == 0 &&
+ ffe_ctl->empty_cluster == 0)
+ return -ENOSPC;
+ ffe_ctl->empty_size = 0;
+ ffe_ctl->empty_cluster = 0;
+ }
+ return 1;
+ }
+ return -ENOSPC;
}
/*
- * helper to process tree block pointer.
+ * walks the btree of allocated extents and find a hole of a given size.
+ * The key ins is changed to record the hole:
+ * ins->objectid == start position
+ * ins->flags = BTRFS_EXTENT_ITEM_KEY
+ * ins->offset == the size of the hole.
+ * Any available blocks before search_start are skipped.
*
- * when wc->stage == DROP_REFERENCE, this function checks
- * reference count of the block pointed to. if the block
- * is shared and we need update back refs for the subtree
- * rooted at the block, this function changes wc->stage to
- * UPDATE_BACKREF. if the block is shared and there is no
- * need to update back, this function drops the reference
- * to the block.
+ * If there is no suitable free space, we will record the max size of
+ * the free space extent currently.
*
- * NOTE: return value 1 means we should stop walking down.
+ * The overall logic and call chain:
+ *
+ * find_free_extent()
+ * |- Iterate through all block groups
+ * | |- Get a valid block group
+ * | |- Try to do clustered allocation in that block group
+ * | |- Try to do unclustered allocation in that block group
+ * | |- Check if the result is valid
+ * | | |- If valid, then exit
+ * | |- Jump to next block group
+ * |
+ * |- Push harder to find free extents
+ * |- If not found, re-iterate all block groups
*/
-static noinline int do_walk_down(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- struct walk_control *wc, int *lookup_info)
+static noinline int find_free_extent(struct btrfs_fs_info *fs_info,
+ u64 ram_bytes, u64 num_bytes, u64 empty_size,
+ u64 hint_byte, struct btrfs_key *ins,
+ u64 flags, int delalloc)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- u64 bytenr;
- u64 generation;
- u64 parent;
- struct btrfs_key key;
- struct btrfs_key first_key;
- struct btrfs_ref ref = { 0 };
- struct extent_buffer *next;
- int level = wc->level;
- int reada = 0;
int ret = 0;
- bool need_account = false;
-
- generation = btrfs_node_ptr_generation(path->nodes[level],
- path->slots[level]);
- /*
- * if the lower level block was created before the snapshot
- * was created, we know there is no need to update back refs
- * for the subtree
- */
- if (wc->stage == UPDATE_BACKREF &&
- generation <= root->root_key.offset) {
- *lookup_info = 1;
- return 1;
- }
+ struct btrfs_free_cluster *last_ptr = NULL;
+ struct btrfs_block_group_cache *block_group = NULL;
+ struct find_free_extent_ctl ffe_ctl = {0};
+ struct btrfs_space_info *space_info;
+ bool use_cluster = true;
+ bool full_search = false;
- bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
- btrfs_node_key_to_cpu(path->nodes[level], &first_key,
- path->slots[level]);
+ WARN_ON(num_bytes < fs_info->sectorsize);
- next = find_extent_buffer(fs_info, bytenr);
- if (!next) {
- next = btrfs_find_create_tree_block(fs_info, bytenr);
- if (IS_ERR(next))
- return PTR_ERR(next);
+ ffe_ctl.ram_bytes = ram_bytes;
+ ffe_ctl.num_bytes = num_bytes;
+ ffe_ctl.empty_size = empty_size;
+ ffe_ctl.flags = flags;
+ ffe_ctl.search_start = 0;
+ ffe_ctl.retry_clustered = false;
+ ffe_ctl.retry_unclustered = false;
+ ffe_ctl.delalloc = delalloc;
+ ffe_ctl.index = btrfs_bg_flags_to_raid_index(flags);
+ ffe_ctl.have_caching_bg = false;
+ ffe_ctl.orig_have_caching_bg = false;
+ ffe_ctl.found_offset = 0;
- btrfs_set_buffer_lockdep_class(root->root_key.objectid, next,
- level - 1);
- reada = 1;
- }
- btrfs_tree_lock(next);
- btrfs_set_lock_blocking_write(next);
+ ins->type = BTRFS_EXTENT_ITEM_KEY;
+ ins->objectid = 0;
+ ins->offset = 0;
- ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1,
- &wc->refs[level - 1],
- &wc->flags[level - 1]);
- if (ret < 0)
- goto out_unlock;
+ trace_find_free_extent(fs_info, num_bytes, empty_size, flags);
- if (unlikely(wc->refs[level - 1] == 0)) {
- btrfs_err(fs_info, "Missing references.");
- ret = -EIO;
- goto out_unlock;
+ space_info = btrfs_find_space_info(fs_info, flags);
+ if (!space_info) {
+ btrfs_err(fs_info, "No space info for %llu", flags);
+ return -ENOSPC;
}
- *lookup_info = 0;
-
- if (wc->stage == DROP_REFERENCE) {
- if (wc->refs[level - 1] > 1) {
- need_account = true;
- if (level == 1 &&
- (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
- goto skip;
-
- if (!wc->update_ref ||
- generation <= root->root_key.offset)
- goto skip;
-
- btrfs_node_key_to_cpu(path->nodes[level], &key,
- path->slots[level]);
- ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
- if (ret < 0)
- goto skip;
- wc->stage = UPDATE_BACKREF;
- wc->shared_level = level - 1;
+ /*
+ * If our free space is heavily fragmented we may not be able to make
+ * big contiguous allocations, so instead of doing the expensive search
+ * for free space, simply return ENOSPC with our max_extent_size so we
+ * can go ahead and search for a more manageable chunk.
+ *
+ * If our max_extent_size is large enough for our allocation simply
+ * disable clustering since we will likely not be able to find enough
+ * space to create a cluster and induce latency trying.
+ */
+ if (unlikely(space_info->max_extent_size)) {
+ spin_lock(&space_info->lock);
+ if (space_info->max_extent_size &&
+ num_bytes > space_info->max_extent_size) {
+ ins->offset = space_info->max_extent_size;
+ spin_unlock(&space_info->lock);
+ return -ENOSPC;
+ } else if (space_info->max_extent_size) {
+ use_cluster = false;
}
- } else {
- if (level == 1 &&
- (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
- goto skip;
+ spin_unlock(&space_info->lock);
}
- if (!btrfs_buffer_uptodate(next, generation, 0)) {
- btrfs_tree_unlock(next);
- free_extent_buffer(next);
- next = NULL;
- *lookup_info = 1;
+ last_ptr = fetch_cluster_info(fs_info, space_info,
+ &ffe_ctl.empty_cluster);
+ if (last_ptr) {
+ spin_lock(&last_ptr->lock);
+ if (last_ptr->block_group)
+ hint_byte = last_ptr->window_start;
+ if (last_ptr->fragmented) {
+ /*
+ * We still set window_start so we can keep track of the
+ * last place we found an allocation to try and save
+ * some time.
+ */
+ hint_byte = last_ptr->window_start;
+ use_cluster = false;
+ }
+ spin_unlock(&last_ptr->lock);
}
- if (!next) {
- if (reada && level == 1)
- reada_walk_down(trans, root, wc, path);
- next = read_tree_block(fs_info, bytenr, generation, level - 1,
- &first_key);
- if (IS_ERR(next)) {
- return PTR_ERR(next);
- } else if (!extent_buffer_uptodate(next)) {
- free_extent_buffer(next);
- return -EIO;
+ ffe_ctl.search_start = max(ffe_ctl.search_start,
+ first_logical_byte(fs_info, 0));
+ ffe_ctl.search_start = max(ffe_ctl.search_start, hint_byte);
+ if (ffe_ctl.search_start == hint_byte) {
+ block_group = btrfs_lookup_block_group(fs_info,
+ ffe_ctl.search_start);
+ /*
+ * we don't want to use the block group if it doesn't match our
+ * allocation bits, or if its not cached.
+ *
+ * However if we are re-searching with an ideal block group
+ * picked out then we don't care that the block group is cached.
+ */
+ if (block_group && block_group_bits(block_group, flags) &&
+ block_group->cached != BTRFS_CACHE_NO) {
+ down_read(&space_info->groups_sem);
+ if (list_empty(&block_group->list) ||
+ block_group->ro) {
+ /*
+ * someone is removing this block group,
+ * we can't jump into the have_block_group
+ * target because our list pointers are not
+ * valid
+ */
+ btrfs_put_block_group(block_group);
+ up_read(&space_info->groups_sem);
+ } else {
+ ffe_ctl.index = btrfs_bg_flags_to_raid_index(
+ block_group->flags);
+ btrfs_lock_block_group(block_group, delalloc);
+ goto have_block_group;
+ }
+ } else if (block_group) {
+ btrfs_put_block_group(block_group);
}
- btrfs_tree_lock(next);
- btrfs_set_lock_blocking_write(next);
}
+search:
+ ffe_ctl.have_caching_bg = false;
+ if (ffe_ctl.index == btrfs_bg_flags_to_raid_index(flags) ||
+ ffe_ctl.index == 0)
+ full_search = true;
+ down_read(&space_info->groups_sem);
+ list_for_each_entry(block_group,
+ &space_info->block_groups[ffe_ctl.index], list) {
+ /* If the block group is read-only, we can skip it entirely. */
+ if (unlikely(block_group->ro))
+ continue;
- level--;
- ASSERT(level == btrfs_header_level(next));
- if (level != btrfs_header_level(next)) {
- btrfs_err(root->fs_info, "mismatched level");
- ret = -EIO;
- goto out_unlock;
- }
- path->nodes[level] = next;
- path->slots[level] = 0;
- path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
- wc->level = level;
- if (wc->level == 1)
- wc->reada_slot = 0;
- return 0;
-skip:
- wc->refs[level - 1] = 0;
- wc->flags[level - 1] = 0;
- if (wc->stage == DROP_REFERENCE) {
- if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
- parent = path->nodes[level]->start;
- } else {
- ASSERT(root->root_key.objectid ==
- btrfs_header_owner(path->nodes[level]));
- if (root->root_key.objectid !=
- btrfs_header_owner(path->nodes[level])) {
- btrfs_err(root->fs_info,
- "mismatched block owner");
- ret = -EIO;
- goto out_unlock;
- }
- parent = 0;
- }
+ btrfs_grab_block_group(block_group, delalloc);
+ ffe_ctl.search_start = block_group->key.objectid;
/*
- * If we had a drop_progress we need to verify the refs are set
- * as expected. If we find our ref then we know that from here
- * on out everything should be correct, and we can clear the
- * ->restarted flag.
+ * this can happen if we end up cycling through all the
+ * raid types, but we want to make sure we only allocate
+ * for the proper type.
*/
- if (wc->restarted) {
- ret = check_ref_exists(trans, root, bytenr, parent,
- level - 1);
- if (ret < 0)
- goto out_unlock;
- if (ret == 0)
- goto no_delete;
+ if (!block_group_bits(block_group, flags)) {
+ u64 extra = BTRFS_BLOCK_GROUP_DUP |
+ BTRFS_BLOCK_GROUP_RAID1_MASK |
+ BTRFS_BLOCK_GROUP_RAID56_MASK |
+ BTRFS_BLOCK_GROUP_RAID10;
+
+ /*
+ * if they asked for extra copies and this block group
+ * doesn't provide them, bail. This does allow us to
+ * fill raid0 from raid1.
+ */
+ if ((flags & extra) && !(block_group->flags & extra))
+ goto loop;
+
+ /*
+ * This block group has different flags than we want.
+ * It's possible that we have MIXED_GROUP flag but no
+ * block group is mixed. Just skip such block group.
+ */
+ btrfs_release_block_group(block_group, delalloc);
+ continue;
+ }
+
+have_block_group:
+ ffe_ctl.cached = btrfs_block_group_cache_done(block_group);
+ if (unlikely(!ffe_ctl.cached)) {
+ ffe_ctl.have_caching_bg = true;
+ ret = btrfs_cache_block_group(block_group, 0);
+ BUG_ON(ret < 0);
ret = 0;
- wc->restarted = 0;
}
+ if (unlikely(block_group->cached == BTRFS_CACHE_ERROR))
+ goto loop;
+
/*
- * Reloc tree doesn't contribute to qgroup numbers, and we have
- * already accounted them at merge time (replace_path),
- * thus we could skip expensive subtree trace here.
+ * Ok we want to try and use the cluster allocator, so
+ * lets look there
*/
- if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
- need_account) {
- ret = btrfs_qgroup_trace_subtree(trans, next,
- generation, level - 1);
- if (ret) {
- btrfs_err_rl(fs_info,
- "Error %d accounting shared subtree. Quota is out of sync, rescan required.",
- ret);
+ if (last_ptr && use_cluster) {
+ struct btrfs_block_group_cache *cluster_bg = NULL;
+
+ ret = find_free_extent_clustered(block_group, last_ptr,
+ &ffe_ctl, &cluster_bg);
+
+ if (ret == 0) {
+ if (cluster_bg && cluster_bg != block_group) {
+ btrfs_release_block_group(block_group,
+ delalloc);
+ block_group = cluster_bg;
+ }
+ goto checks;
+ } else if (ret == -EAGAIN) {
+ goto have_block_group;
+ } else if (ret > 0) {
+ goto loop;
}
+ /* ret == -ENOENT case falls through */
}
- /*
- * We need to update the next key in our walk control so we can
- * update the drop_progress key accordingly. We don't care if
- * find_next_key doesn't find a key because that means we're at
- * the end and are going to clean up now.
