Merge tag 'for-5.6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull more btrfs updates from David Sterba:
 "Fixes that arrived after the merge window freeze, mostly stable
  material.

   - fix race in tree-mod-log element tracking

   - fix bio flushing inside extent writepages

   - fix assertion when in-memory tracking of discarded extents finds an
     empty tree (eg. after adding a new device)

   - update logic of temporary read-only block groups to take into
     account overcommit

   - fix some fixup worker corner cases:
       - page could not go through proper COW cycle and the dirty status
         is lost due to page migration
       - deadlock if delayed allocation is performed under page lock

   - fix send emitting invalid clones within the same file

   - fix statfs reporting 0 free space when global block reserve size is
     larger than remaining free space but there is still space for new
     chunks"

* tag 'for-5.6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: do not zero f_bavail if we have available space
  Btrfs: send, fix emission of invalid clone operations within the same file
  btrfs: do not do delalloc reservation under page lock
  btrfs: drop the -EBUSY case in __extent_writepage_io
  Btrfs: keep pages dirty when using btrfs_writepage_fixup_worker
  btrfs: take overcommit into account in inc_block_group_ro
  btrfs: fix force usage in inc_block_group_ro
  btrfs: Correctly handle empty trees in find_first_clear_extent_bit
  btrfs: flush write bio if we loop in extent_write_cache_pages
  Btrfs: fix race between adding and putting tree mod seq elements and nodes

Author: torvalds

fs/btrfs/block-group.c

@@ -1191,7 +1191,6 @@ static int inc_block_group_ro(struct btrfs_block_group *cache, int force)
 {
 	struct btrfs_space_info *sinfo = cache->space_info;
 	u64 num_bytes;
-	u64 sinfo_used;
 	int ret = -ENOSPC;
 
 	spin_lock(&sinfo->lock);
@@ -1205,29 +1204,45 @@ static int inc_block_group_ro(struct btrfs_block_group *cache, int force)
 
 	num_bytes = cache->length - cache->reserved - cache->pinned -
 		    cache->bytes_super - cache->used;
-	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.
+	 * Data never overcommits, even in mixed mode, so do just the straight
+	 * check of left over space in how much we have allocated.
 	 */
-	if (sinfo_used + num_bytes <= sinfo->total_bytes) {
+	if (force) {
+		ret = 0;
+	} else if (sinfo->flags & BTRFS_BLOCK_GROUP_DATA) {
+		u64 sinfo_used = btrfs_space_info_used(sinfo, true);
+
+		/*
+		 * Here we make sure if we mark this bg RO, we still have enough
+		 * free space as buffer.
+		 */
+		if (sinfo_used + num_bytes <= sinfo->total_bytes)
+			ret = 0;
+	} else {
+		/*
+		 * We overcommit metadata, so we need to do the
+		 * btrfs_can_overcommit check here, and we need to pass in
+		 * BTRFS_RESERVE_NO_FLUSH to give ourselves the most amount of
+		 * leeway to allow us to mark this block group as read only.
+		 */
+		if (btrfs_can_overcommit(cache->fs_info, sinfo, num_bytes,
+					 BTRFS_RESERVE_NO_FLUSH))
+			ret = 0;
+	}
+
+	if (!ret) {
 		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->start);
-		btrfs_info(cache->fs_info,
-			"sinfo_used=%llu bg_num_bytes=%llu",
-			sinfo_used, num_bytes);
 		btrfs_dump_space_info(cache->fs_info, cache->space_info, 0, 0);
 	}
 	return ret;
@@ -2225,7 +2240,7 @@ int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
 		}
 	}
 
-	ret = inc_block_group_ro(cache, !do_chunk_alloc);
+	ret = inc_block_group_ro(cache, 0);
 	if (!do_chunk_alloc)
 		goto unlock_out;
 	if (!ret)

fs/btrfs/ctree.c

@@ -326,12 +326,10 @@ u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
 			   struct seq_list *elem)
 {
 	write_lock(&fs_info->tree_mod_log_lock);
-	spin_lock(&fs_info->tree_mod_seq_lock);
 	if (!elem->seq) {
 		elem->seq = btrfs_inc_tree_mod_seq(fs_info);
 		list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
 	}
-	spin_unlock(&fs_info->tree_mod_seq_lock);
 	write_unlock(&fs_info->tree_mod_log_lock);
 
 	return elem->seq;
@@ -351,7 +349,7 @@ void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
 	if (!seq_putting)
 		return;
 
-	spin_lock(&fs_info->tree_mod_seq_lock);
+	write_lock(&fs_info->tree_mod_log_lock);
 	list_del(&elem->list);
 	elem->seq = 0;
 
@@ -362,19 +360,17 @@ void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
 				 * blocker with lower sequence number exists, we
 				 * cannot remove anything from the log
 				 */
-				spin_unlock(&fs_info->tree_mod_seq_lock);
+				write_unlock(&fs_info->tree_mod_log_lock);
 				return;
 			}
 			min_seq = cur_elem->seq;
 		}
 	}
-	spin_unlock(&fs_info->tree_mod_seq_lock);
 
