Re: [PATCH v3 11/28] iomap: pass readpage operation to read path

From: Andrey Albershteyn <aalbersh@redhat.com>
To: Christoph Hellwig <hch@infradead.org>
Cc: linux-xfs@vger.kernel.org, linux-fsdevel@vger.kernel.org,
	 fsverity@lists.linux.dev, djwong@kernel.org,
	ebiggers@kernel.org, david@fromorbit.com,  dchinner@redhat.com
Subject: Re: [PATCH v3 11/28] iomap: pass readpage operation to read path
Date: Mon, 16 Oct 2023 14:32:25 +0200	[thread overview]
Message-ID: <xic2dogypvli45ku7nasuuszslvv55tadj6etpl7r3gubvw2o6@hms55yjiqsiq> (raw)
In-Reply-To: <ZSz/FLK+tNIQzOA/@infradead.org>

On 2023-10-16 02:15:00, Christoph Hellwig wrote:
> Looking at the entire series, it seems like the only XFS-specific
> part of the fsverity processing in iomap is the per-sb workqueue
> now that the fsverity interfaces were cleaned up.
> 
> Based on that it seems like we'd be much better just doing all the
> work inside iomap, and just allow XFS to pass the workqueue to
> iomap_read_folio and iomap_readahead.  The patch below does that
> as an untested WIP on top of your branch.
> 
> If we go down that way I suspect the better interface would be
> to allocate the iomap_readpage_ctx in the callers of these functions
> instead of passing an extra argument, but I'm not entirely sure
> about that yet.

From the discussion in v2 [1] I understood that btrfs would like to
have this bio_set/submit_io interface for iomap transition; and any
other filesystem deferrals would also be possible. Is it no more the
case with btrfs? Would fs-verity verification in iomap_read_end_io
combine both solutions (fs-verity verification in iomap +
submit_io/bio_set interface).

[1]: https://lore.kernel.org/linux-xfs/ZCxEHkWayQyGqnxL@infradead.org/#t

-- 
- Andrey