- */
- wc->drop_level = level;
- find_next_key(path, level, &wc->drop_progress);
+ ret = find_free_extent_unclustered(block_group, last_ptr,
+ &ffe_ctl);
+ if (ret == -EAGAIN)
+ goto have_block_group;
+ else if (ret > 0)
+ goto loop;
+ /* ret == 0 case falls through */
+checks:
+ ffe_ctl.search_start = round_up(ffe_ctl.found_offset,
+ fs_info->stripesize);
- btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
- fs_info->nodesize, parent);
- btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid);
- ret = btrfs_free_extent(trans, &ref);
- if (ret)
- goto out_unlock;
+ /* move on to the next group */
+ if (ffe_ctl.search_start + num_bytes >
+ block_group->key.objectid + block_group->key.offset) {
+ btrfs_add_free_space(block_group, ffe_ctl.found_offset,
+ num_bytes);
+ goto loop;
+ }
+
+ if (ffe_ctl.found_offset < ffe_ctl.search_start)
+ btrfs_add_free_space(block_group, ffe_ctl.found_offset,
+ ffe_ctl.search_start - ffe_ctl.found_offset);
+
+ ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
+ num_bytes, delalloc);
+ if (ret == -EAGAIN) {
+ btrfs_add_free_space(block_group, ffe_ctl.found_offset,
+ num_bytes);
+ goto loop;
+ }
+ btrfs_inc_block_group_reservations(block_group);
+
+ /* we are all good, lets return */
+ ins->objectid = ffe_ctl.search_start;
+ ins->offset = num_bytes;
+
+ trace_btrfs_reserve_extent(block_group, ffe_ctl.search_start,
+ num_bytes);
+ btrfs_release_block_group(block_group, delalloc);
+ break;
+loop:
+ ffe_ctl.retry_clustered = false;
+ ffe_ctl.retry_unclustered = false;
+ BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) !=
+ ffe_ctl.index);
+ btrfs_release_block_group(block_group, delalloc);
+ cond_resched();
}
-no_delete:
- *lookup_info = 1;
- ret = 1;
+ up_read(&space_info->groups_sem);
-out_unlock:
- btrfs_tree_unlock(next);
- free_extent_buffer(next);
+ ret = find_free_extent_update_loop(fs_info, last_ptr, ins, &ffe_ctl,
+ full_search, use_cluster);
+ if (ret > 0)
+ goto search;
+ if (ret == -ENOSPC) {
+ /*
+ * Use ffe_ctl->total_free_space as fallback if we can't find
+ * any contiguous hole.
+ */
+ if (!ffe_ctl.max_extent_size)
+ ffe_ctl.max_extent_size = ffe_ctl.total_free_space;
+ spin_lock(&space_info->lock);
+ space_info->max_extent_size = ffe_ctl.max_extent_size;
+ spin_unlock(&space_info->lock);
+ ins->offset = ffe_ctl.max_extent_size;
+ }
return ret;
}
/*
- * helper to process tree block while walking up the tree.
+ * btrfs_reserve_extent - entry point to the extent allocator. Tries to find a
+ * hole that is at least as big as @num_bytes.
*
- * when wc->stage == DROP_REFERENCE, this function drops
- * reference count on the block.
+ * @root - The root that will contain this extent
+ *
+ * @ram_bytes - The amount of space in ram that @num_bytes take. This
+ * is used for accounting purposes. This value differs
+ * from @num_bytes only in the case of compressed extents.
+ *
+ * @num_bytes - Number of bytes to allocate on-disk.
+ *
+ * @min_alloc_size - Indicates the minimum amount of space that the
+ * allocator should try to satisfy. In some cases
+ * @num_bytes may be larger than what is required and if
+ * the filesystem is fragmented then allocation fails.
+ * However, the presence of @min_alloc_size gives a
+ * chance to try and satisfy the smaller allocation.
+ *
+ * @empty_size - A hint that you plan on doing more COW. This is the
+ * size in bytes the allocator should try to find free
+ * next to the block it returns. This is just a hint and
+ * may be ignored by the allocator.
+ *
+ * @hint_byte - Hint to the allocator to start searching above the byte
+ * address passed. It might be ignored.
+ *
+ * @ins - This key is modified to record the found hole. It will
+ * have the following values:
+ * ins->objectid == start position
+ * ins->flags = BTRFS_EXTENT_ITEM_KEY
+ * ins->offset == the size of the hole.
+ *
+ * @is_data - Boolean flag indicating whether an extent is
+ * allocated for data (true) or metadata (false)
+ *
+ * @delalloc - Boolean flag indicating whether this allocation is for
+ * delalloc or not. If 'true' data_rwsem of block groups
+ * is going to be acquired.
*
- * when wc->stage == UPDATE_BACKREF, this function changes
- * wc->stage back to DROP_REFERENCE if we changed wc->stage
- * to UPDATE_BACKREF previously while processing the block.
*
- * NOTE: return value 1 means we should stop walking up.
+ * Returns 0 when an allocation succeeded or < 0 when an error occurred. In
+ * case -ENOSPC is returned then @ins->offset will contain the size of the
+ * largest available hole the allocator managed to find.
*/
-static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- struct walk_control *wc)
+int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
+ u64 num_bytes, u64 min_alloc_size,
+ u64 empty_size, u64 hint_byte,
+ struct btrfs_key *ins, int is_data, int delalloc)
{
struct btrfs_fs_info *fs_info = root->fs_info;
+ bool final_tried = num_bytes == min_alloc_size;
+ u64 flags;
int ret;
- int level = wc->level;
- struct extent_buffer *eb = path->nodes[level];
- u64 parent = 0;
-
- if (wc->stage == UPDATE_BACKREF) {
- BUG_ON(wc->shared_level < level);
- if (level < wc->shared_level)
- goto out;
-
- ret = find_next_key(path, level + 1, &wc->update_progress);
- if (ret > 0)
- wc->update_ref = 0;
-
- wc->stage = DROP_REFERENCE;
- wc->shared_level = -1;
- path->slots[level] = 0;
- /*
- * check reference count again if the block isn't locked.
- * we should start walking down the tree again if reference
- * count is one.
- */
- if (!path->locks[level]) {
- BUG_ON(level == 0);
- btrfs_tree_lock(eb);
- btrfs_set_lock_blocking_write(eb);
- path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
+ flags = get_alloc_profile_by_root(root, is_data);
+again:
+ WARN_ON(num_bytes < fs_info->sectorsize);
+ ret = find_free_extent(fs_info, ram_bytes, num_bytes, empty_size,
+ hint_byte, ins, flags, delalloc);
+ if (!ret && !is_data) {
+ btrfs_dec_block_group_reservations(fs_info, ins->objectid);
+ } else if (ret == -ENOSPC) {
+ if (!final_tried && ins->offset) {
+ num_bytes = min(num_bytes >> 1, ins->offset);
+ num_bytes = round_down(num_bytes,
+ fs_info->sectorsize);
+ num_bytes = max(num_bytes, min_alloc_size);
+ ram_bytes = num_bytes;
+ if (num_bytes == min_alloc_size)
+ final_tried = true;
+ goto again;
+ } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
+ struct btrfs_space_info *sinfo;
- ret = btrfs_lookup_extent_info(trans, fs_info,
- eb->start, level, 1,
- &wc->refs[level],
- &wc->flags[level]);
- if (ret < 0) {
- btrfs_tree_unlock_rw(eb, path->locks[level]);
- path->locks[level] = 0;
- return ret;
- }
- BUG_ON(wc->refs[level] == 0);
- if (wc->refs[level] == 1) {
- btrfs_tree_unlock_rw(eb, path->locks[level]);
- path->locks[level] = 0;
- return 1;
- }
+ sinfo = btrfs_find_space_info(fs_info, flags);
+ btrfs_err(fs_info,
+ "allocation failed flags %llu, wanted %llu",
+ flags, num_bytes);
+ if (sinfo)
+ btrfs_dump_space_info(fs_info, sinfo,
+ num_bytes, 1);
}
}
- /* wc->stage == DROP_REFERENCE */
- BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
+ return ret;
+}
- if (wc->refs[level] == 1) {
- if (level == 0) {
- if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
- ret = btrfs_dec_ref(trans, root, eb, 1);
- else
- ret = btrfs_dec_ref(trans, root, eb, 0);
- BUG_ON(ret); /* -ENOMEM */
- if (is_fstree(root->root_key.objectid)) {
- ret = btrfs_qgroup_trace_leaf_items(trans, eb);
- if (ret) {
- btrfs_err_rl(fs_info,
- "error %d accounting leaf items, quota is out of sync, rescan required",
- ret);
- }
- }
- }
- /* make block locked assertion in btrfs_clean_tree_block happy */
- if (!path->locks[level] &&
- btrfs_header_generation(eb) == trans->transid) {
- btrfs_tree_lock(eb);
- btrfs_set_lock_blocking_write(eb);
- path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
- }
- btrfs_clean_tree_block(eb);
+static int __btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
+ u64 start, u64 len,
+ int pin, int delalloc)
+{
+ struct btrfs_block_group_cache *cache;
+ int ret = 0;
+
+ cache = btrfs_lookup_block_group(fs_info, start);
+ if (!cache) {
+ btrfs_err(fs_info, "Unable to find block group for %llu",
+ start);
+ return -ENOSPC;
}
- if (eb == root->node) {
- if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
- parent = eb->start;
- else if (root->root_key.objectid != btrfs_header_owner(eb))
- goto owner_mismatch;
- } else {
- if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
- parent = path->nodes[level + 1]->start;
- else if (root->root_key.objectid !=
- btrfs_header_owner(path->nodes[level + 1]))
- goto owner_mismatch;
+ if (pin)
+ pin_down_extent(cache, start, len, 1);
+ else {
+ if (btrfs_test_opt(fs_info, DISCARD))
+ ret = btrfs_discard_extent(fs_info, start, len, NULL);
+ btrfs_add_free_space(cache, start, len);
+ btrfs_free_reserved_bytes(cache, len, delalloc);
+ trace_btrfs_reserved_extent_free(fs_info, start, len);
}
- btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
-out:
- wc->refs[level] = 0;
- wc->flags[level] = 0;
- return 0;
+ btrfs_put_block_group(cache);
+ return ret;
+}
-owner_mismatch:
- btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu",
- btrfs_header_owner(eb), root->root_key.objectid);
- return -EUCLEAN;
+int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
+ u64 start, u64 len, int delalloc)
+{
+ return __btrfs_free_reserved_extent(fs_info, start, len, 0, delalloc);
}
-static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- struct walk_control *wc)
+int btrfs_free_and_pin_reserved_extent(struct btrfs_fs_info *fs_info,
+ u64 start, u64 len)
{
- int level = wc->level;
- int lookup_info = 1;
+ return __btrfs_free_reserved_extent(fs_info, start, len, 1, 0);
+}
+
+static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
+ u64 parent, u64 root_objectid,
+ u64 flags, u64 owner, u64 offset,
+ struct btrfs_key *ins, int ref_mod)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
int ret;
+ struct btrfs_extent_item *extent_item;
+ struct btrfs_extent_inline_ref *iref;
+ struct btrfs_path *path;
+ struct extent_buffer *leaf;
+ int type;
+ u32 size;
- while (level >= 0) {
- ret = walk_down_proc(trans, root, path, wc, lookup_info);
- if (ret > 0)
- break;
+ if (parent > 0)
+ type = BTRFS_SHARED_DATA_REF_KEY;
+ else
+ type = BTRFS_EXTENT_DATA_REF_KEY;
- if (level == 0)
- break;
+ size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
- if (path->slots[level] >=
- btrfs_header_nritems(path->nodes[level]))
- break;
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
- ret = do_walk_down(trans, root, path, wc, &lookup_info);
- if (ret > 0) {
- path->slots[level]++;
- continue;
- } else if (ret < 0)
- return ret;
- level = wc->level;
+ path->leave_spinning = 1;
+ ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
+ ins, size);
+ if (ret) {
+ btrfs_free_path(path);
+ return ret;
}
- return 0;
-}
-static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- struct walk_control *wc, int max_level)
-{
- int level = wc->level;
- int ret;
+ leaf = path->nodes[0];
+ extent_item = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_item);
+ btrfs_set_extent_refs(leaf, extent_item, ref_mod);
+ btrfs_set_extent_generation(leaf, extent_item, trans->transid);
+ btrfs_set_extent_flags(leaf, extent_item,
+ flags | BTRFS_EXTENT_FLAG_DATA);
- path->slots[level] = btrfs_header_nritems(path->nodes[level]);
- while (level < max_level && path->nodes[level]) {
- wc->level = level;
- if (path->slots[level] + 1 <
- btrfs_header_nritems(path->nodes[level])) {
- path->slots[level]++;
- return 0;
- } else {
- ret = walk_up_proc(trans, root, path, wc);
- if (ret > 0)
- return 0;
- if (ret < 0)
- return ret;
+ iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
+ btrfs_set_extent_inline_ref_type(leaf, iref, type);
+ if (parent > 0) {
+ struct btrfs_shared_data_ref *ref;
+ ref = (struct btrfs_shared_data_ref *)(iref + 1);
+ btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
+ btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
+ } else {
+ struct btrfs_extent_data_ref *ref;
+ ref = (struct btrfs_extent_data_ref *)(&iref->offset);
+ btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
+ btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
+ btrfs_set_extent_data_ref_offset(leaf, ref, offset);
+ btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
+ }
+
+ btrfs_mark_buffer_dirty(path->nodes[0]);
+ btrfs_free_path(path);
- if (path->locks[level]) {
- btrfs_tree_unlock_rw(path->nodes[level],
- path->locks[level]);
- path->locks[level] = 0;
- }
- free_extent_buffer(path->nodes[level]);
- path->nodes[level] = NULL;
- level++;
- }
+ ret = remove_from_free_space_tree(trans, ins->objectid, ins->offset);
+ if (ret)
+ return ret;
+
+ ret = btrfs_update_block_group(trans, ins->objectid, ins->offset, 1);
+ if (ret) { /* -ENOENT, logic error */
+ btrfs_err(fs_info, "update block group failed for %llu %llu",
+ ins->objectid, ins->offset);
+ BUG();
}
- return 1;
+ trace_btrfs_reserved_extent_alloc(fs_info, ins->objectid, ins->offset);
+ return ret;
}
-/*
- * drop a subvolume tree.
