On Thu, Sep 15, 2011 at 09:37:48AM +0200, Alexander Graf wrote:
>
> On 15.09.2011, at 05:19, David Gibson wrote:
>
> > On Wed, Sep 14, 2011 at 10:42:52AM +0200, Alexander Graf wrote:
> >> We currently load a device tree blob and then just take its size x2 to
> >> account for modifications we do inside. While this is nice and great,
> >> it fails when we have a small device tree as blob and lots of nodes added
> >> in machine init code.
> >>
> >> So for now, just make it 20k bigger than it was before. We maybe want to
> >> be more clever about this later.
> >
> > In fact, one of the few things I can think of that might justify
> > qemu's "abstraction" of the libfdt interface, is that the wrappers
> > could be modified to detect -FDT_ERR_NOSPACE and realloc()
> > appropriately.
>
> Oh, yeah, that sounds like a very good idea!
>
> > Otherwise the wrappers, which are limited and not notably simpler to
> > use than the raw libfdt functions seem pretty pointless to me.
> >
> > Not that I'm biased as the author of libfdt or anything :).
>
> I agree that the wrappers are not all that overly useful atm. I was
> actually very close to just ripping them out completely instead of
> extending them for new functionality. I did have the feeling that
> wrapping libfdt would give us a few benefits, maybe even the chance
> of getting rid of #ifdefs in target code.
Hrm, maybe. Can't really see it. Of course, my preference would be
to get rid of those #ifdefs by embedding libfdt in qemu so it's always
there.
> Could you please put this on your todo list? We should probably
> force every target code in QEMU to only use the wrappers and
> dynamically realloc() in them.
Uh, sure, but it's a long list and it won't be near the top.
The wrappers would need to be a lot more extensive to do this. I use
libfdt directly in the spapr code for a reason.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
