Re: Embedded planet 860T CLLF HELP!

Re: Embedded planet 860T CLLF HELP!

[Dan Malek]

Alan Mimms wrote:

> I built the 2.2.13 from the source tarball, only changing .config to
> 860T CPU support,

Well, here is a lesson.  You just can't edit .config.  You have
to run 'make config', as there are more files than just .config
that have context about the kernel configuration.

You could also _ask_ before starting, as I have a ready to
go set of sources for the CLLF on my file server.  You can
use anonymous FTP to mail.netx4.com and pick up cllf-2.2.13.tgz.

I don't know how long I will keep this stuff there, but it is
there now if you want it (along with some other things).


FYI, I _really_ have the FEC driver screwed up for the CLLF, so
don't bother trying to use it.  I made lots of changes yesterday,
it still isn't perfect, and I am travelling for the next ten
days or so.  I will finish it when I return, and may have the
opportunity to work on it during my travel.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[David Clear]

Dan,

Thanks for your help.  I followed a different path to Alan, and came up
with different failures.

I ran 'make config' to set up the RPX CLLF.  The kernel built fine, but
then I found I had to edit the Makefile to make it build coffboot and
then mbxboot.  That gave me a zImage file that I could convert into
zi.cllf.

I got this to boot - it mounted its NFS filesystem fine.  But after a
while (typing ls -lR usually does the trick) it crashes with things like:

NIP: A000FD38 XER: A000FD39 LR: A000FD39 REGS: c0dc5cd0 TRAP: 0400
MSR: 08209032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
TASK = c0dc4000[7] 'ls' mm->pgd c0dc3000 Last syscall: 4
last math 00000000
GPR00: A000FD39 C0DC5D80 C0DC4000 00009032 00000000 00000002 00000020
C00BFC90
GPR08: 00000020 00000000 00000000 C0DC5D80 39000030 00254D6C C012B8A8
00000043
GPR16: FFF29664 00000000 0000003D 3000D000 C01E8BE8 C01F1B54 C009B7C8
3000D000
GPR24: C0DC5DD8 C0DC5DD8 C01E9060 C01E8000 00000FF0 00009032 20203139
00000000
Call backtrace:
A000FD39 C000CA7C C009B9FC C0096E58 C0025174 C00022D8 FFFFFFFF
3009D660 3009D9AC 3009E398 00204704 00203504 0020185C 002012B8
Kernel panic: kernel access of bad area pc a000fd38 lr a000fd39 address
A000FD38
 tsk ls/7

and

NIP: C0006AC4 XER: A000FD39 LR: C00A90D8 REGS: c0dc5a90 TRAP: 0300
MSR: 00009032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
TASK = c0dc4000[13] 'ls' mm->pgd c0dc3000 Last syscall: 90
last math 00000000
GPR00: 00036150 C0DC5B40 C0DC4000 00000070 00000000 C0DC5BD0 C01F2920
00000000
GPR08: FA200000 00000129 00000000 FA200A00 95500033 00254D6C C012B8A8
00000043
GPR16: FFF29664 00000000 00000004 00000000 00001032 00DC5BC0 00000000
C0002544
GPR24: C00043A4 C01F26E8 00000000 00000002 C00CCC74 FA202890 00000000
C0135260
Scheduling in interrupt
page fault in interrupt handler, addr=0
NIP: C000CBF8 XER: 8000FD39 LR: C000CBF8 REGS: c0dc5950 TRAP: 0300
MSR: 00009032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
TASK = c0dc4000[13] 'ls' mm->pgd c0dc3000 Last syscall: 90
last math 00000000
GPR00: C000CBF8 C0DC5A00 C0DC4000 00000018 00000001 00000018 C0138720
FA202830
GPR08: 00000018 C00E0000 FA202838 C0DC5940 0000000D 00254D6C C012B8A8
00000043
GPR16: FFF29664 00000000 00000004 00000000 00009032 00DC5A80 00000000
C0002544
GPR24: C00090F8 C01F26E8 C0DC5A38 C0DC5A38 C00D0000 00000000 00000000
C0DC5A00
Scheduling in interrupt
... forever.

