Re: [RFC PATCH] Added the logic to decode 32 bit ldr/str post-indexing instructions

[Julien Grall <julien.grall.oss@xxxxxxxxx>]

Hi,

On Mon, 22 Nov 2021 at 19:59, Stefano Stabellini <sstabellini@xxxxxxxxxx> wrote:
>
> On Mon, 22 Nov 2021, Ayan Kumar Halder wrote:
> > Stefano > It doesn't look like we are setting dabt->write anywhere.
> >
> > Ayan > Yes, this is a miss. Depending on the opc, this should be upadeted
> > accordingly in decode_64bit_loadstore_postindexing().
> >
> > Stefano > Also, is info.gpa in try_handle_mmio already updated in the
> > pre-index
> > case? If not, do we need to apply the offset manually here?
> >
> > Ayan > Sorry, I did not understand you. This patch is only related to the 
> > post
> > indexing ldr/str issue. Why do we need to check for pre-indexing ?
>
> I thought you were trying to handle both post-indexing and pre-indexing.
> It is OK if you intend to only handle post-indexing but considering that
> most of the code is shared between the two, we might as well also make
> pre-indexing work (unless it turns out it is more difficult).

Wouldn't this effectively be dead code?

>
> In the pre-indexing case, I would imagine we need to update the base
> address before taking any other actions.

>From my understanding, this would have already been performed by the
HW when the syndrome is valid. This may also be the case for
the non-valid case, but I haven't checked the Arm Arm.

Cheers,