Re: [XEN PATCH v5 7/8] x86/mm: add defensive return

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Fri, Aug 30, 2024 at 08:17:40PM +0100, Andrew Cooper wrote:
> On 30/08/2024 10:18 am, Roger Pau Monné wrote:
> > On Thu, Aug 29, 2024 at 09:04:37AM +0200, Jan Beulich wrote:
> >> On 29.08.2024 02:35, Stefano Stabellini wrote:
> >>> On Mon, 29 Jul 2024, Stefano Stabellini wrote:
> >>>> On Mon, 29 Jul 2024, Federico Serafini wrote:
> >>>>> Add defensive return statement at the end of an unreachable
> >>>>> default case. Other than improve safety, this meets the requirements
> >>>>> to deviate a violation of MISRA C Rule 16.3: "An unconditional `break'
> >>>>> statement shall terminate every switch-clause".
> >>>>>
> >>>>> Signed-off-by: Federico Serafini <federico.serafini@xxxxxxxxxxx>
> >>>> Reviewed-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> >>>>
> >>>>> ---
> >>>>> No changes from v3 and v4, further feedback on this thread would be 
> >>>>> appreciated:
> >>>>> https://lists.xenproject.org/archives/html/xen-devel/2024-07/msg00474.html
> >>> Looking at the older threads, I looks like Jan suggested EACCES, I also
> >>> think it is marginally better. My R-b stands also for EACCES. Jan, I
> >>> think you should go ahead and fix on commit
> >> No, I very definitely want a 2nd x86 maintainer opinion here. Or a better
> >> suggestion for the error code to use by anyone. After all, as you confirm,
> >> EACCES is only marginally better.
> > Hm, the only alternative I could suggest is using ERANGE, to signal
> > the value of opt_mmio_relax is out of the expected range, however that
> > could be confusing for the callers?
> >
> > One benefit of using ERANGE is that there's currently no return path
> > in get_page_from_l1e() with that error code, so it would be very easy
> > to spot when an unexpected value of opt_mmio_relax is found.  However
> > there are no guarantees that further error paths might use that error
> > code.
> 
> EPERM and EACCES are both very wrong here.  They imply something that is
> simply not appropriate in this context.
> 
> We use EILSEQ elsewhere for believed-impossible conditions where we need
> an errno of some kind.  I suggest we use it here too.

Fine with me.  With that:

Reviewed-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>

Thanks.