Re: [PATCH v7] iomap: make inline data support more flexible

From: Gao Xiang <hsiangkao@linux.alibaba.com>
To: Matthew Wilcox <willy@infradead.org>,
	"Darrick J . Wong" <djwong@kernel.org>
Cc: Andreas Gruenbacher <agruenba@redhat.com>,
	linux-kernel@vger.kernel.org,
	Andreas Gruenbacher <andreas.gruenbacher@gmail.com>,
	dsterba@suse.cz, linux-fsdevel@vger.kernel.org,
	linux-erofs@lists.ozlabs.org, Christoph Hellwig <hch@lst.de>
Subject: Re: [PATCH v7] iomap: make inline data support more flexible
Date: Tue, 27 Jul 2021 23:04:36 +0800	[thread overview]
Message-ID: <YQAghPSTWdTGYAm5@B-P7TQMD6M-0146.local> (raw)
In-Reply-To: <YQALsvt0UWGW+iMw@casper.infradead.org>

On Tue, Jul 27, 2021 at 02:35:46PM +0100, Matthew Wilcox wrote:
> On Tue, Jul 27, 2021 at 10:20:42AM +0200, David Sterba wrote:
> > On Mon, Jul 26, 2021 at 02:17:02PM +0200, Christoph Hellwig wrote:
> > > > Subject: iomap: Support tail packing
> > > 
> > > I can't say I like this "tail packing" language here when we have the
> > > perfectly fine inline wording.  Same for various comments in the actual
> > > code.
> > 
> > Yes please, don't call it tail-packing when it's an inline extent, we'll
> > use that for btrfs eventually and conflating the two terms has been
> > cofusing users. Except reiserfs, no linux filesystem does tail-packing.
> 
> Hmm ... I see what reiserfs does as packing tails of multiple files into
> one block.  What gfs2 (and ext4) do is inline data.  Erofs packs the
> tail of a single file into the same block as the inode.  If I understand

Plus each erofs block can have multiple inodes (thus multi-tail blocks) 
oo as long as the meta block itself can fit.

No matter what it's called, it's a kind of inline data (I think inline
means that data mixes with metadata according to [1]). I was called it
tail-block inline initially... whatever.

Hopefully, Darrick could update the v8 title if some concern here.

[1] https://www.kernel.org/doc/Documentation/filesystems/fiemap.txt

Thanks,
Gao Xiang

> what btrfs does correctly, it stores data in the btree.  But (like
> gfs2/ext4), it's only for the entire-file-is-small case, not for
> its-just-ten-bytes-into-the-last-block case.
> 
> So what would you call what erofs is doing if not tail-packing?
> Wikipedia calls it https://en.wikipedia.org/wiki/Block_suballocation
> which doesn't quite fit.  We need a phrase which means "this isn't
> just for small files but for small tails of large files".