Re: [PATCH 04/10] btrfs: split btrfs_submit_data_bio

[Christoph Hellwig <hch@lst.de>]

On Wed, May 04, 2022 at 08:38:23PM +0800, Qu Wenruo wrote:
>> +	struct inode *inode = tree->private_data;
>
> I guess we shouldn't use the extent_io_tree for bio->bi_private at all
> if we're just tring to grab an inode.
>
> In fact, for all submit_one_bio() callers, we are handling buffered
> read/write, thus we can grab inode using
> bio_first_page_all(bio)->mapping->host.
>
> No need for such weird io_tree based dance.

Yes, we can eventully.  Not for this series, though.

>> -	if (is_data_inode(tree->private_data))
>> -		btrfs_submit_data_bio(tree->private_data, bio, mirror_num,
>> -					    compress_type);
>> +	if (!is_data_inode(tree->private_data))
>> +		btrfs_submit_metadata_bio(inode, bio, mirror_num);
>
> Can we just call btrfs_submit_metadata_bio() and return?
>
> Every time I see an "if () else if ()", I can't stop thinking if we have
> some corner cases not taken into consideration.

I generally agree with you, but for this case I think it is pretty
simple.  But a few more series down the road these helpers will change
a bit anyway, so we can revisit it.