Re: linux-next: manual merge of the block tree with the mm tree

[Andrew Morton]

On Tue, 20 Oct 2026 11:13:23 -0600 Jens Axboe <axboe@xxxxxxxxx> wrote:

> On 10/20/26 11:00 AM, Andrew Morton wrote:
> > On Tue, 20 Oct 2026 06:45:16 -0600 Jens Axboe <axboe@xxxxxxxxx> wrote:
> > 
> >> On 10/20/26 6:27 AM, Mark Brown wrote:
> >>> Hi all,
> >>>
> >>> Today's linux-next merge of the block tree got a conflict in:
> >>>
> >>>   block/bio.c
> >>>
> >>> between commit:
> >>>
> >>>   d4ee6948bcabb ("block: restore the iterator when rejecting a misaligned bio")
> >>>
> >>> from the mm tree and commit:
> >>>
> >>>   73a0ff6030afc ("block: fail a short atomic pin in bio_iov_iter_get_pages()")
> >>>
> >>> from the block tree.
> >>
> >> I suspected as much when I saw Andrew pick it up. Andrew, perhaps
> >> let's queue it where it belongs?
> > 
> > I was hoping to get some review feedback before sending it over
> > formally.
> > 
> > I'll send it now.
> 
> Applied - can you double check? Flow is a bit different in there
> compared to the one you patched against, entirely hand applied.
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux.git/commit/?h=for-7.4/block&id=40d070cbb1906896d6fd65c39722d2411830ed9f

Integration lgtm, thanks.


The error path leaves bio->bi_iter.bi_size still reflecting the bytes
extracted before the failure which is untidy, but no caller appears to
use it.


I'll also send over Daehyeon's "iov_iter: validate extracted segment
for bvec alignment", v2.

He showed that we can allow misaligned direct I/O to be accepted when
it should have failed with -EINVAL.  Daehyeon reproduced the problem
and showed misaligned bvecs reaching a bio-based driver.  We shouldn't
rely on drivers to enforce an alignment requirement which should have
been checked earlier.