Re: [PATCH 1/4] xen/arm: traps: Fix MISRA C 2012 Rule 8.4 violations

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Tue, 5 Jul 2022, Julien Grall wrote:
> Hi Xenia,
> 
> On 05/07/2022 22:49, Xenia Ragiadakou wrote:
> > On 7/6/22 00:28, Julien Grall wrote:
> > > Hi Xenia,
> > > 
> > > On 05/07/2022 22:02, Xenia Ragiadakou wrote:
> > > > Add the function prototypes of the functions below in <asm/processor.h>
> > > > header
> > > > file so that they are visible before the function definitions in
> > > > traps.c.
> > > > enter_hypervisor_from_guest_preirq()
> > > > enter_hypervisor_from_guest()
> > > > do_trap_hyp_sync()
> > > > do_trap_guest_sync()
> > > > do_trap_irq()
> > > > do_trap_fiq()
> > > > leave_hypervisor_to_guest()
> > > 
> > > I have actually came across those missing prototypes in the past and I
> > > have refrained to add them because they are not meant to be called from C.
> > > 
> > > While I agree this is a violation of Misra C, I am still not convinced
> > > they should be added (there is no need to verify the prototype).
> > 
> > Yes, there is no need. Here, I decided to follow the example of
> > do_trap_hyp_serror() and do_trap_guest_serror() for consistency.
> 
> do_trap_guest_serror() is called from C code (see arch/arm32/traps.c).
> 
> do_trap_hyp_serror() is called only from assembly. I would argue that if you
> want consistency, then you should drop the prototype rather than modifying 90%
> of the other examples.
> 
> Otherwise, this is not really consistency but more a design choice (you want
> to be fully compliant with MISRA).

Actually I am not sure that MISRA requires prototypes for assembly
functions only meant to be called from assembly. I think MISRA requires
detailed docs for anything assembly, but I don't think it requires C
prototypes for all assembly functions.

So I think we could skip them for now. At some point we'll discuss what
we need to do for the assembly code but we haven't tackled that yet.