Re: [PATCH v2] rust: dma: return zero for Coherent reads past EOF

On Sun, 02 Aug 2026 16:54:40 +0300=0D
Onur =C3=96zkan <[email protected]> wrote:=0D
=0D
> On Thu, 30 Jul 2026 18:34:38 +0200=0D
> Younes Akhouayri via B4 Relay <[email protected]> wrote:=0D
> =0D
> > From: Younes Akhouayri <[email protected]>=0D
> > =0D
> > Coherent<T>::write_to_slice() calculates a zero-byte copy when the file=
=0D
> > offset is beyond the allocation, but still calls=0D
> > UserSliceWriter::write_dma(). The latter rejects offsets beyond the=0D
> > allocation even when the copy length is zero, so a debugfs read past EO=
F=0D
> > returns -ERANGE.=0D
> > =0D
> > Return before calling write_dma() when the offset is at or beyond the=0D
> > allocation, matching simple_read_from_buffer() EOF semantics.=0D
> > =0D
> > Fixes: 016818513936 ("rust: dma: implement BinaryWriter for Coherent<[u=
8]>")=0D
> > Cc: [email protected]=0D
> > Link: https://rust-for-linux.zulipchat.com/#narrow/channel/291566-Libra=
ry/topic/.E2.9C.94.20Possible.20past-EOF.20bug.20in.20Coherent.3CT.3E.3A.3A=
write_to_slice/near/611677095=0D
> > Signed-off-by: Younes Akhouayri <[email protected]>=0D
> > ---=0D
> > Changes in v2:=0D
> > - Use ordinary subtraction after the explicit EOF check, which makes=0D
> >   underflow impossible.=0D
> > - Link to v1: https://patch.msgid.link/20260720-fix-dma-coherent-eof-v1=
[email protected]=0D
> > ---=0D
> >  rust/kernel/dma.rs | 6 +++++-=0D
> >  1 file changed, 5 insertions(+), 1 deletion(-)=0D
> > =0D
> > diff --git a/rust/kernel/dma.rs b/rust/kernel/dma.rs=0D
> > index 200def84fb69..556424b88fb9 100644=0D
> > --- a/rust/kernel/dma.rs=0D
> > +++ b/rust/kernel/dma.rs=0D
> > @@ -1005,7 +1005,11 @@ fn write_to_slice(=0D
> >              return Ok(0);=0D
> >          };=0D
> >  =0D
> > -        let count =3D self.size().saturating_sub(offset_val).min(write=
r.len());=0D
> =0D
> Nit: I would probably just check if `count` is zero next to this line.=0D
=0D
I just noticed there is already been quite a bit of discussion about this t=
opic=0D
so you can ignore my comment since it was just a "nit" anyway. I don't thin=
k=0D
there's much difference either way, so there's probably no need to reignite=
 the=0D
discussion.=0D
=0D
Thanks for the fix.=0D
=0D
Reviewed-by: Onur =C3=96zkan <[email protected]>=0D
=0D
> =0D
> > +        if offset_val >=3D self.size() {=0D
> > +            return Ok(0);=0D
> > +        }=0D
> > +=0D
> > +        let count =3D (self.size() - offset_val).min(writer.len());=0D
> >  =0D
> >          writer.write_dma(self, offset_val, count)?;=0D
> >  =0D
> > =0D
> > ---=0D
> > base-commit: 667d0fb32149f023b8b34a1f6f3d384556eafb5a=0D
> > change-id: 20260720-fix-dma-coherent-eof-d0d08c91e013=0D
> > =0D
> > Best regards,=0D
> > --  =0D
> > Younes Akhouayri <[email protected]>=0D
> > =0D
> > =0D