Re: FirePower (was Re: TODO list for ofppc)

[Tim Rightnour <root%garbled.net@localhost>]

On 23-Jan-2008 Izumi Tsutsui wrote:
> Hmm, but isn't it better to adjust scells?
> It's 2, but sizes look to appear at offset 4 (acells is 3 there).
> Anyway, I've tried a kernel without the "+ 1".

On my machines, there is another field in there.  I'm assuming it's some kind
of parent field.  Thats why I have the +1.  We may end up just needing a
model-based quirk table or something.


> On FirePower:
> found a map reclen=5 cur=0 len=160
Looks correct.
> FOUND PCI RANGE
It found 8, thats good.
> returning with CUR=7
> cur == 7
> addr=0x80800800 size=0x800 type=0
> addr=0x80801000 size=0x1000 type=0
> addr=0x80802000 size=0x1800 type=0
> addr=0x80804000 size=0x2000 type=0
> addr=0x80808000 size=0x1000000 type=0
> addr=0x80000000 size=0x1000000 type=1
> addr=0x80100000 size=0x2000000 type=1

And this is where it all goes wrong.

00000000 00000000 00000000 80800800 00000800
00000800 00000000 00000000 80801000 00000800
00001000 00000000 00000000 80802000 00000800
00001800 00000000 00000000 80804000 00000800
00002000 00000000 00000000 80808000 00000800
01000000 00000000 00000000 80000000 00010000
01000000 00000000 00010000 80100000 3e800000
02000000 00000000 00000000 c0000000 3e000000

Its not picking up the sizes correctly:

> addr=0x80000000 size=0x2100000 type=1

That should have been from 80000000 with a size of somewhere in the 3e900000
area.  It's not clear to me where it got 0x2100000 from.  Possibly more
off-by-one errors.

> HOLES FOUND
> Data access exception

Even then however, I wonder why it took a data access exception there.  I
wonder what it was attempting to do.

> On ESTRELLA:
01000000 00000000 00000000 80000000 00010000
01000000 00000000 01000000 81000000 3e800000
02000000 00000000 00000000 c0000000 3c000000

Again, it gets the sizes wrong.  I assume due to the same problem.

> Unhandled pic node type node=7e5ac40

Here, it is searching for things marked interrupt-controller, and then trying
to guess what kind of interrupt controller it is.  It failed.  I'll need the
.properties of whatever node has that marking to try and point it at the right
thing.

> So both have #address-cells "1".

Sigh.

---
Tim Rightnour <root%garbled.net@localhost>
NetBSD: Free multi-architecture OS http://www.netbsd.org/
Genecys: Open Source 3D MMORPG: http://www.genecys.org/