Re: [PATCH 2/5] btrfs: zoned: drop stale dirty btree buffers at close_ctree()

[Qu Wenruo <wqu@suse.com>]

在 2026/7/16 21:26, Johannes Thumshirn 写道:
> On a zoned filesystem btrfs_clear_buffer_dirty() keeps a freed-but-dirty
> tree block dirty (EXTENT_BUFFER_ZONED_ZEROOUT) so its zero-out keeps the
> zone write pointer moving. A block left ahead of the block group's
> meta_write_pointer (e.g. a tree-log block freed before being written) can
> never be written by btree_writepages() and survives to unmount.
> 
> close_ctree() stops the endio workqueues before the final iput() of the
> btree inode but frees the block groups only afterwards. Once the block
> group is gone btrfs_check_meta_write_pointer() no longer defers the buffer,
> so the iput() submits it for writeback; its completion can no longer be
> queued on the destroyed endio_meta_workers and umount hangs on writeback
> that never finishes.

Can we just trigger a btree inode write back for zoned cases?

That sounds much simpler and less hacky.

Thanks,
Qu
> 
> btrfs_free_block_groups() has to stay after btrfs_stop_all_workers()
> (see also commit 5cdd7db6c5c9), so instead drop the dirty state of such
> stale buffers in invalidate_and_check_btree_folios(), which still runs
> while the workqueues are alive.
> 
> Assisted-by: LLM (debugging, commit message)
> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> ---
>   fs/btrfs/disk-io.c   |  4 ++++
>   fs/btrfs/extent_io.c | 12 ++++++++++++
>   fs/btrfs/extent_io.h |  1 +
>   3 files changed, 17 insertions(+)
> 
> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
> index acff40469589..b66525701b33 100644
> --- a/fs/btrfs/disk-io.c
> +++ b/fs/btrfs/disk-io.c
> @@ -3317,6 +3317,10 @@ static void invalidate_and_check_btree_folios(struct btrfs_fs_info *fs_info)
>   		if (test_bit(EXTENT_BUFFER_READING, &eb->bflags))
>   			wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_READING,
>   				       TASK_UNINTERRUPTIBLE);
> +
> +		if (btrfs_is_zoned(fs_info) &&
> +		    test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags))
> +			btrfs_clear_stale_buffer_dirty(eb);
>   		/*
>   		 * The refs threshold is 2, one held by us at the beginning
>   		 * of the loop, one for the ownership in the buffer tree.
> diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
> index 4dd8fc5ad247..480d4ec8ca7f 100644
> --- a/fs/btrfs/extent_io.c
> +++ b/fs/btrfs/extent_io.c
> @@ -3841,6 +3841,18 @@ void btrfs_clear_buffer_dirty(struct btrfs_trans_handle *trans,
>   	__btrfs_clear_buffer_dirty(eb);
>   }
>   
> +/*
> + * Drop the dirty state of a stale, already-freed metadata buffer, bypassing
> + * the zoned EXTENT_BUFFER_ZONED_ZEROOUT deferral of btrfs_clear_buffer_dirty().
> + */
> +void btrfs_clear_stale_buffer_dirty(struct extent_buffer *eb)
> +{
> +	btrfs_tree_lock(eb);
> +	clear_bit(EXTENT_BUFFER_ZONED_ZEROOUT, &eb->bflags);
> +	__btrfs_clear_buffer_dirty(eb);
> +	btrfs_tree_unlock(eb);
> +}
> +
>   void set_extent_buffer_dirty(struct extent_buffer *eb)
>   {
>   	bool was_dirty;
> diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
> index 869925337699..290d6cff6ddb 100644
> --- a/fs/btrfs/extent_io.h
> +++ b/fs/btrfs/extent_io.h
> @@ -393,6 +393,7 @@ void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
>   				  u32 bits_to_clear, unsigned long page_ops);
>   void btrfs_clear_buffer_dirty(struct btrfs_trans_handle *trans,
>   			      struct extent_buffer *buf);
> +void btrfs_clear_stale_buffer_dirty(struct extent_buffer *eb);
>   
>   static inline void btrfs_clear_folio_dirty_tag(struct folio *folio)
>   {