Patch for Cross-Reference Phandles

[Adrian Chadd]

... holy crap. :-)

erm, try writing some code that does something with it:

int foo[2] = { 0, 0 };

main()
{
    int i;
    for (i = 0; i < 2; i++) {
        foo[i] = i;
    }
}

.. and then make foo[2] be foo[512] and explicitly set it up to be zero'ed.
see what happens.



-a


On 14 September 2013 10:55, Ian Lepore <ian at freebsd.org> wrote:

> On Sat, 2013-09-14 at 10:27 -0700, John-Mark Gurney wrote:
> > Ian Lepore wrote this message on Sat, Sep 14, 2013 at 09:43 -0600:
> > > On Sat, 2013-09-14 at 09:06 -0500, Nathan Whitehorn wrote:
> > > > On 09/14/13 08:31, Marius Strobl wrote:
> > > > > On Thu, Sep 12, 2013 at 10:07:18AM -0500, Nathan Whitehorn wrote:
> > > > >> On 09/08/13 13:54, Nathan Whitehorn wrote:
> > > > >>> Open Firmware has three namespaces for handles:
> > > > >>> 1. Instance handles, for open devices
> > > > >>> 2. Package handles for the client interface
> > > > >>> 3. Package handles for device tree cross references
> > > > >>>
> > > > >>> On Powermac hardware, we assume that (2) and (3) are identical
> and
> > > > >>> call both phandles. On embedded FDT systems, you can't open
> devices
> > > > >>> and so we abuse ihandle_t for (3). IBM pSeries hardware,
> however, has
> > > > >>> all three things. With that in mind, I'd like to start separating
> > > > >>> them. The patch at
> > > > >>> http://people.freebsd.org/~nwhitehorn/xref_phandle.diff adds a
> new
> > > > >>> function (OF_child_xref_phandle) that takes a phandle of type
> (3) and
> > > > >>> turns into one of type (2) by searching for entries named
> "phandle",
> > > > >>> "ibm,phandle", or "linux,phandle" in the tree. This should work
> for
> > > > >>> FDT as well, but is not connected in the patch to anything
> actually
> > > > >>> FDT related.
> > > > >>>
> > > > >>> Comments would be appreciated. I'd like to get to get this as in
> as
> > > > >>> soon as possible (given the HEAD freeze) otherwise.
> > > > >>> -Nathan
> > > > >> Since I haven't heard anything, it shouldn't affect any existing
> > > > >> platforms or code, and it would be nice to have this interface
> available
> > > > >> in the 10.x series, I plan to ask re@ for commit approval
> tomorrow.
> > > > >> Please let me know if there are any objections.
> > > > > Technically that patch is fine. It would be nice if you could fix
> the
> > > > > style bugs before committing, though. In OF_child_xref_phandle(),
> the
> > > > > variables should go above the comment and you don't need to
> initialize
> > > > > the static global one in ofw_pcibus.c to zero explicitly. If you do
> > > > > nevertheless, spaces should go before and after the '='.
> > > > >
> > > >
> > > > Thanks! I think the explicit zero initialization adds some clarity
> for
> > > > people reading the code, so I'd like to keep it. The rest of the
> changes
> > > > have been made.
> > > > -Nathan
> > >
> > > The explicit zero init also makes it 4 bytes harder to fit the kernel
> > > into a small flash on embedded systems.  We should be eliminating
> > > zero-inits that move items from bss to data, not adding new ones.
> >
> > I just ran a simple test:
> > int foo = 0;
> >
> > and compiled it, both with:
> > FreeBSD clang version 3.3 (tags/RELEASE_33/final 183502) 20130610
> > Target: x86_64-unknown-freebsd10.0
> > Thread model: posix
> >
> > and:
> > FreeBSD clang version 3.3 (trunk 178860) 20130405
> > Target: armv6--freebsd
> > Thread model: posix
> >
> > and:
> > gcc (GCC) 4.2.1 20070831 patched [FreeBSD]
> > Copyright (C) 2007 Free Software Foundation, Inc.
> > This is free software; see the source for copying conditions.  There is
> NO
> > warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
> PURPOSE.
> >
> > all are smart enough to know that zero init is the same as locating it
> > in bss, and do so... and not put it in data...  Oh, these were all
> > compiled w/o any options, just <compile> -c foo.c...
> >
>
> Huh.  That's pretty surprising to me, because IMO, an explicit zero-init
> is NOT the same as putting it in bss, it's explicitly a way to say "I
> want this in the data segment."  I guess there are other ways to control
> placement these days.
>
> -- Ian
>
>
> _______________________________________________
> freebsd-ppc at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-ppc
> To unsubscribe, send any mail to "freebsd-ppc-unsubscribe at freebsd.org"
>