- *
- * this function traverses the tree freeing any blocks that only
- * referenced by the tree.
- *
- * when a shared tree block is found. this function decreases its
- * reference count by one. if update_ref is true, this function
- * also make sure backrefs for the shared block and all lower level
- * blocks are properly updated.
- *
- * If called with for_reloc == 0, may exit early with -EAGAIN
- */
-int btrfs_drop_snapshot(struct btrfs_root *root,
- struct btrfs_block_rsv *block_rsv, int update_ref,
- int for_reloc)
+static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_node *node,
+ struct btrfs_delayed_extent_op *extent_op)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- struct btrfs_path *path;
- struct btrfs_trans_handle *trans;
- struct btrfs_root *tree_root = fs_info->tree_root;
- struct btrfs_root_item *root_item = &root->root_item;
- struct walk_control *wc;
- struct btrfs_key key;
- int err = 0;
+ struct btrfs_fs_info *fs_info = trans->fs_info;
int ret;
- int level;
- bool root_dropped = false;
+ struct btrfs_extent_item *extent_item;
+ struct btrfs_key extent_key;
+ struct btrfs_tree_block_info *block_info;
+ struct btrfs_extent_inline_ref *iref;
+ struct btrfs_path *path;
+ struct extent_buffer *leaf;
+ struct btrfs_delayed_tree_ref *ref;
+ u32 size = sizeof(*extent_item) + sizeof(*iref);
+ u64 num_bytes;
+ u64 flags = extent_op->flags_to_set;
+ bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
- btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid);
+ ref = btrfs_delayed_node_to_tree_ref(node);
- path = btrfs_alloc_path();
- if (!path) {
- err = -ENOMEM;
- goto out;
+ extent_key.objectid = node->bytenr;
+ if (skinny_metadata) {
+ extent_key.offset = ref->level;
+ extent_key.type = BTRFS_METADATA_ITEM_KEY;
+ num_bytes = fs_info->nodesize;
+ } else {
+ extent_key.offset = node->num_bytes;
+ extent_key.type = BTRFS_EXTENT_ITEM_KEY;
+ size += sizeof(*block_info);
+ num_bytes = node->num_bytes;
}
- wc = kzalloc(sizeof(*wc), GFP_NOFS);
- if (!wc) {
- btrfs_free_path(path);
- err = -ENOMEM;
- goto out;
- }
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
- trans = btrfs_start_transaction(tree_root, 0);
- if (IS_ERR(trans)) {
- err = PTR_ERR(trans);
- goto out_free;
+ path->leave_spinning = 1;
+ ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
+ &extent_key, size);
+ if (ret) {
+ btrfs_free_path(path);
+ return ret;
}
- err = btrfs_run_delayed_items(trans);
- if (err)
- goto out_end_trans;
+ leaf = path->nodes[0];
+ extent_item = btrfs_item_ptr(leaf, path->slots[0],
+ struct btrfs_extent_item);
+ btrfs_set_extent_refs(leaf, extent_item, 1);
+ btrfs_set_extent_generation(leaf, extent_item, trans->transid);
+ btrfs_set_extent_flags(leaf, extent_item,
+ flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
- if (block_rsv)
- trans->block_rsv = block_rsv;
+ if (skinny_metadata) {
+ iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
+ } else {
+ block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
+ btrfs_set_tree_block_key(leaf, block_info, &extent_op->key);
+ btrfs_set_tree_block_level(leaf, block_info, ref->level);
+ iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
+ }
- /*
- * This will help us catch people modifying the fs tree while we're
- * dropping it. It is unsafe to mess with the fs tree while it's being
- * dropped as we unlock the root node and parent nodes as we walk down
- * the tree, assuming nothing will change. If something does change
- * then we'll have stale information and drop references to blocks we've
- * already dropped.
- */
- set_bit(BTRFS_ROOT_DELETING, &root->state);
- if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
- level = btrfs_header_level(root->node);
- path->nodes[level] = btrfs_lock_root_node(root);
- btrfs_set_lock_blocking_write(path->nodes[level]);
- path->slots[level] = 0;
- path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
- memset(&wc->update_progress, 0,
- sizeof(wc->update_progress));
+ if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) {
+ BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
+ btrfs_set_extent_inline_ref_type(leaf, iref,
+ BTRFS_SHARED_BLOCK_REF_KEY);
+ btrfs_set_extent_inline_ref_offset(leaf, iref, ref->parent);
} else {
- btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
- memcpy(&wc->update_progress, &key,
- sizeof(wc->update_progress));
+ btrfs_set_extent_inline_ref_type(leaf, iref,
+ BTRFS_TREE_BLOCK_REF_KEY);
+ btrfs_set_extent_inline_ref_offset(leaf, iref, ref->root);
+ }
- level = root_item->drop_level;
- BUG_ON(level == 0);
- path->lowest_level = level;
- ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
- path->lowest_level = 0;
- if (ret < 0) {
- err = ret;
- goto out_end_trans;
- }
- WARN_ON(ret > 0);
+ btrfs_mark_buffer_dirty(leaf);
+ btrfs_free_path(path);
- /*
- * unlock our path, this is safe because only this
- * function is allowed to delete this snapshot
- */
- btrfs_unlock_up_safe(path, 0);
+ ret = remove_from_free_space_tree(trans, extent_key.objectid,
+ num_bytes);
+ if (ret)
+ return ret;
- level = btrfs_header_level(root->node);
- while (1) {
- btrfs_tree_lock(path->nodes[level]);
- btrfs_set_lock_blocking_write(path->nodes[level]);
- path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
+ ret = btrfs_update_block_group(trans, extent_key.objectid,
+ fs_info->nodesize, 1);
+ if (ret) { /* -ENOENT, logic error */
+ btrfs_err(fs_info, "update block group failed for %llu %llu",
+ extent_key.objectid, extent_key.offset);
+ BUG();
+ }
- ret = btrfs_lookup_extent_info(trans, fs_info,
- path->nodes[level]->start,
- level, 1, &wc->refs[level],
- &wc->flags[level]);
- if (ret < 0) {
- err = ret;
- goto out_end_trans;
- }
- BUG_ON(wc->refs[level] == 0);
+ trace_btrfs_reserved_extent_alloc(fs_info, extent_key.objectid,
+ fs_info->nodesize);
+ return ret;
+}
- if (level == root_item->drop_level)
- break;
+int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u64 owner,
+ u64 offset, u64 ram_bytes,
+ struct btrfs_key *ins)
+{
+ struct btrfs_ref generic_ref = { 0 };
+ int ret;
- btrfs_tree_unlock(path->nodes[level]);
- path->locks[level] = 0;
- WARN_ON(wc->refs[level] != 1);
- level--;
- }
- }
+ BUG_ON(root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID);
- wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
- wc->level = level;
- wc->shared_level = -1;
- wc->stage = DROP_REFERENCE;
- wc->update_ref = update_ref;
- wc->keep_locks = 0;
- wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
+ btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
+ ins->objectid, ins->offset, 0);
+ btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, offset);
+ btrfs_ref_tree_mod(root->fs_info, &generic_ref);
+ ret = btrfs_add_delayed_data_ref(trans, &generic_ref,
+ ram_bytes, NULL, NULL);
+ return ret;
+}
- while (1) {
+/*
+ * this is used by the tree logging recovery code. It records that
+ * an extent has been allocated and makes sure to clear the free
+ * space cache bits as well
+ */
+int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
+ u64 root_objectid, u64 owner, u64 offset,
+ struct btrfs_key *ins)
+{
+ struct btrfs_fs_info *fs_info = trans->fs_info;
+ int ret;
+ struct btrfs_block_group_cache *block_group;
+ struct btrfs_space_info *space_info;
- ret = walk_down_tree(trans, root, path, wc);
- if (ret < 0) {
- err = ret;
- break;
- }
+ /*
+ * Mixed block groups will exclude before processing the log so we only
+ * need to do the exclude dance if this fs isn't mixed.
+ */
+ if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
+ ret = __exclude_logged_extent(fs_info, ins->objectid,
+ ins->offset);
+ if (ret)
+ return ret;
+ }
- ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
- if (ret < 0) {
- err = ret;
- break;
- }
+ block_group = btrfs_lookup_block_group(fs_info, ins->objectid);
+ if (!block_group)
+ return -EINVAL;
- if (ret > 0) {
- BUG_ON(wc->stage != DROP_REFERENCE);
- break;
- }
+ space_info = block_group->space_info;
+ spin_lock(&space_info->lock);
+ spin_lock(&block_group->lock);
+ space_info->bytes_reserved += ins->offset;
+ block_group->reserved += ins->offset;
+ spin_unlock(&block_group->lock);
+ spin_unlock(&space_info->lock);
- if (wc->stage == DROP_REFERENCE) {
- wc->drop_level = wc->level;
- btrfs_node_key_to_cpu(path->nodes[wc->drop_level],
- &wc->drop_progress,
- path->slots[wc->drop_level]);
- }
- btrfs_cpu_key_to_disk(&root_item->drop_progress,
- &wc->drop_progress);
- root_item->drop_level = wc->drop_level;
+ ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner,
+ offset, ins, 1);
+ btrfs_put_block_group(block_group);
+ return ret;
+}
- BUG_ON(wc->level == 0);
- if (btrfs_should_end_transaction(trans) ||
- (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) {
- ret = btrfs_update_root(trans, tree_root,
- &root->root_key,
- root_item);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- err = ret;
- goto out_end_trans;
- }
+static struct extent_buffer *
+btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+ u64 bytenr, int level, u64 owner)
+{
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ struct extent_buffer *buf;
- btrfs_end_transaction_throttle(trans);
- if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) {
- btrfs_debug(fs_info,
- "drop snapshot early exit");
- err = -EAGAIN;
- goto out_free;
- }
+ buf = btrfs_find_create_tree_block(fs_info, bytenr);
+ if (IS_ERR(buf))
+ return buf;
- trans = btrfs_start_transaction(tree_root, 0);
- if (IS_ERR(trans)) {
- err = PTR_ERR(trans);
- goto out_free;
- }
- if (block_rsv)
- trans->block_rsv = block_rsv;
- }
+ /*
+ * Extra safety check in case the extent tree is corrupted and extent
+ * allocator chooses to use a tree block which is already used and
+ * locked.
+ */
+ if (buf->lock_owner == current->pid) {
+ btrfs_err_rl(fs_info,
+"tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected",
+ buf->start, btrfs_header_owner(buf), current->pid);
+ free_extent_buffer(buf);
+ return ERR_PTR(-EUCLEAN);
}
- btrfs_release_path(path);
- if (err)
- goto out_end_trans;
- ret = btrfs_del_root(trans, &root->root_key);
- if (ret) {
- btrfs_abort_transaction(trans, ret);
- err = ret;
- goto out_end_trans;
- }
+ btrfs_set_buffer_lockdep_class(root->root_key.objectid, buf, level);
+ btrfs_tree_lock(buf);
+ btrfs_clean_tree_block(buf);
+ clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
- if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
- ret = btrfs_find_root(tree_root, &root->root_key, path,
- NULL, NULL);
- if (ret < 0) {
- btrfs_abort_transaction(trans, ret);
- err = ret;
- goto out_end_trans;
- } else if (ret > 0) {
- /* if we fail to delete the orphan item this time
- * around, it'll get picked up the next time.
- *
- * The most common failure here is just -ENOENT.
- */
- btrfs_del_orphan_item(trans, tree_root,
- root->root_key.objectid);
- }
- }
+ btrfs_set_lock_blocking_write(buf);
+ set_extent_buffer_uptodate(buf);
- if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state)) {
- btrfs_add_dropped_root(trans, root);
+ memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header));
+ btrfs_set_header_level(buf, level);
+ btrfs_set_header_bytenr(buf, buf->start);
+ btrfs_set_header_generation(buf, trans->transid);
+ btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
+ btrfs_set_header_owner(buf, owner);
+ write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid);
+ write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid);
+ if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
+ buf->log_index = root->log_transid % 2;
+ /*
+ * we allow two log transactions at a time, use different
+ * EXTENT bit to differentiate dirty pages.
+ */
+ if (buf->log_index == 0)
+ set_extent_dirty(&root->dirty_log_pages, buf->start,
+ buf->start + buf->len - 1, GFP_NOFS);
+ else
+ set_extent_new(&root->dirty_log_pages, buf->start,
+ buf->start + buf->len - 1);
} else {
- free_extent_buffer(root->node);
- free_extent_buffer(root->commit_root);
- btrfs_put_fs_root(root);
+ buf->log_index = -1;
+ set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
+ buf->start + buf->len - 1, GFP_NOFS);
}
- root_dropped = true;
-out_end_trans:
- btrfs_end_transaction_throttle(trans);
-out_free:
- kfree(wc);
- btrfs_free_path(path);
-out:
- /*
- * So if we need to stop dropping the snapshot for whatever reason we
- * need to make sure to add it back to the dead root list so that we
- * keep trying to do the work later. This also cleans up roots if we
- * don't have it in the radix (like when we recover after a power fail
- * or unmount) so we don't leak memory.
- */
- if (!for_reloc && !root_dropped)
- btrfs_add_dead_root(root);
- if (err && err != -EAGAIN)
- btrfs_handle_fs_error(fs_info, err, NULL);
- return err;
+ trans->dirty = true;
+ /* this returns a buffer locked for blocking */
+ return buf;
}
/*
- * drop subtree rooted at tree block 'node'.
- *
- * NOTE: this function will unlock and release tree block 'node'
- * only used by relocation code
+ * finds a free extent and does all the dirty work required for allocation
+ * returns the tree buffer or an ERR_PTR on error.
