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

From: Gao Xiang <hsiangkao@linux.alibaba.com>
To: Matthew Wilcox <willy@infradead.org>
Cc: linux-erofs@lists.ozlabs.org, linux-fsdevel@vger.kernel.org,
	LKML <linux-kernel@vger.kernel.org>,
	Christoph Hellwig <hch@lst.de>,
	"Darrick J . Wong" <djwong@kernel.org>,
	Andreas Gruenbacher <andreas.gruenbacher@gmail.com>,
	Huang Jianan <huangjianan@oppo.com>
Subject: Re: [PATCH v7] iomap: make inline data support more flexible
Date: Sat, 24 Jul 2021 08:54:10 +0800	[thread overview]
Message-ID: <YPtksjmvVbcsKwlK@B-P7TQMD6M-0146.local> (raw)
In-Reply-To: <YPsbQzcNz+r4V7P2@casper.infradead.org>

Hi Matthew,

On Fri, Jul 23, 2021 at 08:40:51PM +0100, Matthew Wilcox wrote:
> On Sat, Jul 24, 2021 at 01:41:31AM +0800, Gao Xiang wrote:
> > Add support for reading inline data content into the page cache from
> > nonzero page-aligned file offsets.  This enables the EROFS tailpacking
> > mode where the last few bytes of the file are stored right after the
> > inode.
> > 
> > The buffered write path remains untouched since EROFS cannot be used
> > for testing. It'd be better to be implemented if upcoming real users
> > care and provide a real pattern rather than leave untested dead code
> > around.
> 
> My one complaint with this version is the subject line.  It's a bit vague.
> I went with:
> 
> iomap: Support file tail packing
> 
> I also wrote a changelog entry that reads:
>     The existing inline data support only works for cases where the entire
>     file is stored as inline data.  For larger files, EROFS stores the
>     initial blocks separately and then can pack a small tail adjacent to
>     the inode.  Generalise inline data to allow for tail packing.  Tails
>     may not cross a page boundary in memory.
>

Yeah, we could complete the commit message like this.

Actually EROFS inode base is only 32-byte or 64-byte (so the maximum could
not be exactly small), compared to using another tail block or storing other
(maybe) irrelevant inodes. According to cache locality principle, a strategy
can be selected by mkfs to load tail block with the inode base itself to the
page cache by the tail-packing inline and so reduce I/O and fragmentation.

> ... but I'm not sure that's necessarily better than what you've written
> here.
> 
> > Cc: Christoph Hellwig <hch@lst.de>
> > Cc: Darrick J. Wong <djwong@kernel.org>
> > Cc: Matthew Wilcox <willy@infradead.org>
> > Cc: Andreas Gruenbacher <andreas.gruenbacher@gmail.com>
> > Tested-by: Huang Jianan <huangjianan@oppo.com> # erofs
> > Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
> 
> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>

Many thanks for the review!

Thanks,
Gao Xiang