Re: [PATCH 3/7] btrfs: Use filemap_invalidate_pages()

On Thu, Aug 20, 2026 at 08:33:36PM +0100, Matthew Wilcox (Oracle) wrote:
> btrfs relies on invalidate_inode_pages2() /
> invalidate_inode_pages2_range() doing writeback by calling
> btrfs_launder_folio().  While this works, it is inefficient as each

I don't believe btrfs_launder_folio() is doing writeback, it is just
dropping otherwise leaked qgroup reservation. So this part of the
commit message feels inaccurate to me, at least.

> folio is written back and waited for individually.  Far better to call
> filemap_invalidate_pages() which will do a bulk write first, then remove
> the page cache.
> 
> With this done, btrfs_launder_folio() no longer needs to exist so
> delete it.
> 
> Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
> ---
>  fs/btrfs/direct-io.c        |  2 +-
>  fs/btrfs/disk-io.c          | 11 +++++++----
>  fs/btrfs/free-space-cache.c |  4 ++--
>  fs/btrfs/inode.c            | 12 ++----------
>  fs/btrfs/volumes.c          |  4 ++--
>  5 files changed, 14 insertions(+), 19 deletions(-)
> 
> diff --git a/fs/btrfs/direct-io.c b/fs/btrfs/direct-io.c
> index 460326d34143..4bb5890d0898 100644
> --- a/fs/btrfs/direct-io.c
> +++ b/fs/btrfs/direct-io.c
> @@ -115,7 +115,7 @@ static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend,
>  			/*
>  			 * We could trigger writeback for this range (and wait
>  			 * for it to complete) and then invalidate the pages for
> -			 * this range (through invalidate_inode_pages2_range()),
> +			 * this range (through filemap_invalidate_pages()),
>  			 * but that can lead us to a deadlock with a concurrent
>  			 * call to readahead (a buffered read or a defrag call
>  			 * triggered a readahead) on a page lock due to an
> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
> index 2f1666d9544e..2cf9189225c9 100644
> --- a/fs/btrfs/disk-io.c
> +++ b/fs/btrfs/disk-io.c
> @@ -3304,7 +3304,8 @@ static void invalidate_and_check_btree_folios(struct btrfs_fs_info *fs_info)
>  	struct extent_buffer *eb;
>  	int ret;
>  
> -	ret = invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
> +	ret = filemap_invalidate_pages(fs_info->btree_inode->i_mapping, 0,
> +			OFFSET_MAX);
>  	if (likely(ret == 0))
>  		return;
>  
> @@ -3339,7 +3340,8 @@ static void invalidate_and_check_btree_folios(struct btrfs_fs_info *fs_info)
>  		rcu_read_lock();
>  	}
>  	rcu_read_unlock();
> -	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
> +	filemap_invalidate_pages(fs_info->btree_inode->i_mapping, 0,
> +			OFFSET_MAX);
>  }
>  
>  static u32 calc_block_max_order(u32 sectorsize_bits)
> @@ -4739,7 +4741,8 @@ static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
>  			unsigned int nofs_flag;
>  
>  			nofs_flag = memalloc_nofs_save();
> -			invalidate_inode_pages2(inode->i_mapping);
> +			filemap_invalidate_pages(inode->i_mapping, 0,
> +					OFFSET_MAX);

I believe that this was the original motivating call to
invalidate_inode_pages2 that I cared about. The context it runs in is
when the filesystem has failed on a transaction and needs to invalidate
the remaining dirty pages that have delalloc associated with them.

I am trying to figure out if trying to force the writeback in this
semi-failed context will also result in freeing the reservation, and
also testing this patch series on the test that I originally fixed with
launder_folio().

With all that said, thank you for working on cleaning up the mess I made
by adding this "creative" usage of ->launder_folio(). Hopefully we can
figure this out cleanly.

