we ethernet on prep

rootgarbled.net wrote:

> However, your we card doesn't get
> picked up, so you still need the weisa
attachment.  I'm not sure how to ensure
> that if the card gets picked up by pnpbus, it won't
get picked up again by isa.

Can we use the direct config by isa_attach_knowndevs() in
sys/dev/isa/isa.c?
(we could put device info from residual data in
isa_attach_hook())
---
Izumi Tsutsui

On 10-Jun-2006 Izumi Tsutsui wrote:
> It looks ISA mem space conflicts against PCI memory
space.
> The attached patch fixes the problem on my 7248-133.
> (I'm not sure why extent_alloc_region(9) in
memio_map() doesn't fail though)

That patch made your we card work?  I tried it and still get
the exact same
problem...

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

rootgarbled.net wrote:

> > It looks ISA mem space conflicts against PCI
memory space.
> > The attached patch fixes the problem on my
7248-133.

> That patch made your we card work?  I tried it and
still get the exact same
> problem...

Umm, after I update sources and recompile a kernel,
now we(4) works even *without* the patch. Weird.

Anyway, I still have a concern that the PCI VGA device
requires large chunk of PCI memory space (~8MB),
so I wonder the BAT can handle such mappings near
other small mappings.
(I'm not familiar with powerpc though)

Isn't it better to map static 16MB region for ISA mem space
rather than mapping regions with BAT on each
bus_space_map(9) call?
---
Izumi Tsutsui

On 13-Jun-2006 Izumi Tsutsui wrote:
> Umm, after I update sources and recompile a kernel,
> now we(4) works even *without* the patch. Weird.

I don't understand what is going on with we at all here. 
Mine is completely
unusable, and I can't tell if thats because it's broken,
or because prep is
broken.. but yours works occasionally.. which horrifies me
even more.

> Anyway, I still have a concern that the PCI VGA device
> requires large chunk of PCI memory space (~8MB),
> so I wonder the BAT can handle such mappings near
> other small mappings.
> (I'm not familiar with powerpc though)

I doubt it's a problem..

> Isn't it better to map static 16MB region for ISA mem
space
> rather than mapping regions with BAT on each
bus_space_map(9) call?

Well.. it might be..  are you looking at the prep code, or
the powerpc shared
code?  BAT mappings are really fast.. and the preferred way
to wire things up
on powerpc.  Are you talking about how we batmap the whole
PCI IO and MEM
spaces in machdep, or something else?  The big batmaps in
machdep.c are probably
the ideal way to do things.

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

rootgarbled.net wrote:

> Well.. it might be..  are you looking at the prep code,
or the powerpc shared
> code?  BAT mappings are really fast.. and the preferred
way to wire things up
> on powerpc.  Are you talking about how we batmap the
whole PCI IO and MEM
> spaces in machdep, or something else?  The big batmaps
in machdep.c are probably
> the ideal way to do things.

Ah, I just misread the code. I thought
powerpc/bus_space.c:memio_map()
did BAT mappings, but now I see it just looks up BAT mapped
regions.
Sorry for the confusion.

Anyway, I'll check more what trigges the problem on my
7248.
---
Izumi Tsutsui

I wrote:

> Anyway, I'll check more what trigges the problem on my
7248.

Hmm, it looks an old bootloader triggers the we(4) problem.
A kernel loaded by the following version have the problem:
---
>> NetBSD/prep BOOT, Revision 1.5
>> (tsutsuimirage, Tue Apr  6 22:42:26 JST 2004)
---
(sorry for too old one)

But the we(4) works fine on the same kernel loaded by the
latest loader:
---
>> NetBSD/prep BOOT, Revision 1.7
>> (tsutsuimirage, Sat Jun 10 10:38:59 JST 2006)
---

Note I booted kernels via network (pcn0) and
used the same latest nbprep-mkbootimage.
---
Izumi Tsutsui

On 14-Jun-2006 Izumi Tsutsui wrote:
> Hmm, it looks an old bootloader triggers the we(4)
problem.
> A kernel loaded by the following version have the
problem:

Wow.  umm.. wow.  I need to try that.  Thats crazy.

Even crazier.. I still do my test netboots with 1.5... oh
geez.

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

rootgarbled.net wrote:

> Even crazier.. I still do my test netboots with 1.5...
oh geez.

I also tried much older boot (not boot_com0) and the same
kernel:
---
>> NetBSD/prep BOOT, Revision 1.2
>> (tsutsuimirage, Sun Jun 28 20:43:51 JST 2001)
---
then we(4) also works fine on the kernel.

Attaching uninitilized VGA (S3 86C864-1 on my 7248)
prevents some ISA memory space access?
---
Izumi Tsutsui