Re: [PATCHv2 1/7] block: check for valid bio while splitting

[Keith Busch <kbusch@kernel.org>]

On Sun, Aug 10, 2025 at 07:37:36AM -0700, Christoph Hellwig wrote:
> On Tue, Aug 05, 2025 at 07:11:17AM -0700, Keith Busch wrote:
> > @@ -341,6 +344,8 @@ int bio_split_rw_at(struct bio *bio, const struct queue_limits *lim,
> >  	 * we do not use the full hardware limits.
> >  	 */
> >  	bytes = ALIGN_DOWN(bytes, bio_split_alignment(bio, lim));
> > +	if (!bytes)
> > +		return -EINVAL;
> 
> How is this related to the other hunk and the patch description?

The patchset allows you to submit an io with vectors that are partial
logical blocks. Misuse could create a bio that exceeds the device max
vectors or introduces virtual boundary gaps, requiring a split into
something that is smaller than a block size. This check catches that.

Quick example: nvme with a 4k logical block size, and the usual 4k
virtual boundary. Send an io with four vectors iov_len=1k. The total
size is block sized, but there's no way that could split into a valid
io. There's a test specifically for this in my reply about xfstests.