Thanks,
Boris

>  			memalloc_nofs_restore(nofs_flag);
>  			iput(inode);
>  		}
> @@ -4837,7 +4840,7 @@ static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache)
>  		unsigned int nofs_flag;
>  
>  		nofs_flag = memalloc_nofs_save();
> -		invalidate_inode_pages2(inode->i_mapping);
> +		filemap_invalidate_pages(inode->i_mapping, 0, OFFSET_MAX);
>  		memalloc_nofs_restore(nofs_flag);
>  
>  		BTRFS_I(inode)->generation = 0;
> diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
> index e2af75a205ea..a5d5eba2c85d 100644
> --- a/fs/btrfs/free-space-cache.c
> +++ b/fs/btrfs/free-space-cache.c
> @@ -1307,7 +1307,7 @@ static int __btrfs_wait_cache_io(struct btrfs_root *root,
>  				io_ctl->entries, io_ctl->bitmaps);
>  out:
>  	if (ret) {
> -		invalidate_inode_pages2(inode->i_mapping);
> +		filemap_invalidate_pages(inode->i_mapping, 0, OFFSET_MAX);
>  		BTRFS_I(inode)->generation = 0;
>  		if (block_group)
>  			btrfs_debug(root->fs_info,
> @@ -1500,7 +1500,7 @@ static int __btrfs_write_out_cache(struct inode *inode,
>  	io_ctl->inode = NULL;
>  	io_ctl_free(io_ctl);
>  	if (ret) {
> -		invalidate_inode_pages2(inode->i_mapping);
> +		filemap_invalidate_pages(inode->i_mapping, 0, OFFSET_MAX);
>  		BTRFS_I(inode)->generation = 0;
>  	}
>  	btrfs_update_inode(trans, BTRFS_I(inode));
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 2534cd9284d5..a1bb73f083aa 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -7616,12 +7616,6 @@ static void wait_subpage_spinlock(struct folio *folio)
>  	spin_unlock_irq(&bfs->lock);
>  }
>  
> -static int btrfs_launder_folio(struct folio *folio)
> -{
> -	return btrfs_qgroup_free_data(folio_to_inode(folio), NULL, folio_pos(folio),
> -				      folio_size(folio), NULL);
> -}
> -
>  static bool __btrfs_release_folio(struct folio *folio, gfp_t gfp_flags)
>  {
>  	if (try_release_extent_mapping(folio, gfp_flags)) {
> @@ -10027,9 +10021,8 @@ ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
>  		ret = btrfs_wait_ordered_range(inode, start, num_bytes);
>  		if (ret)
>  			goto out_cb;
> -		ret = invalidate_inode_pages2_range(inode->vfs_inode.i_mapping,
> -						    start >> PAGE_SHIFT,
> -						    end >> PAGE_SHIFT);
> +		ret = filemap_invalidate_pages(inode->vfs_inode.i_mapping,
> +				start, end);
>  		if (ret)
>  			goto out_cb;
>  		btrfs_lock_extent(io_tree, start, end, &cached_state);
> @@ -10782,7 +10775,6 @@ static const struct address_space_operations btrfs_aops = {
>  	.writepages	= btrfs_writepages,
>  	.readahead	= btrfs_readahead,
>  	.invalidate_folio = btrfs_invalidate_folio,
> -	.launder_folio	= btrfs_launder_folio,
>  	.release_folio	= btrfs_release_folio,
>  	.migrate_folio	= btrfs_migrate_folio,
>  	.dirty_folio	= btrfs_data_dirty_folio,
> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> index 6eab4cc73ce4..0301465ab87f 100644
> --- a/fs/btrfs/volumes.c
> +++ b/fs/btrfs/volumes.c
> @@ -1370,8 +1370,8 @@ struct btrfs_super_block *btrfs_read_disk_super(struct block_device *bdev,
>  		 * Drop the page of the primary superblock, so later read will
>  		 * always read from the device.
>  		 */
> -		invalidate_inode_pages2_range(mapping, bytenr >> PAGE_SHIFT,
> -				      (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT);
> +		filemap_invalidate_pages(mapping, bytenr,
> +				bytenr + BTRFS_SUPER_INFO_SIZE - 1);
>  	}
>  
>  	filemap_invalidate_lock_shared(mapping);
> -- 
> 2.47.3
>