 	/*
 	 * anything that's lower than the lowest existing (read: blocked)
 	 * sequence number can be removed from the tree.
 	 */
-	write_lock(&fs_info->tree_mod_log_lock);
 	tm_root = &fs_info->tree_mod_log;
 	for (node = rb_first(tm_root); node; node = next) {
 		next = rb_next(node);

fs/btrfs/ctree.h

@@ -714,14 +714,12 @@ struct btrfs_fs_info {
 	atomic_t nr_delayed_iputs;
 	wait_queue_head_t delayed_iputs_wait;
 
-	/* this protects tree_mod_seq_list */
-	spinlock_t tree_mod_seq_lock;
 	atomic64_t tree_mod_seq;
-	struct list_head tree_mod_seq_list;
 
-	/* this protects tree_mod_log */
+	/* this protects tree_mod_log and tree_mod_seq_list */
 	rwlock_t tree_mod_log_lock;
 	struct rb_root tree_mod_log;
+	struct list_head tree_mod_seq_list;
 
 	atomic_t async_delalloc_pages;
 

fs/btrfs/delayed-ref.c

@@ -492,15 +492,15 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
 	if (head->is_data)
 		return;
 
-	spin_lock(&fs_info->tree_mod_seq_lock);
+	read_lock(&fs_info->tree_mod_log_lock);
 	if (!list_empty(&fs_info->tree_mod_seq_list)) {
 		struct seq_list *elem;
 
 		elem = list_first_entry(&fs_info->tree_mod_seq_list,
 					struct seq_list, list);
 		seq = elem->seq;
 	}
-	spin_unlock(&fs_info->tree_mod_seq_lock);
+	read_unlock(&fs_info->tree_mod_log_lock);
 
 again:
 	for (node = rb_first_cached(&head->ref_tree); node;
@@ -518,7 +518,7 @@ int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq)
 	struct seq_list *elem;
 	int ret = 0;
 
-	spin_lock(&fs_info->tree_mod_seq_lock);
+	read_lock(&fs_info->tree_mod_log_lock);
 	if (!list_empty(&fs_info->tree_mod_seq_list)) {
 		elem = list_first_entry(&fs_info->tree_mod_seq_list,
 					struct seq_list, list);
@@ -531,7 +531,7 @@ int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq)
 		}
 	}
 
-	spin_unlock(&fs_info->tree_mod_seq_lock);
+	read_unlock(&fs_info->tree_mod_log_lock);
 	return ret;
 }
 

fs/btrfs/disk-io.c

@@ -2697,7 +2697,6 @@ int __cold open_ctree(struct super_block *sb,
 	spin_lock_init(&fs_info->fs_roots_radix_lock);
 	spin_lock_init(&fs_info->delayed_iput_lock);
 	spin_lock_init(&fs_info->defrag_inodes_lock);
-	spin_lock_init(&fs_info->tree_mod_seq_lock);
 	spin_lock_init(&fs_info->super_lock);
 	spin_lock_init(&fs_info->buffer_lock);
 	spin_lock_init(&fs_info->unused_bgs_lock);

fs/btrfs/extent_io.c

@@ -1593,21 +1593,25 @@ void find_first_clear_extent_bit(struct extent_io_tree *tree, u64 start,
 	/* Find first extent with bits cleared */
 	while (1) {
 		node = __etree_search(tree, start, &next, &prev, NULL, NULL);
-		if (!node) {
+		if (!node && !next && !prev) {
+			/*
+			 * Tree is completely empty, send full range and let
+			 * caller deal with it
+			 */
+			*start_ret = 0;
+			*end_ret = -1;
+			goto out;
+		} else if (!node && !next) {
+			/*
+			 * We are past the last allocated chunk, set start at
+			 * the end of the last extent.
+			 */
+			state = rb_entry(prev, struct extent_state, rb_node);
+			*start_ret = state->end + 1;
+			*end_ret = -1;
+			goto out;
+		} else if (!node) {
 			node = next;
-			if (!node) {
-				/*
-				 * We are past the last allocated chunk,
-				 * set start at the end of the last extent. The
-				 * device alloc tree should never be empty so
-				 * prev is always set.
-				 */
-				ASSERT(prev);
-				state = rb_entry(prev, struct extent_state, rb_node);
-				*start_ret = state->end + 1;
-				*end_ret = -1;
-				goto out;
-			}
 		}
 		/*
 		 * At this point 'node' either contains 'start' or start is
@@ -3438,11 +3442,7 @@ static noinline_for_stack int __extent_writepage_io(struct inode *inode,
 	ret = btrfs_writepage_cow_fixup(page, start, page_end);
 	if (ret) {
 		/* Fixup worker will requeue */
-		if (ret == -EBUSY)
-			wbc->pages_skipped++;
-		else
-			redirty_page_for_writepage(wbc, page);
-
+		redirty_page_for_writepage(wbc, page);
 		update_nr_written(wbc, nr_written);
 		unlock_page(page);
 		return 1;
@@ -4166,7 +4166,16 @@ static int extent_write_cache_pages(struct address_space *mapping,
 		 */
 		scanned = 1;
 		index = 0;
-		goto retry;
+
+		/*
+		 * If we're looping we could run into a page that is locked by a
+		 * writer and that writer could be waiting on writeback for a
+		 * page in our current bio, and thus deadlock, so flush the
+		 * write bio here.
+		 */
+		ret = flush_write_bio(epd);
+		if (!ret)
+			goto retry;
 	}
 
 	if (wbc->range_cyclic || (wbc->nr_to_write > 0 && range_whole))