Re: [OpenBIOS] [PATCH 0/2] Sparc64: virtio-block and interrupt support

[Artyom Tarasenko]

On Sun, May 13, 2012 at 7:56 AM, Jakub Jermar  wrote:
> On 05/13/2012 12:00 AM, Artyom Tarasenko wrote:
> > On Sat, May 12, 2012 at 7:06 PM, Artyom Tarasenko  wrote:
> > > On Sat, May 12, 2012 at 6:29 PM, Blue Swirl  wrote:
> > > > On Sat, May 12, 2012 at 3:57 PM, Artyom Tarasenko  wrote:
> > > > > On Sat, May 12, 2012 at 1:42 PM, Blue Swirl  wrote:
> > > > > > Based on Artyom's Sparc64 work (thanks a lot!), I added some minimal
> > > > > > support for virtio-block and interrupt nodes.
> > > > > >
> > > > > > With these patches and QEMU HEAD, Linux can boot like Artyom describes in
> > > > > > http://tyom.blogspot.de/2012_05_01_archive.html
> > > > > > but without any manual OpenBIOS setup.
> > > > >
> > > > > My test kernel - 3.3.2 - still needs the ebus interrupt mapping properties:
> > > > >
> > > > > cd /pci@1fe,0/ebus
> > > > > 000001fe encode-int 020003f8 encode-int encode+ 1 encode-int encode+
> > > > > ffe29140 encode-int encode+ 2b encode-int encode+ " interrupt-map" property
> > > > > 000001ff encode-int ffffffff encode-int encode+ 00000003 encode-int encode+
> > > > > " interrupt-map-mask" property
> > > > > device-end
> > > > > boot
> > > >
> > > > Updated patch attached. This change seems to help the serial driver.
> > >
> > > Looks good. The result seems to be identical with my hack, but there
> > > is one point I'd like to discuss:
> > >
> > > cd /pci@1fe,0/pci-ata  ok
> > > .properties
> > > name                      "pci-ata"
> > > [...]
> > > interrupts                0
> > >
> > > I'm not sure that this is a valid value for the interrupts property.
> >
> > Tarl, Jakub, can you comment whether 0 is allowed?
>
> According to the UltraSPARC IIi User's Manual, section 19.3.3, INO 0
> (again, relative to the parent node mapping), would be INTA# on bus 0
> and INO 1 would be INTB#.

Yes, this is how it's wired in qemu too. But I meant the OBP level.
What worries me is that this construction seems to be never used in
sun4u OBP:

$  grep interrupts linux/kernel/git/davem/prtconfs/* | grep -c 00000001
402
$ grep interrupts linux/kernel/git/davem/prtconfs/* | grep -c 00000000
2

And the both hits are sun4v:

$  grep interrupts ~/openbios/linux/kernel/git/davem/prtconfs/* | grep 00000000
linux/kernel/git/davem/prtconfs/t1000:            interrupts:  00000000
linux/kernel/git/davem/prtconfs/t2000_1:            interrupts:  00000000

Can it be, that 0 is (or used to be) some magic constant?
> > > I guess it has to be changed back to 1, and the
> > > additional interrupt mappings have to be added to the root pci node.

Artyom