Re: [Xen-devel] [PATCH v1 2/8] xen/arm: io: Extend write/read handler to pass the register in parameter

[Ian Campbell <ian.campbell@xxxxxxxxxx>]

On Tue, 2015-09-29 at 12:00 +0100, Julien Grall wrote:
> Hi Ian,
> 
> On 29/09/15 11:51, Ian Campbell wrote:
> > On Mon, 2015-09-28 at 17:35 +0100, Julien Grall wrote:
> > > Hi Ian,
> > > 
> > > On 25/09/15 17:36, Ian Campbell wrote:
> > > > On Fri, 2015-09-25 at 15:51 +0100, Julien Grall wrote:
> > > > > From: Julien Grall <julien.grall@xxxxxxxxxx>
> > > > > 
> > > > > Rather than letting each handler to retrieve the register used by
> > > > > the
> > > > > I/O access, add a new parameter to pass the register in
> > > > > parameter.
> > > > > 
> > > > > This will help to implement generic register manipulation on I/O
> > > > > access
> > > > > such as sign-extension and endianess.
> > > > > 
> > > > > Read handlers need to modify the value of the register, so a
> > > > > pointer
> > > > > to
> > > > > it is given in argument. Write handlers shouldn't modify the
> > > > > register,
> > > > > therfore only a plain value is given.
> > > > 
> > > > "therefore"
> > > > > 
> > > > > Signed-off-by: Julien Grall <julien.grall@xxxxxxxxxx>
> > > > 
> > > > Acked-by: Ian Campbell <ian.campbell@xxxxxxxxxxx
> > > > 
> > > > Is it the case that read handlers today only update *r on success? 
> > > 
> > > Yes, each read handler only update the register mentioned in
> > > dabt.reg.
> > 
> > I meant that they do not touch it on failure.
> 
> Sorry, somehow I read as "Is only the register pointed by *r updated by
> the read handlers"?

Yes, my original comment was ambiguously parsable that way, sorry.

> AFAICT, it's always the case. When the handler returns 0 (i.e failure),
> a data abort will be injected to the domain unless the handler decide to
> crash the domain.

The important thing is whether any path sets *r and then returns 0, I
think.

> I don't think it's necessary to mandate/enforce as the worst that can
> happen is a guest register is corrupted. Although, the guest will
> unlikely get continue after a data abort on MMIO we emulate.

True.

> 
> Regards,
> 

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel