Re: [PATCH net v3] ethernet: ionic: Fix DMA mapping tests

[Brett Creeley]

On 6/20/2025 3:51 AM, Simon Horman wrote:
> Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.
>
>
> On Thu, Jun 19, 2025 at 03:28:06PM -0700, Brett Creeley wrote:
> > On 6/19/2025 2:45 AM, Thomas Fourier wrote:
> > > Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.
> > >
> > >
> > > Change error values of `ionic_tx_map_single()` and `ionic_tx_map_frag()`
> > > from 0 to `DMA_MAPPING_ERROR` to prevent collision with 0 as a valid
> > > address.
> > >
> > > This also fixes the use of `dma_mapping_error()` to test against 0 in
> > > `ionic_xdp_post_frame()`
> > >
> > > Fixes: 0f3154e6bcb3 ("ionic: Add Tx and Rx handling")
> >
> > I'm not sure the Fixes commit above should be in the list. Functionally it's
> > correct, except there being multiple calls to dma_mapping_error() on the
> > same dma_addr.
> >
> > Other than the minor nit above the commit looks good. Thanks again for
> > fixing this.
> >
> > Reviewed-by: Brett Creeley <brett.creeley@xxxxxxx>
>
> Hi Brett and Thomas,
>
> Maybe I misunderstand things, if so I apologise.
>
> If this patch fixes a bug - e.g. the may observe a system crash -
> then it should be targeted at net and have a Fixes tag. Where the
> Fixes tag generally cites the first commit in which the user may
> experience the bug.
>
> If, on the other hand, this does not fix a bug then the patch
> should be targeted at net-next and should not have a Fixes tag.
>
> In that case, commits may be cited using following form in
> the commit message (before the Signed-off-by and other tags).
> And, unlike tags, it may be line wrapped.
>
> commit 0f3154e6bcb3 ("ionic: Add Tx and Rx handling")
>
> E.g.: This was introduce by commit 0f3154e6bcb3 ("ionic: Add Tx and Rx
> handling").
>
> I hope this helps. If not, sorry for the noise.

Simon,

I suspect you are right and this probably shouldn't be categorized as a 
bug fix since the change only addresses a corner case that would happen 
if the DMA mapping API(s) return 0 as a valid adddress, which wouldn't 
cause a crash with/without this patch.

Thanks for the feedback.

Brett

> ...