*/
-int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct extent_buffer *node,
- struct extent_buffer *parent)
+struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 parent, u64 root_objectid,
+ const struct btrfs_disk_key *key,
+ int level, u64 hint,
+ u64 empty_size)
{
struct btrfs_fs_info *fs_info = root->fs_info;
- struct btrfs_path *path;
- struct walk_control *wc;
- int level;
- int parent_level;
- int ret = 0;
- int wret;
-
- BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
+ struct btrfs_key ins;
+ struct btrfs_block_rsv *block_rsv;
+ struct extent_buffer *buf;
+ struct btrfs_delayed_extent_op *extent_op;
+ struct btrfs_ref generic_ref = { 0 };
+ u64 flags = 0;
+ int ret;
+ u32 blocksize = fs_info->nodesize;
+ bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
- wc = kzalloc(sizeof(*wc), GFP_NOFS);
- if (!wc) {
- btrfs_free_path(path);
- return -ENOMEM;
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+ if (btrfs_is_testing(fs_info)) {
+ buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
+ level, root_objectid);
+ if (!IS_ERR(buf))
+ root->alloc_bytenr += blocksize;
+ return buf;
}
+#endif
- btrfs_assert_tree_locked(parent);
- parent_level = btrfs_header_level(parent);
- extent_buffer_get(parent);
- path->nodes[parent_level] = parent;
- path->slots[parent_level] = btrfs_header_nritems(parent);
+ block_rsv = btrfs_use_block_rsv(trans, root, blocksize);
+ if (IS_ERR(block_rsv))
+ return ERR_CAST(block_rsv);
- btrfs_assert_tree_locked(node);
- level = btrfs_header_level(node);
- path->nodes[level] = node;
- path->slots[level] = 0;
- path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
+ ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
+ empty_size, hint, &ins, 0, 0);
+ if (ret)
+ goto out_unuse;
- wc->refs[parent_level] = 1;
- wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
- wc->level = level;
- wc->shared_level = -1;
- wc->stage = DROP_REFERENCE;
- wc->update_ref = 0;
- wc->keep_locks = 1;
- wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
+ buf = btrfs_init_new_buffer(trans, root, ins.objectid, level,
+ root_objectid);
+ if (IS_ERR(buf)) {
+ ret = PTR_ERR(buf);
+ goto out_free_reserved;
+ }
- while (1) {
- wret = walk_down_tree(trans, root, path, wc);
- if (wret < 0) {
- ret = wret;
- break;
+ if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
+ if (parent == 0)
+ parent = ins.objectid;
+ flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
+ } else
+ BUG_ON(parent > 0);
+
+ if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
+ extent_op = btrfs_alloc_delayed_extent_op();
+ if (!extent_op) {
+ ret = -ENOMEM;
+ goto out_free_buf;
}
+ if (key)
+ memcpy(&extent_op->key, key, sizeof(extent_op->key));
+ else
+ memset(&extent_op->key, 0, sizeof(extent_op->key));
+ extent_op->flags_to_set = flags;
+ extent_op->update_key = skinny_metadata ? false : true;
+ extent_op->update_flags = true;
+ extent_op->is_data = false;
+ extent_op->level = level;
- wret = walk_up_tree(trans, root, path, wc, parent_level);
- if (wret < 0)
- ret = wret;
- if (wret != 0)
- break;
+ btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
+ ins.objectid, ins.offset, parent);
+ generic_ref.real_root = root->root_key.objectid;
+ btrfs_init_tree_ref(&generic_ref, level, root_objectid);
+ btrfs_ref_tree_mod(fs_info, &generic_ref);
+ ret = btrfs_add_delayed_tree_ref(trans, &generic_ref,
+ extent_op, NULL, NULL);
+ if (ret)
+ goto out_free_delayed;
}
+ return buf;
- kfree(wc);
- btrfs_free_path(path);
- return ret;
+out_free_delayed:
+ btrfs_free_delayed_extent_op(extent_op);
+out_free_buf:
+ free_extent_buffer(buf);
+out_free_reserved:
+ btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0);
+out_unuse:
+ btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize);
+ return ERR_PTR(ret);
}
-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;
+struct walk_control {
+ u64 refs[BTRFS_MAX_LEVEL];
+ u64 flags[BTRFS_MAX_LEVEL];
+ struct btrfs_key update_progress;
+ struct btrfs_key drop_progress;
+ int drop_level;
+ int stage;
+ int level;
+ int shared_level;
+ int update_ref;
+ int keep_locks;
+ int reada_slot;
+ int reada_count;
+ int restarted;
+};
- if (num_devices == 1) {
- stripped |= BTRFS_BLOCK_GROUP_DUP;
- stripped = flags & ~stripped;
+#define DROP_REFERENCE 1
+#define UPDATE_BACKREF 2
- /* turn raid0 into single device chunks */
- if (flags & BTRFS_BLOCK_GROUP_RAID0)
- return stripped;
+static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct walk_control *wc,
+ struct btrfs_path *path)
+{
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ u64 bytenr;
+ u64 generation;
+ u64 refs;
+ u64 flags;
+ u32 nritems;
+ struct btrfs_key key;
+ struct extent_buffer *eb;
+ int ret;
+ int slot;
+ int nread = 0;
- /* turn mirroring into duplication */
- if (flags & (BTRFS_BLOCK_GROUP_RAID1_MASK |
- BTRFS_BLOCK_GROUP_RAID10))
- return stripped | BTRFS_BLOCK_GROUP_DUP;
+ if (path->slots[wc->level] < wc->reada_slot) {
+ wc->reada_count = wc->reada_count * 2 / 3;
+ wc->reada_count = max(wc->reada_count, 2);
} 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 */
+ wc->reada_count = wc->reada_count * 3 / 2;
+ wc->reada_count = min_t(int, wc->reada_count,
+ BTRFS_NODEPTRS_PER_BLOCK(fs_info));
}
- return flags;
-}
+ eb = path->nodes[wc->level];
+ nritems = btrfs_header_nritems(eb);
-static int inc_block_group_ro(struct btrfs_block_group_cache *cache, int force)
-{
- struct btrfs_space_info *sinfo = cache->space_info;
- u64 num_bytes;
- u64 sinfo_used;
- u64 min_allocable_bytes;
- int ret = -ENOSPC;
+ for (slot = path->slots[wc->level]; slot < nritems; slot++) {
+ if (nread >= wc->reada_count)
+ break;
- /*
- * We need some metadata space and system metadata space for
- * allocating chunks in some corner cases until we force to set
- * it to be readonly.
- */
- if ((sinfo->flags &
- (BTRFS_BLOCK_GROUP_SYSTEM | BTRFS_BLOCK_GROUP_METADATA)) &&
- !force)
- min_allocable_bytes = SZ_1M;
- else
- min_allocable_bytes = 0;
+ cond_resched();
+ bytenr = btrfs_node_blockptr(eb, slot);
+ generation = btrfs_node_ptr_generation(eb, slot);
- spin_lock(&sinfo->lock);
- spin_lock(&cache->lock);
+ if (slot == path->slots[wc->level])
+ goto reada;
- if (cache->ro) {
- cache->ro++;
- ret = 0;
- goto out;
- }
+ if (wc->stage == UPDATE_BACKREF &&
+ generation <= root->root_key.offset)
+ continue;
- num_bytes = cache->key.offset - cache->reserved - cache->pinned -
- cache->bytes_super - btrfs_block_group_used(&cache->item);
- sinfo_used = btrfs_space_info_used(sinfo, true);
+ /* We don't lock the tree block, it's OK to be racy here */
+ ret = btrfs_lookup_extent_info(trans, fs_info, bytenr,
+ wc->level - 1, 1, &refs,
+ &flags);
+ /* We don't care about errors in readahead. */
+ if (ret < 0)
+ continue;
+ BUG_ON(refs == 0);
- if (sinfo_used + num_bytes + min_allocable_bytes <=
- sinfo->total_bytes) {
- sinfo->bytes_readonly += num_bytes;
- cache->ro++;
- list_add_tail(&cache->ro_list, &sinfo->ro_bgs);
- ret = 0;
- }
-out:
- spin_unlock(&cache->lock);
- spin_unlock(&sinfo->lock);
- if (ret == -ENOSPC && btrfs_test_opt(cache->fs_info, ENOSPC_DEBUG)) {
- btrfs_info(cache->fs_info,
- "unable to make block group %llu ro",
- cache->key.objectid);
- btrfs_info(cache->fs_info,
- "sinfo_used=%llu bg_num_bytes=%llu min_allocable=%llu",
- sinfo_used, num_bytes, min_allocable_bytes);
- btrfs_dump_space_info(cache->fs_info, cache->space_info, 0, 0);
+ if (wc->stage == DROP_REFERENCE) {
+ if (refs == 1)
+ goto reada;
+
+ if (wc->level == 1 &&
+ (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+ continue;
+ if (!wc->update_ref ||
+ generation <= root->root_key.offset)
+ continue;
+ btrfs_node_key_to_cpu(eb, &key, slot);
+ ret = btrfs_comp_cpu_keys(&key,
+ &wc->update_progress);
+ if (ret < 0)
+ continue;
+ } else {
+ if (wc->level == 1 &&
+ (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+ continue;
+ }
+reada:
+ readahead_tree_block(fs_info, bytenr);
+ nread++;
}
- return ret;
+ wc->reada_slot = slot;
}
-int btrfs_inc_block_group_ro(struct btrfs_block_group_cache *cache)
-
+/*
+ * helper to process tree block while walking down the tree.
+ *
+ * when wc->stage == UPDATE_BACKREF, this function updates
+ * back refs for pointers in the block.
+ *
+ * NOTE: return value 1 means we should stop walking down.
+ */
+static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct walk_control *wc, int lookup_info)
{
- struct btrfs_fs_info *fs_info = cache->fs_info;
- struct btrfs_trans_handle *trans;
- u64 alloc_flags;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ int level = wc->level;
+ struct extent_buffer *eb = path->nodes[level];
+ u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
int ret;
-again:
- trans = btrfs_join_transaction(fs_info->extent_root);
- if (IS_ERR(trans))
- return PTR_ERR(trans);
+ if (wc->stage == UPDATE_BACKREF &&
+ btrfs_header_owner(eb) != root->root_key.objectid)
+ return 1;
/*
- * 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
+ * when reference count of tree block is 1, it won't increase
+ * again. once full backref flag is set, we never clear it.
*/
- 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 (lookup_info &&
+ ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
+ (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
+ BUG_ON(!path->locks[level]);
+ ret = btrfs_lookup_extent_info(trans, fs_info,
+ eb->start, level, 1,
+ &wc->refs[level],
+ &wc->flags[level]);
+ BUG_ON(ret == -ENOMEM);
if (ret)
return ret;
- goto again;
- }
-
- /*
- * 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);
- if (alloc_flags != cache->flags) {
- ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
- /*
- * ENOSPC is allowed here, we may have enough space
- * already allocated at the new raid level to
- * carry on
- */
- if (ret == -ENOSPC)
- ret = 0;
- if (ret < 0)
- goto out;
- }
-
- ret = inc_block_group_ro(cache, 0);
- if (!ret)
- goto out;
- alloc_flags = get_alloc_profile(fs_info, cache->space_info->flags);
- ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
- if (ret < 0)
- goto out;
- ret = inc_block_group_ro(cache, 0);
-out:
- if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {
- alloc_flags = update_block_group_flags(fs_info, cache->flags);
- mutex_lock(&fs_info->chunk_mutex);
- check_system_chunk(trans, alloc_flags);
- mutex_unlock(&fs_info->chunk_mutex);
+ BUG_ON(wc->refs[level] == 0);
}
- mutex_unlock(&fs_info->ro_block_group_mutex);
-
- btrfs_end_transaction(trans);
- return ret;
-}
-
-int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type)
-{
- u64 alloc_flags = get_alloc_profile(trans->fs_info, type);
-
- return btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
-}
-
-/*
- * helper to account the unused space of all the readonly block group in the
- * space_info. takes mirrors into account.
- */
-u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
-{
- struct btrfs_block_group_cache *block_group;
- u64 free_bytes = 0;
- int factor;
-
- /* It's df, we don't care if it's racy */
- if (list_empty(&sinfo->ro_bgs))
- return 0;
- spin_lock(&sinfo->lock);
- list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
- spin_lock(&block_group->lock);
+ if (wc->stage == DROP_REFERENCE) {
+ if (wc->refs[level] > 1)
+ return 1;
- if (!block_group->ro) {
- spin_unlock(&block_group->lock);
- continue;
+ if (path->locks[level] && !wc->keep_locks) {
+ btrfs_tree_unlock_rw(eb, path->locks[level]);
+ path->locks[level] = 0;
}
+ return 0;
+ }
- factor = btrfs_bg_type_to_factor(block_group->flags);
- free_bytes += (block_group->key.offset -
- btrfs_block_group_used(&block_group->item)) *
- factor;
-
- spin_unlock(&block_group->lock);
+ /* wc->stage == UPDATE_BACKREF */
+ if (!(wc->flags[level] & flag)) {
+ BUG_ON(!path->locks[level]);
+ ret = btrfs_inc_ref(trans, root, eb, 1);
+ BUG_ON(ret); /* -ENOMEM */
+ ret = btrfs_dec_ref(trans, root, eb, 0);
+ BUG_ON(ret); /* -ENOMEM */
+ ret = btrfs_set_disk_extent_flags(trans, eb->start,
+ eb->len, flag,
+ btrfs_header_level(eb), 0);
+ BUG_ON(ret); /* -ENOMEM */
+ wc->flags[level] |= flag;
}
- spin_unlock(&sinfo->lock);
- return free_bytes;
+ /*
+ * the block is shared by multiple trees, so it's not good to
+ * keep the tree lock
+ */
+ if (path->locks[level] && level > 0) {
+ btrfs_tree_unlock_rw(eb, path->locks[level]);
+ path->locks[level] = 0;
+ }
+ return 0;
}
-void btrfs_dec_block_group_ro(struct btrfs_block_group_cache *cache)
+/*
+ * This is used to verify a ref exists for this root to deal with a bug where we
+ * would have a drop_progress key that hadn't been updated properly.
