Re: [PATCH v2 4/4] iomap: don't lose a failed direct I/O bio's error when zeroing the tail

[Andrea Parri]

On Thu, Sep 24, 2026 at 11:53:17AM -0700, Darrick J. Wong wrote:
> > Cc: stable@xxxxxxxxxxxxxxx
> 
> Cc: <stable@xxxxxxxxxxxxxxx> # v6.12

Fixed for v3.

> > @@ -582,8 +578,7 @@ static int iomap_dio_bio_iter(struct iomap_iter *iter, struct iomap_dio *dio)
> >  		/* zero out from the end of the write to the end of the block */
> >  		pad = pos & (fs_block_size - 1);
> >  		if (pad)
> > -			ret = iomap_dio_zero(iter, dio, pos,
> > -					     fs_block_size - pad);
> > +			iomap_dio_zero(iter, dio, pos, fs_block_size - pad);
> 
> I think this is the original cause of the bug, right?  We might have
> already had a nonzero ret, and the assignment here blows that away.
> Right?

Right.  When iomap_dio_bio_iter_one() fails we break out of the loop
with ret < 0 and fall through to the tail zeroing, and this assignment
replaced that error with iomap_dio_zero()'s 0.

> If the answer to that is yes, then
> Reviewed-by: "Darrick J. Wong" <djwong@xxxxxxxxxx>

Thanks, picked up for v3.

> (I should probably whine about how changing the function signature of
> iomap_dio_zero should be a separate patch to make it more obvious what's
> the actual fix...)

Fair point.  FWIW, before 10553a91652d iomap_dio_zero() returned void
and neither call site touched ret; that commit made it return int and
added both "ret = iomap_dio_zero(...)" assignments.  So the signature
change here is really undoing the part of 10553a91652d that introduced
the bug, and keeping it in one patch makes the stable backport
self-contained.  If you'd still prefer a one-line fix followed by the
void conversion, I'm happy to split it.

Thanks!
  Andrea