+ */
+static int check_ref_exists(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u64 bytenr, u64 parent,
+ int level)
{
- struct btrfs_space_info *sinfo = cache->space_info;
- u64 num_bytes;
+ struct btrfs_path *path;
+ struct btrfs_extent_inline_ref *iref;
+ int ret;
- BUG_ON(!cache->ro);
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
- spin_lock(&sinfo->lock);
- spin_lock(&cache->lock);
- if (!--cache->ro) {
- num_bytes = cache->key.offset - cache->reserved -
- cache->pinned - cache->bytes_super -
- btrfs_block_group_used(&cache->item);
- sinfo->bytes_readonly -= num_bytes;
- list_del_init(&cache->ro_list);
- }
- spin_unlock(&cache->lock);
- spin_unlock(&sinfo->lock);
+ ret = lookup_extent_backref(trans, path, &iref, bytenr,
+ root->fs_info->nodesize, parent,
+ root->root_key.objectid, level, 0);
+ btrfs_free_path(path);
+ if (ret == -ENOENT)
+ return 0;
+ if (ret < 0)
+ return ret;
+ return 1;
}
/*
- * Checks to see if it's even possible to relocate this block group.
+ * helper to process tree block pointer.
*
- * @return - -1 if it's not a good idea to relocate this block group, 0 if its
- * ok to go ahead and try.
+ * when wc->stage == DROP_REFERENCE, this function checks
+ * reference count of the block pointed to. if the block
+ * is shared and we need update back refs for the subtree
+ * rooted at the block, this function changes wc->stage to
+ * UPDATE_BACKREF. if the block is shared and there is no
+ * need to update back, this function drops the reference
+ * to the block.
+ *
+ * NOTE: return value 1 means we should stop walking down.
*/
-int btrfs_can_relocate(struct btrfs_fs_info *fs_info, u64 bytenr)
+static noinline int do_walk_down(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct walk_control *wc, int *lookup_info)
{
- struct btrfs_block_group_cache *block_group;
- struct btrfs_space_info *space_info;
- struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
- struct btrfs_device *device;
- u64 min_free;
- u64 dev_min = 1;
- u64 dev_nr = 0;
- u64 target;
- int debug;
- int index;
- int full = 0;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ u64 bytenr;
+ u64 generation;
+ u64 parent;
+ struct btrfs_key key;
+ struct btrfs_key first_key;
+ struct btrfs_ref ref = { 0 };
+ struct extent_buffer *next;
+ int level = wc->level;
+ int reada = 0;
int ret = 0;
+ bool need_account = false;
- debug = btrfs_test_opt(fs_info, ENOSPC_DEBUG);
-
- block_group = btrfs_lookup_block_group(fs_info, bytenr);
-
- /* odd, couldn't find the block group, leave it alone */
- if (!block_group) {
- if (debug)
- btrfs_warn(fs_info,
- "can't find block group for bytenr %llu",
- bytenr);
- return -1;
- }
-
- min_free = btrfs_block_group_used(&block_group->item);
-
- /* no bytes used, we're good */
- if (!min_free)
- goto out;
-
- space_info = block_group->space_info;
- spin_lock(&space_info->lock);
-
- full = space_info->full;
-
+ generation = btrfs_node_ptr_generation(path->nodes[level],
+ path->slots[level]);
/*
- * if this is the last block group we have in this space, we can't
- * relocate it unless we're able to allocate a new chunk below.
- *
- * Otherwise, we need to make sure we have room in the space to handle
- * all of the extents from this block group. If we can, we're good
+ * if the lower level block was created before the snapshot
+ * was created, we know there is no need to update back refs
+ * for the subtree
*/
- if ((space_info->total_bytes != block_group->key.offset) &&
- (btrfs_space_info_used(space_info, false) + min_free <
- space_info->total_bytes)) {
- spin_unlock(&space_info->lock);
- goto out;
+ if (wc->stage == UPDATE_BACKREF &&
+ generation <= root->root_key.offset) {
+ *lookup_info = 1;
+ return 1;
}
- spin_unlock(&space_info->lock);
- /*
- * ok we don't have enough space, but maybe we have free space on our
- * devices to allocate new chunks for relocation, so loop through our
- * alloc devices and guess if we have enough space. if this block
- * group is going to be restriped, run checks against the target
- * profile instead of the current one.
- */
- ret = -1;
+ bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
+ btrfs_node_key_to_cpu(path->nodes[level], &first_key,
+ path->slots[level]);
- /*
- * index:
- * 0: raid10
- * 1: raid1
- * 2: dup
- * 3: raid0
- * 4: single
- */
- target = get_restripe_target(fs_info, block_group->flags);
- if (target) {
- index = btrfs_bg_flags_to_raid_index(extended_to_chunk(target));
- } else {
- /*
- * this is just a balance, so if we were marked as full
- * we know there is no space for a new chunk
- */
- if (full) {
- if (debug)
- btrfs_warn(fs_info,
- "no space to alloc new chunk for block group %llu",
- block_group->key.objectid);
- goto out;
- }
+ next = find_extent_buffer(fs_info, bytenr);
+ if (!next) {
+ next = btrfs_find_create_tree_block(fs_info, bytenr);
+ if (IS_ERR(next))
+ return PTR_ERR(next);
- index = btrfs_bg_flags_to_raid_index(block_group->flags);
+ btrfs_set_buffer_lockdep_class(root->root_key.objectid, next,
+ level - 1);
+ reada = 1;
}
+ btrfs_tree_lock(next);
+ btrfs_set_lock_blocking_write(next);
+
+ ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1,
+ &wc->refs[level - 1],
+ &wc->flags[level - 1]);
+ if (ret < 0)
+ goto out_unlock;
- if (index == BTRFS_RAID_RAID10) {
- dev_min = 4;
- /* Divide by 2 */
- min_free >>= 1;
- } else if (index == BTRFS_RAID_RAID1) {
- dev_min = 2;
- } else if (index == BTRFS_RAID_DUP) {
- /* Multiply by 2 */
- min_free <<= 1;
- } else if (index == BTRFS_RAID_RAID0) {
- dev_min = fs_devices->rw_devices;
- min_free = div64_u64(min_free, dev_min);
+ if (unlikely(wc->refs[level - 1] == 0)) {
+ btrfs_err(fs_info, "Missing references.");
+ ret = -EIO;
+ goto out_unlock;
}
+ *lookup_info = 0;
- mutex_lock(&fs_info->chunk_mutex);
- list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
- u64 dev_offset;
+ if (wc->stage == DROP_REFERENCE) {
+ if (wc->refs[level - 1] > 1) {
+ need_account = true;
+ if (level == 1 &&
+ (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+ goto skip;
- /*
- * check to make sure we can actually find a chunk with enough
- * space to fit our block group in.
- */
- if (device->total_bytes > device->bytes_used + min_free &&
- !test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
- ret = find_free_dev_extent(device, min_free,
- &dev_offset, NULL);
- if (!ret)
- dev_nr++;
+ if (!wc->update_ref ||
+ generation <= root->root_key.offset)
+ goto skip;
- if (dev_nr >= dev_min)
- break;
+ btrfs_node_key_to_cpu(path->nodes[level], &key,
+ path->slots[level]);
+ ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
+ if (ret < 0)
+ goto skip;
- ret = -1;
+ wc->stage = UPDATE_BACKREF;
+ wc->shared_level = level - 1;
}
+ } else {
+ if (level == 1 &&
+ (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+ goto skip;
}
- if (debug && ret == -1)
- btrfs_warn(fs_info,
- "no space to allocate a new chunk for block group %llu",
- block_group->key.objectid);
- mutex_unlock(&fs_info->chunk_mutex);
-out:
- btrfs_put_block_group(block_group);
- 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 = fs_info->extent_root;
- int ret = 0;
- struct btrfs_key found_key;
- struct extent_buffer *leaf;
- struct btrfs_block_group_item bg;
- u64 flags;
- int slot;
- ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
- if (ret < 0)
- goto out;
+ if (!btrfs_buffer_uptodate(next, generation, 0)) {
+ btrfs_tree_unlock(next);
+ free_extent_buffer(next);
+ next = NULL;
+ *lookup_info = 1;
+ }
- 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;
+ if (!next) {
+ if (reada && level == 1)
+ reada_walk_down(trans, root, wc, path);
+ next = read_tree_block(fs_info, bytenr, generation, level - 1,
+ &first_key);
+ if (IS_ERR(next)) {
+ return PTR_ERR(next);
+ } else if (!extent_buffer_uptodate(next)) {
+ free_extent_buffer(next);
+ return -EIO;
}
- btrfs_item_key_to_cpu(leaf, &found_key, slot);
-
- 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_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;
- }
+ btrfs_tree_lock(next);
+ btrfs_set_lock_blocking_write(next);
+ }
+
+ level--;
+ ASSERT(level == btrfs_header_level(next));
+ if (level != btrfs_header_level(next)) {
+ btrfs_err(root->fs_info, "mismatched level");
+ ret = -EIO;
+ goto out_unlock;
+ }
+ path->nodes[level] = next;
+ path->slots[level] = 0;
+ path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
+ wc->level = level;
+ if (wc->level == 1)
+ wc->reada_slot = 0;
+ return 0;
+skip:
+ wc->refs[level - 1] = 0;
+ wc->flags[level - 1] = 0;
+ if (wc->stage == DROP_REFERENCE) {
+ if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
+ parent = path->nodes[level]->start;
+ } else {
+ ASSERT(root->root_key.objectid ==
+ btrfs_header_owner(path->nodes[level]));
+ if (root->root_key.objectid !=
+ btrfs_header_owner(path->nodes[level])) {
+ btrfs_err(root->fs_info,
+ "mismatched block owner");
+ ret = -EIO;
+ goto out_unlock;
+ }
+ parent = 0;
+ }
+
+ /*
+ * If we had a drop_progress we need to verify the refs are set
+ * as expected. If we find our ref then we know that from here
+ * on out everything should be correct, and we can clear the
+ * ->restarted flag.
+ */
+ if (wc->restarted) {
+ ret = check_ref_exists(trans, root, bytenr, parent,
+ level - 1);
+ if (ret < 0)
+ goto out_unlock;
+ if (ret == 0)
+ goto no_delete;
+ ret = 0;
+ wc->restarted = 0;
+ }
+
+ /*
+ * Reloc tree doesn't contribute to qgroup numbers, and we have
+ * already accounted them at merge time (replace_path),
+ * thus we could skip expensive subtree trace here.
+ */
+ if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
+ need_account) {
+ ret = btrfs_qgroup_trace_subtree(trans, next,
+ generation, level - 1);
+ if (ret) {
+ btrfs_err_rl(fs_info,
+ "Error %d accounting shared subtree. Quota is out of sync, rescan required.",
+ ret);
}
- free_extent_map(em);
- goto out;
}
- path->slots[0]++;
- }
-out:
- return ret;
-}
-void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
-{
- struct btrfs_block_group_cache *block_group;
- u64 last = 0;
+ /*
+ * We need to update the next key in our walk control so we can
+ * update the drop_progress key accordingly. We don't care if
+ * find_next_key doesn't find a key because that means we're at
+ * the end and are going to clean up now.
+ */
+ wc->drop_level = level;
+ find_next_key(path, level, &wc->drop_progress);
- while (1) {
- struct inode *inode;
+ btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
+ fs_info->nodesize, parent);
+ btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid);
+ ret = btrfs_free_extent(trans, &ref);
+ if (ret)
+ goto out_unlock;
+ }
+no_delete:
+ *lookup_info = 1;
+ ret = 1;
- block_group = btrfs_lookup_first_block_group(info, last);
- while (block_group) {
- wait_block_group_cache_done(block_group);
- spin_lock(&block_group->lock);
- if (block_group->iref)
- break;
- spin_unlock(&block_group->lock);
- block_group = next_block_group(block_group);
- }
- if (!block_group) {
- if (last == 0)
- break;
- last = 0;
- continue;
- }
+out_unlock:
+ btrfs_tree_unlock(next);
+ free_extent_buffer(next);
- 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->key.objectid + block_group->key.offset;
- btrfs_put_block_group(block_group);
- }
+ return ret;
}
/*
- * Must be called only after stopping all workers, since we could have block
- * group caching kthreads running, and therefore they could race with us if we
- * freed the block groups before stopping them.
+ * helper to process tree block while walking up the tree.
+ *
+ * when wc->stage == DROP_REFERENCE, this function drops
+ * reference count on the block.
+ *
+ * when wc->stage == UPDATE_BACKREF, this function changes
+ * wc->stage back to DROP_REFERENCE if we changed wc->stage
+ * to UPDATE_BACKREF previously while processing the block.
+ *
+ * NOTE: return value 1 means we should stop walking up.
*/
-int btrfs_free_block_groups(struct btrfs_fs_info *info)
+static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct walk_control *wc)
{
- struct btrfs_block_group_cache *block_group;
- struct btrfs_space_info *space_info;
- struct btrfs_caching_control *caching_ctl;
- struct rb_node *n;
-
- down_write(&info->commit_root_sem);
- 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);
- put_caching_control(caching_ctl);
- }
- up_write(&info->commit_root_sem);
-
- spin_lock(&info->unused_bgs_lock);
- while (!list_empty(&info->unused_bgs)) {
- block_group = list_first_entry(&info->unused_bgs,
- struct btrfs_block_group_cache,
- 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) {
- block_group = rb_entry(n, struct btrfs_block_group_cache,
- cache_node);
- rb_erase(&block_group->cache_node,
- &info->block_group_cache_tree);
- RB_CLEAR_NODE(&block_group->cache_node);
- spin_unlock(&info->block_group_cache_lock);
-
- down_write(&block_group->space_info->groups_sem);
- list_del(&block_group->list);
- up_write(&block_group->space_info->groups_sem);
-
- /*
- * We haven't cached this block group, which means we could
- * possibly have excluded extents on this block group.
- */
- if (block_group->cached == BTRFS_CACHE_NO ||
- block_group->cached == BTRFS_CACHE_ERROR)
- free_excluded_extents(block_group);
-
- btrfs_remove_free_space_cache(block_group);
- ASSERT(block_group->cached != BTRFS_CACHE_STARTED);
- 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);
- btrfs_put_block_group(block_group);
-
- spin_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();
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ int ret;
+ int level = wc->level;
+ struct extent_buffer *eb = path->nodes[level];
+ u64 parent = 0;
- btrfs_release_global_block_rsv(info);
+ if (wc->stage == UPDATE_BACKREF) {
+ BUG_ON(wc->shared_level < level);
+ if (level < wc->shared_level)
+ goto out;
- while (!list_empty(&info->space_info)) {
- int i;
+ ret = find_next_key(path, level + 1, &wc->update_progress);
+ if (ret > 0)
+ wc->update_ref = 0;
- space_info = list_entry(info->space_info.next,
- struct btrfs_space_info,
- list);
+ wc->stage = DROP_REFERENCE;
+ wc->shared_level = -1;
+ path->slots[level] = 0;
/*
- * Do not hide this behind enospc_debug, this is actually
- * important and indicates a real bug if this happens.
+ * check reference count again if the block isn't locked.
+ * we should start walking down the tree again if reference
+ * count is one.
*/
- 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);
- list_del(&space_info->list);
- for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
- struct kobject *kobj;
- kobj = space_info->block_group_kobjs[i];
- space_info->block_group_kobjs[i] = NULL;
- if (kobj) {
- kobject_del(kobj);
- kobject_put(kobj);
+ if (!path->locks[level]) {
+ BUG_ON(level == 0);
+ btrfs_tree_lock(eb);
+ btrfs_set_lock_blocking_write(eb);
+ path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
+
+ ret = btrfs_lookup_extent_info(trans, fs_info,
+ eb->start, level, 1,
+ &wc->refs[level],
+ &wc->flags[level]);
+ if (ret < 0) {
+ btrfs_tree_unlock_rw(eb, path->locks[level]);
+ path->locks[level] = 0;
+ return ret;
+ }
+ BUG_ON(wc->refs[level] == 0);
+ if (wc->refs[level] == 1) {
+ btrfs_tree_unlock_rw(eb, path->locks[level]);
+ path->locks[level] = 0;
+ return 1;
}
}
- kobject_del(&space_info->kobj);
- kobject_put(&space_info->kobj);
}
- return 0;
-}
-static void link_block_group(struct btrfs_block_group_cache *cache)
-{
- struct btrfs_space_info *space_info = cache->space_info;
- struct btrfs_fs_info *fs_info = cache->fs_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) {
- struct raid_kobject *rkobj;
- unsigned int nofs_flag;
- int ret;
+ /* wc->stage == DROP_REFERENCE */
+ BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
- /*
- * Setup a NOFS context because kobject_add(), deep in its call
- * chain, does GFP_KERNEL allocations, and we are often called
- * in a context where if reclaim is triggered we can deadlock
- * (we are either holding a transaction handle or some lock
- * required for a transaction commit).
- */
- nofs_flag = memalloc_nofs_save();
- rkobj = kzalloc(sizeof(*rkobj), GFP_KERNEL);
- if (!rkobj) {
- memalloc_nofs_restore(nofs_flag);
- btrfs_warn(cache->fs_info,
- "couldn't alloc memory for raid level kobject");
- return;
+ if (wc->refs[level] == 1) {
+ if (level == 0) {
+ if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+ ret = btrfs_dec_ref(trans, root, eb, 1);
+ else
+ ret = btrfs_dec_ref(trans, root, eb, 0);
+ BUG_ON(ret); /* -ENOMEM */
+ if (is_fstree(root->root_key.objectid)) {
+ ret = btrfs_qgroup_trace_leaf_items(trans, eb);
+ if (ret) {
+ btrfs_err_rl(fs_info,
+ "error %d accounting leaf items, quota is out of sync, rescan required",
+ ret);
+ }
+ }
}
- rkobj->flags = cache->flags;
- kobject_init(&rkobj->kobj, &btrfs_raid_ktype);
- ret = kobject_add(&rkobj->kobj, &space_info->kobj, "%s",
- btrfs_bg_type_to_raid_name(rkobj->flags));
- memalloc_nofs_restore(nofs_flag);
- if (ret) {
- kobject_put(&rkobj->kobj);
- btrfs_warn(fs_info,
- "failed to add kobject for block cache, ignoring");
- return;
+ /* make block locked assertion in btrfs_clean_tree_block happy */
+ if (!path->locks[level] &&
+ btrfs_header_generation(eb) == trans->transid) {
+ btrfs_tree_lock(eb);
+ btrfs_set_lock_blocking_write(eb);
+ path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
}
- space_info->block_group_kobjs[index] = &rkobj->kobj;
+ btrfs_clean_tree_block(eb);
}
-}
-
-static struct btrfs_block_group_cache *
-btrfs_create_block_group_cache(struct btrfs_fs_info *fs_info,
- u64 start, u64 size)
-{
- struct btrfs_block_group_cache *cache;
-
- cache = kzalloc(sizeof(*cache), GFP_NOFS);
- if (!cache)
- return NULL;
- cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
- GFP_NOFS);
- if (!cache->free_space_ctl) {
- kfree(cache);
- return NULL;
+ if (eb == root->node) {
+ if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+ parent = eb->start;
+ else if (root->root_key.objectid != btrfs_header_owner(eb))
+ goto owner_mismatch;
+ } else {
+ if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+ parent = path->nodes[level + 1]->start;
+ else if (root->root_key.objectid !=
+ btrfs_header_owner(path->nodes[level + 1]))
+ goto owner_mismatch;
}
- cache->key.objectid = start;
- cache->key.offset = size;
- cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
-
- cache->fs_info = fs_info;
- cache->full_stripe_len = btrfs_full_stripe_len(fs_info, start);
- set_free_space_tree_thresholds(cache);
-
- atomic_set(&cache->count, 1);
- spin_lock_init(&cache->lock);
- init_rwsem(&cache->data_rwsem);
- INIT_LIST_HEAD(&cache->list);
- INIT_LIST_HEAD(&cache->cluster_list);
- INIT_LIST_HEAD(&cache->bg_list);
- INIT_LIST_HEAD(&cache->ro_list);
- INIT_LIST_HEAD(&cache->dirty_list);
- INIT_LIST_HEAD(&cache->io_list);
- btrfs_init_free_space_ctl(cache);
- atomic_set(&cache->trimming, 0);
- mutex_init(&cache->free_space_lock);
- btrfs_init_full_stripe_locks_tree(&cache->full_stripe_locks_root);
-
- return cache;
-}
-
-
-/*
- * Iterate all chunks and verify that each of them has the corresponding block
- * group
- */
-static int check_chunk_block_group_mappings(struct btrfs_fs_info *fs_info)
-{
- struct extent_map_tree *map_tree = &fs_info->mapping_tree;
- struct extent_map *em;
- struct btrfs_block_group_cache *bg;
- u64 start = 0;
- int ret = 0;
-
- while (1) {
- read_lock(&map_tree->lock);
- /*
- * lookup_extent_mapping will return the first extent map
- * intersecting the range, so setting @len to 1 is enough to
- * get the first chunk.
- */
- em = lookup_extent_mapping(map_tree, start, 1);
- read_unlock(&map_tree->lock);
- if (!em)
- break;
+ btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
+out:
+ wc->refs[level] = 0;
+ wc->flags[level] = 0;
+ return 0;
- bg = btrfs_lookup_block_group(fs_info, em->start);
- if (!bg) {
- btrfs_err(fs_info,
- "chunk start=%llu len=%llu doesn't have corresponding block group",
- em->start, em->len);
- ret = -EUCLEAN;
- free_extent_map(em);
- break;
- }
- if (bg->key.objectid != em->start ||
- bg->key.offset != em->len ||
- (bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK) !=
- (em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK)) {
- btrfs_err(fs_info,
-"chunk start=%llu len=%llu flags=0x%llx doesn't match block group start=%llu len=%llu flags=0x%llx",
- em->start, em->len,
- em->map_lookup->type & BTRFS_BLOCK_GROUP_TYPE_MASK,
- bg->key.objectid, bg->key.offset,
- bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK);
- ret = -EUCLEAN;
- free_extent_map(em);
- btrfs_put_block_group(bg);
- break;
- }
- start = em->start + em->len;
- free_extent_map(em);
- btrfs_put_block_group(bg);
- }
- return ret;
+owner_mismatch:
+ btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu",
+ btrfs_header_owner(eb), root->root_key.objectid);
+ return -EUCLEAN;
}
-int btrfs_read_block_groups(struct btrfs_fs_info *info)
+static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct walk_control *wc)
{
- struct btrfs_path *path;
+ int level = wc->level;
+ int lookup_info = 1;
int ret;
- struct btrfs_block_group_cache *cache;
- struct btrfs_space_info *space_info;
- struct btrfs_key key;
- struct btrfs_key found_key;
- struct extent_buffer *leaf;
- int need_clear = 0;
- u64 cache_gen;
- u64 feature;
- int mixed;
-
- feature = btrfs_super_incompat_flags(info->super_copy);
- mixed = !!(feature & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS);
-
- 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) &&
- btrfs_super_generation(info->super_copy) != cache_gen)
- need_clear = 1;
- if (btrfs_test_opt(info, CLEAR_CACHE))
- need_clear = 1;
-
- while (1) {
- ret = find_first_block_group(info, path, &key);
+ while (level >= 0) {
+ ret = walk_down_proc(trans, root, path, wc, lookup_info);
if (ret > 0)
break;
- if (ret != 0)
- goto error;
-
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-
- cache = btrfs_create_block_group_cache(info, found_key.objectid,
- found_key.offset);
- if (!cache) {
- ret = -ENOMEM;
- goto error;
- }
-
- if (need_clear) {
- /*
- * When we mount with old space cache, we need to
- * set BTRFS_DC_CLEAR and set dirty flag.
- *
- * a) Setting 'BTRFS_DC_CLEAR' makes sure that we
- * truncate the old free space cache inode and
- * setup a new one.
- * b) Setting 'dirty flag' makes sure that we flush
- * the new space cache info onto disk.
- */
- if (btrfs_test_opt(info, SPACE_CACHE))
- cache->disk_cache_state = BTRFS_DC_CLEAR;
- }
-
- read_extent_buffer(leaf, &cache->item,
- btrfs_item_ptr_offset(leaf, path->slots[0]),
- sizeof(cache->item));
- cache->flags = btrfs_block_group_flags(&cache->item);
- if (!mixed &&
- ((cache->flags & BTRFS_BLOCK_GROUP_METADATA) &&
- (cache->flags & BTRFS_BLOCK_GROUP_DATA))) {
- btrfs_err(info,
-"bg %llu is a mixed block group but filesystem hasn't enabled mixed block groups",
- cache->key.objectid);
- ret = -EINVAL;
- goto error;
- }
-
- key.objectid = found_key.objectid + found_key.offset;
- btrfs_release_path(path);
-
- /*
- * 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 than we actually do.
- */
- ret = exclude_super_stripes(cache);
- if (ret) {
- /*
- * We may have excluded something, so call this just in
- * case.
- */
- free_excluded_extents(cache);
- btrfs_put_block_group(cache);
- goto error;
- }
-
- /*
- * 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 (found_key.offset == btrfs_block_group_used(&cache->item)) {
- cache->last_byte_to_unpin = (u64)-1;
- cache->cached = BTRFS_CACHE_FINISHED;
- free_excluded_extents(cache);
- } else if (btrfs_block_group_used(&cache->item) == 0) {
- cache->last_byte_to_unpin = (u64)-1;
- cache->cached = BTRFS_CACHE_FINISHED;
- add_new_free_space(cache, found_key.objectid,
- found_key.objectid +
- found_key.offset);
- free_excluded_extents(cache);
- }
-
- ret = btrfs_add_block_group_cache(info, cache);
- if (ret) {
- btrfs_remove_free_space_cache(cache);
- btrfs_put_block_group(cache);
- goto error;
- }
- trace_btrfs_add_block_group(info, cache, 0);
- btrfs_update_space_info(info, cache->flags, found_key.offset,
- btrfs_block_group_used(&cache->item),
- cache->bytes_super, &space_info);
-
- cache->space_info = space_info;
+ if (level == 0)
+ break;
- link_block_group(cache);
+ if (path->slots[level] >=
+ btrfs_header_nritems(path->nodes[level]))
+ break;
- set_avail_alloc_bits(info, cache->flags);
- if (btrfs_chunk_readonly(info, cache->key.objectid)) {
- inc_block_group_ro(cache, 1);
- } else if (btrfs_block_group_used(&cache->item) == 0) {
- ASSERT(list_empty(&cache->bg_list));
- btrfs_mark_bg_unused(cache);
- }
+ ret = do_walk_down(trans, root, path, wc, &lookup_info);
+ if (ret > 0) {
+ path->slots[level]++;
+ continue;
+ } else if (ret < 0)
+ return ret;
+ level = wc->level;
}
-
- list_for_each_entry_rcu(space_info, &info->space_info, list) {
- if (!(get_alloc_profile(info, space_info->flags) &
- (BTRFS_BLOCK_GROUP_RAID10 |
- BTRFS_BLOCK_GROUP_RAID1_MASK |
- BTRFS_BLOCK_GROUP_RAID56_MASK |
- BTRFS_BLOCK_GROUP_DUP)))
- continue;
- /*
- * avoid allocating from un-mirrored block group if there are
- * mirrored block groups.
- */
- list_for_each_entry(cache,
- &space_info->block_groups[BTRFS_RAID_RAID0],
- list)
- inc_block_group_ro(cache, 1);
- list_for_each_entry(cache,
- &space_info->block_groups[BTRFS_RAID_SINGLE],
- list)
- inc_block_group_ro(cache, 1);
- }
-
- btrfs_init_global_block_rsv(info);
- ret = check_chunk_block_group_mappings(info);
-error:
- btrfs_free_path(path);
- return ret;
+ return 0;
}
-void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans)
+static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct walk_control *wc, int max_level)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_block_group_cache *block_group;
- struct btrfs_root *extent_root = fs_info->extent_root;
- struct btrfs_block_group_item item;
- struct btrfs_key key;
- int ret = 0;
-
- if (!trans->can_flush_pending_bgs)
- return;
-
- while (!list_empty(&trans->new_bgs)) {
- block_group = list_first_entry(&trans->new_bgs,
- struct btrfs_block_group_cache,
- bg_list);
- if (ret)
- goto next;
+ int level = wc->level;
+ int ret;
- spin_lock(&block_group->lock);
- memcpy(&item, &block_group->item, sizeof(item));
- memcpy(&key, &block_group->key, sizeof(key));
- spin_unlock(&block_group->lock);
+ path->slots[level] = btrfs_header_nritems(path->nodes[level]);
+ while (level < max_level && path->nodes[level]) {
+ wc->level = level;
+ if (path->slots[level] + 1 <
+ btrfs_header_nritems(path->nodes[level])) {
+ path->slots[level]++;
+ return 0;
+ } else {
+ ret = walk_up_proc(trans, root, path, wc);
+ if (ret > 0)
+ return 0;
+ if (ret < 0)
+ return ret;
- ret = btrfs_insert_item(trans, extent_root, &key, &item,
- sizeof(item));
- if (ret)
- btrfs_abort_transaction(trans, ret);
- ret = btrfs_finish_chunk_alloc(trans, key.objectid, key.offset);
- if (ret)
- btrfs_abort_transaction(trans, ret);
- add_block_group_free_space(trans, block_group);
- /* already aborted the transaction if it failed. */
-next:
- btrfs_delayed_refs_rsv_release(fs_info, 1);
- list_del_init(&block_group->bg_list);
+ if (path->locks[level]) {
+ btrfs_tree_unlock_rw(path->nodes[level],
+ path->locks[level]);
+ path->locks[level] = 0;
+ }
+ free_extent_buffer(path->nodes[level]);
+ path->nodes[level] = NULL;
+ level++;
+ }
}
- btrfs_trans_release_chunk_metadata(trans);
+ return 1;
}
-int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used,
- u64 type, u64 chunk_offset, u64 size)
+/*
+ * drop a subvolume tree.
+ *
+ * this function traverses the tree freeing any blocks that only
+ * referenced by the tree.
+ *
+ * when a shared tree block is found. this function decreases its
+ * reference count by one. if update_ref is true, this function
+ * also make sure backrefs for the shared block and all lower level
+ * blocks are properly updated.
+ *
+ * If called with for_reloc == 0, may exit early with -EAGAIN
+ */
+int btrfs_drop_snapshot(struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv, int update_ref,
+ int for_reloc)
{
- struct btrfs_fs_info *fs_info = trans->fs_info;
- struct btrfs_block_group_cache *cache;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ struct btrfs_path *path;
+ struct btrfs_trans_handle *trans;
+ struct btrfs_root *tree_root = fs_info->tree_root;
+ struct btrfs_root_item *root_item = &root->root_item;
+ struct walk_control *wc;
+ struct btrfs_key key;
+ int err = 0;
int ret;
+ int level;
+ bool root_dropped = false;
- btrfs_set_log_full_commit(trans);
-
- cache = btrfs_create_block_group_cache(fs_info, chunk_offset, size);
- if (!cache)
- return -ENOMEM;
+ btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid);
- btrfs_set_block_group_used(&cache->item, bytes_used);
- btrfs_set_block_group_chunk_objectid(&cache->item,
- BTRFS_FIRST_CHUNK_TREE_OBJECTID);
- btrfs_set_block_group_flags(&cache->item, type);
+ path = btrfs_alloc_path();
+ if (!path) {
+ err = -ENOMEM;
+ goto out;
+ }
- cache->flags = type;
- cache->last_byte_to_unpin = (u64)-1;
- cache->cached = BTRFS_CACHE_FINISHED;
- cache->needs_free_space = 1;
- ret = exclude_super_stripes(cache);
- if (ret) {
- /*
- * We may have excluded something, so call this just in
- * case.
- */
- free_excluded_extents(cache);
- btrfs_put_block_group(cache);
- return ret;
+ wc = kzalloc(sizeof(*wc), GFP_NOFS);
+ if (!wc) {
+ btrfs_free_path(path);
+ err = -ENOMEM;
+ goto out;
}
- add_new_free_space(cache, chunk_offset, chunk_offset + size);
+ trans = btrfs_start_transaction(tree_root, 0);
+ if (IS_ERR(trans)) {
+ err = PTR_ERR(trans);
+ goto out_free;
+ }
- free_excluded_extents(cache);
+ err = btrfs_run_delayed_items(trans);
+ if (err)
+ goto out_end_trans;
-#ifdef CONFIG_BTRFS_DEBUG
- if (btrfs_should_fragment_free_space(cache)) {
- u64 new_bytes_used = size - bytes_used;
+ if (block_rsv)
+ trans->block_rsv = block_rsv;
- 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
- * with its ->space_info set.
+ * This will help us catch people modifying the fs tree while we're
+ * dropping it. It is unsafe to mess with the fs tree while it's being
+ * dropped as we unlock the root node and parent nodes as we walk down
+ * the tree, assuming nothing will change. If something does change
+ * then we'll have stale information and drop references to blocks we've
+ * already dropped.
*/
- cache->space_info = btrfs_find_space_info(fs_info, cache->flags);
- ASSERT(cache->space_info);
-
- ret = btrfs_add_block_group_cache(fs_info, cache);
- if (ret) {
- btrfs_remove_free_space_cache(cache);
- btrfs_put_block_group(cache);
- return ret;
- }
+ set_bit(BTRFS_ROOT_DELETING, &root->state);
+ if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
+ level = btrfs_header_level(root->node);
+ path->nodes[level] = btrfs_lock_root_node(root);
+ btrfs_set_lock_blocking_write(path->nodes[level]);
+ path->slots[level] = 0;
+ path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
+ memset(&wc->update_progress, 0,
+ sizeof(wc->update_progress));
+ } else {
+ btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
+ memcpy(&wc->update_progress, &key,
+ sizeof(wc->update_progress));
- /*
- * Now that our block group has its ->space_info set and is inserted in
- * 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_update_global_block_rsv(fs_info);
+ level = root_item->drop_level;
+ BUG_ON(level == 0);
+ path->lowest_level = level;
+ ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+ path->lowest_level = 0;
+ if (ret < 0) {
+ err = ret;
+ goto out_end_trans;
+ }
+ WARN_ON(ret > 0);
- link_block_group(cache);
+ /*
+ * unlock our path, this is safe because only this
+ * function is allowed to delete this snapshot
+ */
+ btrfs_unlock_up_safe(path, 0);
- list_add_tail(&cache->bg_list, &trans->new_bgs);
- trans->delayed_ref_updates++;
- btrfs_update_delayed_refs_rsv(trans);
+ level = btrfs_header_level(root->node);
+ while (1) {
+ btrfs_tree_lock(path->nodes[level]);
+ btrfs_set_lock_blocking_write(path->nodes[level]);
+ path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
- set_avail_alloc_bits(fs_info, type);
- return 0;
-}
+ ret = btrfs_lookup_extent_info(trans, fs_info,
+ path->nodes[level]->start,
+ level, 1, &wc->refs[level],
+ &wc->flags[level]);
+ if (ret < 0) {
+ err = ret;
+ goto out_end_trans;
+ }
+ BUG_ON(wc->refs[level] == 0);
-static void clear_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
-{
- u64 extra_flags = chunk_to_extended(flags) &
- BTRFS_EXTENDED_PROFILE_MASK;
-
- write_seqlock(&fs_info->profiles_lock);
- if (flags & BTRFS_BLOCK_GROUP_DATA)
- fs_info->avail_data_alloc_bits &= ~extra_flags;
- if (flags & BTRFS_BLOCK_GROUP_METADATA)
- fs_info->avail_metadata_alloc_bits &= ~extra_flags;
- if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
- fs_info->avail_system_alloc_bits &= ~extra_flags;
- write_sequnlock(&fs_info->profiles_lock);
-}
+ if (level == root_item->drop_level)
+ break;
-/*
- * Clear incompat bits for the following feature(s):
- *
- * - RAID56 - in case there's neither RAID5 nor RAID6 profile block group
- * in the whole filesystem
- */
-static void clear_incompat_bg_bits(struct btrfs_fs_info *fs_info, u64 flags)
-{
- if (flags & BTRFS_BLOCK_GROUP_RAID56_MASK) {
- struct list_head *head = &fs_info->space_info;
- struct btrfs_space_info *sinfo;
-
- list_for_each_entry_rcu(sinfo, head, list) {
- bool found = false;
-
- down_read(&sinfo->groups_sem);
- if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID5]))
- found = true;
- if (!list_empty(&sinfo->block_groups[BTRFS_RAID_RAID6]))
- found = true;
- up_read(&sinfo->groups_sem);
-
- if (found)
- return;
+ btrfs_tree_unlock(path->nodes[level]);
+ path->locks[level] = 0;
+ WARN_ON(wc->refs[level] != 1);
+ level--;
}
- btrfs_clear_fs_incompat(fs_info, RAID56);
}
-}
-
-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_cache *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;
- int index;
- int factor;
- struct btrfs_caching_control *caching_ctl = NULL;
- bool remove_em;
- bool remove_rsv = false;
- block_group = btrfs_lookup_block_group(fs_info, group_start);
- BUG_ON(!block_group);
- BUG_ON(!block_group->ro);
-
- trace_btrfs_remove_block_group(block_group);
- /*
- * Free the reserved super bytes from this block group before
- * remove it.
- */
- free_excluded_extents(block_group);
- btrfs_free_ref_tree_range(fs_info, block_group->key.objectid,
- block_group->key.offset);
+ wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
+ wc->level = level;
+ wc->shared_level = -1;
+ wc->stage = DROP_REFERENCE;
+ wc->update_ref = update_ref;
+ wc->keep_locks = 0;
+ wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
- memcpy(&key, &block_group->key, sizeof(key));
- index = btrfs_bg_flags_to_raid_index(block_group->flags);
- factor = btrfs_bg_type_to_factor(block_group->flags);
+ while (1) {
- /* make sure this block group isn't part of an allocation cluster */
- cluster = &fs_info->data_alloc_cluster;
- spin_lock(&cluster->refill_lock);
- btrfs_return_cluster_to_free_space(block_group, cluster);
- spin_unlock(&cluster->refill_lock);
+ ret = walk_down_tree(trans, root, path, wc);
+ if (ret < 0) {
+ err = ret;
+ break;
+ }
- /*
- * make sure this block group isn't part of a metadata
- * allocation cluster
- */
- cluster = &fs_info->meta_alloc_cluster;
- spin_lock(&cluster->refill_lock);
- btrfs_return_cluster_to_free_space(block_group, cluster);
- spin_unlock(&cluster->refill_lock);
+ ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
+ if (ret < 0) {
+ err = ret;
+ break;
+ }
- path = btrfs_alloc_path();
- if (!path) {
- ret = -ENOMEM;
- goto out;
- }
+ if (ret > 0) {
+ BUG_ON(wc->stage != DROP_REFERENCE);
+ break;
+ }
- /*
- * get the inode first so any iput calls done for the io_list
- * aren't the final iput (no unlinks allowed now)
- */
- inode = lookup_free_space_inode(block_group, path);
+ if (wc->stage == DROP_REFERENCE) {
+ wc->drop_level = wc->level;
+ btrfs_node_key_to_cpu(path->nodes[wc->drop_level],
+ &wc->drop_progress,
+ path->slots[wc->drop_level]);
+ }
+ btrfs_cpu_key_to_disk(&root_item->drop_progress,
+ &wc->drop_progress);
+ root_item->drop_level = wc->drop_level;
- mutex_lock(&trans->transaction->cache_write_mutex);
- /*
- * Make sure our free space cache IO is done before removing the
- * free space inode
- */
- spin_lock(&trans->transaction->dirty_bgs_lock);
- if (!list_empty(&block_group->io_list)) {
- list_del_init(&block_group->io_list);
+ BUG_ON(wc->level == 0);
+ if (btrfs_should_end_transaction(trans) ||
+ (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) {
+ ret = btrfs_update_root(trans, tree_root,
+ &root->root_key,
+ root_item);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ err = ret;
+ goto out_end_trans;
+ }
- WARN_ON(!IS_ERR(inode) && inode != block_group->io_ctl.inode);
+ btrfs_end_transaction_throttle(trans);
+ if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) {
+ btrfs_debug(fs_info,
+ "drop snapshot early exit");
+ err = -EAGAIN;
+ goto out_free;
+ }
- spin_unlock(&trans->transaction->dirty_bgs_lock);
- btrfs_wait_cache_io(trans, block_group, path);
- btrfs_put_block_group(block_group);
- spin_lock(&trans->transaction->dirty_bgs_lock);
+ trans = btrfs_start_transaction(tree_root, 0);
+ if (IS_ERR(trans)) {
+ err = PTR_ERR(trans);
+ goto out_free;
+ }
+ if (block_rsv)
+ trans->block_rsv = block_rsv;
+ }
}
+ btrfs_release_path(path);
+ if (err)
+ goto out_end_trans;
- if (!list_empty(&block_group->dirty_list)) {
- list_del_init(&block_group->dirty_list);
- remove_rsv = true;
- btrfs_put_block_group(block_group);
+ ret = btrfs_del_root(trans, &root->root_key);
+ if (ret) {
+ btrfs_abort_transaction(trans, ret);
+ err = ret;
+ goto out_end_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);
+ if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
+ ret = btrfs_find_root(tree_root, &root->root_key, path,
+ NULL, NULL);
+ if (ret < 0) {
+ btrfs_abort_transaction(trans, ret);
+ err = ret;
+ goto out_end_trans;
+ } else if (ret > 0) {
+ /* if we fail to delete the orphan item this time
+ * around, it'll get picked up the next time.
+ *
+ * The most common failure here is just -ENOENT.
+ */
+ btrfs_del_orphan_item(trans, tree_root,
+ root->root_key.objectid);
}
- /* One for our lookup ref */
- btrfs_add_delayed_iput(inode);
}
- key.objectid = BTRFS_FREE_SPACE_OBJECTID;
- key.offset = block_group->key.objectid;
- key.type = 0;
-
- ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
- if (ret < 0)
- 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);
+ if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state)) {
+ btrfs_add_dropped_root(trans, root);
+ } else {
+ free_extent_buffer(root->node);
+ free_extent_buffer(root->commit_root);
+ btrfs_put_fs_root(root);
}
-
- spin_lock(&fs_info->block_group_cache_lock);
- rb_erase(&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->key.objectid)
- fs_info->first_logical_byte = (u64)-1;
- spin_unlock(&fs_info->block_group_cache_lock);
-
- down_write(&block_group->space_info->groups_sem);
+ root_dropped = true;
+out_end_trans:
+ btrfs_end_transaction_throttle(trans);
+out_free:
+ kfree(wc);
+ btrfs_free_path(path);
+out:
/*
- * we must use list_del_init so people can check to see if they
- * are still on the list after taking the semaphore
+ * So if we need to stop dropping the snapshot for whatever reason we
+ * need to make sure to add it back to the dead root list so that we
+ * keep trying to do the work later. This also cleans up roots if we
+ * don't have it in the radix (like when we recover after a power fail
+ * or unmount) so we don't leak memory.
*/
- list_del_init(&block_group->list);
- if (list_empty(&block_group->space_info->block_groups[index])) {
- kobj = block_group->space_info->block_group_kobjs[index];
- block_group->space_info->block_group_kobjs[index] = NULL;
- clear_avail_alloc_bits(fs_info, block_group->flags);
- }
- up_write(&block_group->space_info->groups_sem);
- clear_incompat_bg_bits(fs_info, block_group->flags);
- if (kobj) {
- kobject_del(kobj);
- kobject_put(kobj);
- }
-
- if (block_group->has_caching_ctl)
- caching_ctl = get_caching_control(block_group);
- if (block_group->cached == BTRFS_CACHE_STARTED)
- 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. */
- put_caching_control(caching_ctl);
- put_caching_control(caching_ctl);
- }
- }
+ if (!for_reloc && !root_dropped)
+ btrfs_add_dead_root(root);
+ if (err && err != -EAGAIN)
+ btrfs_handle_fs_error(fs_info, err, NULL);
+ return err;
+}
- spin_lock(&trans->transaction->dirty_bgs_lock);
- WARN_ON(!list_empty(&block_group->dirty_list));
- WARN_ON(!list_empty(&block_group->io_list));
- spin_unlock(&trans->transaction->dirty_bgs_lock);
+/*
+ * drop subtree rooted at tree block 'node'.
+ *
+ * NOTE: this function will unlock and release tree block 'node'
+ * only used by relocation code
+ */
+int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct extent_buffer *node,
+ struct extent_buffer *parent)
+{
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ struct btrfs_path *path;
+ struct walk_control *wc;
+ int level;
+ int parent_level;
+ int ret = 0;
+ int wret;
- btrfs_remove_free_space_cache(block_group);
+ BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
- spin_lock(&block_group->space_info->lock);
- list_del_init(&block_group->ro_list);
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
- if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
- WARN_ON(block_group->space_info->total_bytes
- < block_group->key.offset);
- WARN_ON(block_group->space_info->bytes_readonly
- < block_group->key.offset);
- WARN_ON(block_group->space_info->disk_total
- < block_group->key.offset * factor);
+ wc = kzalloc(sizeof(*wc), GFP_NOFS);
+ if (!wc) {
+ btrfs_free_path(path);
+ return -ENOMEM;
}
- block_group->space_info->total_bytes -= block_group->key.offset;
- block_group->space_info->bytes_readonly -= block_group->key.offset;
- block_group->space_info->disk_total -= block_group->key.offset * factor;
-
- spin_unlock(&block_group->space_info->lock);
-
- memcpy(&key, &block_group->key, sizeof(key));
- mutex_lock(&fs_info->chunk_mutex);
- spin_lock(&block_group->lock);
- block_group->removed = 1;
- /*
- * 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().
- *
- * 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
- * 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
- * physical device locations unless we take this special care.
- *
- * There may also be an implicit trim operation if the file system
- * is mounted with -odiscard. The same protections must remain
- * in place until the extents have been discarded completely when
- * the transaction commit has completed.
- */
- remove_em = (atomic_read(&block_group->trimming) == 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);
+ btrfs_assert_tree_locked(parent);
+ parent_level = btrfs_header_level(parent);
+ extent_buffer_get(parent);
+ path->nodes[parent_level] = parent;
+ path->slots[parent_level] = btrfs_header_nritems(parent);
- ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
- if (ret > 0)
- ret = -EIO;
- if (ret < 0)
- goto out;
+ btrfs_assert_tree_locked(node);
+ level = btrfs_header_level(node);
+ path->nodes[level] = node;
+ path->slots[level] = 0;
+ path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
- ret = btrfs_del_item(trans, root, path);
- if (ret)
- goto out;
+ wc->refs[parent_level] = 1;
+ wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
+ wc->level = level;
+ wc->shared_level = -1;
+ wc->stage = DROP_REFERENCE;
+ wc->update_ref = 0;
+ wc->keep_locks = 1;
+ wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
- if (remove_em) {
- struct extent_map_tree *em_tree;
+ while (1) {
+ wret = walk_down_tree(trans, root, path, wc);
+ if (wret < 0) {
+ ret = wret;
+ break;
+ }
- em_tree = &fs_info->mapping_tree;
- write_lock(&em_tree->lock);
- remove_extent_mapping(em_tree, em);
- write_unlock(&em_tree->lock);
- /* once for the tree */
- free_extent_map(em);
+ wret = walk_up_tree(trans, root, path, wc, parent_level);
+ if (wret < 0)
+ ret = wret;
+ if (wret != 0)
+ break;
}
-out:
- if (remove_rsv)
- btrfs_delayed_refs_rsv_release(fs_info, 1);
+
+ kfree(wc);
btrfs_free_path(path);
return ret;
}
-struct btrfs_trans_handle *
-btrfs_start_trans_remove_block_group(struct btrfs_fs_info *fs_info,
- const u64 chunk_offset)
-{
- struct extent_map_tree *em_tree = &fs_info->mapping_tree;
- struct extent_map *em;
- struct map_lookup *map;
- unsigned int num_items;
-
- read_lock(&em_tree->lock);
- em = lookup_extent_mapping(em_tree, chunk_offset, 1);
- read_unlock(&em_tree->lock);
- ASSERT(em && em->start == chunk_offset);
-
- /*
- * We need to reserve 3 + N units from the metadata space info in order
- * to remove a block group (done at btrfs_remove_chunk() and at
- * btrfs_remove_block_group()), which are used for:
- *
- * 1 unit for adding the free space inode's orphan (located in the tree
- * of tree roots).
- * 1 unit for deleting the block group item (located in the extent
- * tree).
- * 1 unit for deleting the free space item (located in tree of tree
- * roots).
- * N units for deleting N device extent items corresponding to each
- * stripe (located in the device tree).
- *
- * In order to remove a block group we also need to reserve units in the
- * system space info in order to update the chunk tree (update one or
- * more device items and remove one chunk item), but this is done at
- * btrfs_remove_chunk() through a call to check_system_chunk().
- */
- map = em->map_lookup;
- num_items = 3 + map->num_stripes;
- free_extent_map(em);
-
- return btrfs_start_transaction_fallback_global_rsv(fs_info->extent_root,
- num_items, 1);
-}
-
/*
- * Process the unused_bgs list and remove any that don't have any allocated
- * space inside of them.
+ * helper to account the unused space of all the readonly block group in the
+ * space_info. takes mirrors into account.
*/
-void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info)
+u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
{
struct btrfs_block_group_cache *block_group;
- struct btrfs_space_info *space_info;
- struct btrfs_trans_handle *trans;
- int ret = 0;
-
- if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
- 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,
- struct btrfs_block_group_cache,
- bg_list);
- list_del_init(&block_group->bg_list);
-
- space_info = block_group->space_info;
-
- if (ret || btrfs_mixed_space_info(space_info)) {
- btrfs_put_block_group(block_group);
- continue;
- }
- spin_unlock(&fs_info->unused_bgs_lock);
+ u64 free_bytes = 0;
+ int factor;
- mutex_lock(&fs_info->delete_unused_bgs_mutex);
+ /* It's df, we don't care if it's racy */
+ if (list_empty(&sinfo->ro_bgs))
+ return 0;
- /* Don't want to race with allocators so take the groups_sem */
- down_write(&space_info->groups_sem);
+ spin_lock(&sinfo->lock);
+ list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
spin_lock(&block_group->lock);
- if (block_group->reserved || block_group->pinned ||
- btrfs_block_group_used(&block_group->item) ||
- block_group->ro ||
- list_is_singular(&block_group->list)) {
- /*
- * 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.
- */
- trace_btrfs_skip_unused_block_group(block_group);
- spin_unlock(&block_group->lock);
- up_write(&space_info->groups_sem);
- goto next;
- }
- spin_unlock(&block_group->lock);
-
- /* We don't want to force the issue, only flip if it's ok. */
- ret = inc_block_group_ro(block_group, 0);
- up_write(&space_info->groups_sem);
- if (ret < 0) {
- 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.
- */
- trans = btrfs_start_trans_remove_block_group(fs_info,
- block_group->key.objectid);
- if (IS_ERR(trans)) {
- btrfs_dec_block_group_ro(block_group);
- ret = PTR_ERR(trans);
- goto next;
- }
- /*
- * We could have pending pinned extents for this block group,
- * just delete them, we don't care about them anymore.
- */
- start = block_group->key.objectid;
- end = start + block_group->key.offset - 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);
- 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;
+ if (!block_group->ro) {
+ spin_unlock(&block_group->lock);
+ continue;
}
- mutex_unlock(&fs_info->unused_bg_unpin_mutex);
-
- /* Reset pinned so btrfs_put_block_group doesn't complain */
- spin_lock(&space_info->lock);
- spin_lock(&block_group->lock);
- 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;
+ factor = btrfs_bg_type_to_factor(block_group->flags);
+ free_bytes += (block_group->key.offset -
+ btrfs_block_group_used(&block_group->item)) *
+ factor;
spin_unlock(&block_group->lock);
- spin_unlock(&space_info->lock);
-
- /* DISCARD can flip during remount */
- trimming = btrfs_test_opt(fs_info, DISCARD);
-
- /* Implicit trim during transaction commit. */
- if (trimming)
- btrfs_get_block_group_trimming(block_group);
-
- /*
- * Btrfs_remove_chunk will abort the transaction if things go
- * horribly wrong.
- */
- ret = btrfs_remove_chunk(trans, block_group->key.objectid);
-
- if (ret) {
- if (trimming)
- btrfs_put_block_group_trimming(block_group);
- goto end_trans;
- }
-
- /*
- * If we're not mounted with -odiscard, we can just forget
- * about this block group. Otherwise we'll need to wait
- * until transaction commit to do the actual discard.
- */
- if (trimming) {
- spin_lock(&fs_info->unused_bgs_lock);
- /*
- * A concurrent scrub might have added us to the list
- * fs_info->unused_bgs, so use a list_move operation
- * to add the block group to the deleted_bgs list.
- */
- list_move(&block_group->bg_list,
- &trans->transaction->deleted_bgs);
- spin_unlock(&fs_info->unused_bgs_lock);
- btrfs_get_block_group(block_group);
- }
-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);
+ spin_unlock(&sinfo->lock);
+
+ return free_bytes;
}
int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
return -EINVAL;
cache = btrfs_lookup_first_block_group(fs_info, range->start);
- for (; cache; cache = next_block_group(cache)) {
+ for (; cache; cache = btrfs_next_block_group(cache)) {
if (cache->key.objectid >= range_end) {
btrfs_put_block_group(cache);
break;
end = min(range_end, cache->key.objectid + cache->key.offset);
if (end - start >= range->minlen) {
- if (!block_group_cache_done(cache)) {
- ret = cache_block_group(cache, 0);
+ if (!btrfs_block_group_cache_done(cache)) {
+ ret = btrfs_cache_block_group(cache, 0);
if (ret) {
bg_failed++;
bg_ret = ret;
continue;
}
- ret = wait_block_group_cache_done(cache);
+ ret = btrfs_wait_block_group_cache_done(cache);
if (ret) {
bg_failed++;
bg_ret = ret;
!atomic_read(&root->will_be_snapshotted));
}
}
-
-void btrfs_mark_bg_unused(struct btrfs_block_group_cache *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_unused_block_group(bg);
- list_add_tail(&bg->bg_list, &fs_info->unused_bgs);
- }
- spin_unlock(&fs_info->unused_bgs_lock);
-}