Anyhow, I've pulled your CLLF sources and am trying it now.

Thanks for your help.

David.


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Dan Malek]

David Clear wrote:


> In retrospect, I think "make zImage" would have helped me out too...

Ya' know, there are help files on ppc.kernel.org/embedded that
describe how to do all of this in detail.  Even one specifically
for the CLLF.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Alan Mimms]

YEYEYEYEHAHAHAHAHA!  I got to the bash prompt with Dan's new source 
tree.  Possibly a driver bug in the source tree I had downloaded from 
the linuxppc.org embedded PPC FTP area?  Now to work....  Thanks very 
much Dan.

a

>>>>>>>>>>>>>>>>>> Original Message <<<<<<<<<<<<<<<<<<

On 12/30/99, 4:31:38 PM, David Clear <davidc@uswest.net> wrote regarding 
Re: Embedded planet 860T CLLF HELP!:


> David Clear wrote:

> > I ran 'make config' to set up the RPX CLLF.  The kernel built fine, 
but
> > then I found I had to edit the Makefile to make it build coffboot and
> > then mbxboot.  That gave me a zImage file that I could convert into
> > zi.cllf.

> In retrospect, I think "make zImage" would have helped me out too...

> sigh...

> David.


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Graham Stoney]

Dan Malek writes:
> You could also _ask_ before starting, as I have a ready to
> go set of sources for the CLLF on my file server.  You can
> use anonymous FTP to mail.netx4.com and pick up cllf-2.2.13.tgz.

Excellent! I can confirm that this kernel appears to have fixed the crashing
problems I was having on one of our CLLF boards.

> FYI, I _really_ have the FEC driver screwed up for the CLLF, so
> don't bother trying to use it.  I made lots of changes yesterday,
> it still isn't perfect, and I am travelling for the next ten
> days or so.  I will finish it when I return, and may have the
> opportunity to work on it during my travel.

Are the FEC changes purely related to the different PHY, or is there more to
it than that?

Thanks!
Graham

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Dan Malek]

Graham Stoney wrote:


> Are the FEC changes purely related to the different PHY, or is there more to
> it than that?

Just the PHY, and that I should really make the driver a little
smarter about automatically configuring them.

By the way, with help from Jim Lewis, we now have this
driver working (undocumented bits in the PHY that need to be
set accordingly).  I will just drop this on the server, and
also include it in the source tree updates.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Graham Stoney]

Dan Malek writes:
> By the way, with help from Jim Lewis, we now have this
> driver working (undocumented bits in the PHY that need to be
> set accordingly).  I will just drop this on the server, and
> also include it in the source tree updates.

Even more excellent! We came across these undocumented bits too while trying
to get it talking to the PHY, but it sounds like you beat us to it.

Which ftp server is it on? Do you mean mail.netx4.com or linuxppc.cs.nmt.edu?

Thanks,
Graham

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[duncanp]

On 13 Jan, Dan Malek wrote:
> Just the PHY, and that I should really make the driver a little
> smarter about automatically configuring them.
>
> By the way, with help from Jim Lewis, we now have this
> driver working (undocumented bits in the PHY that need to be
> set accordingly).  I will just drop this on the server, and
> also include it in the source tree updates.

Cool. I was fiddling with this just before christmas, and the
undocumented bits were making life interesting...

This code doesn't seem to have made it into your source tree on netx4.
Have you checked it into the CVS archive yet? If so, could someone
enlighten me on the address of the CVS server, as the link on
ppc.kernel.org with the details has been down for the last coupla days.

Cheers,

Dunk.

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Dan Malek]

I just placed the latest MPC860T FEC driver on ppc.kernel.org.
This seems to work with the Embedded Planet CLLF.  It is a
pretty bad hack, as the driver should really determine many
of the things in #ifdef's dynamically.

Another thing.  On the "back" side of the CLLF, at the edge
near the four switch pack, you will find an unpopulated R08.
Place a jumper wire across this, as it is the source of the
PHY link change interrupt to the Port C pin.  Future CLLF boards
will have this populated.

And another thing.  This driver doesn't seem to work correctly
on the 860P version of the boards.  I thought I found all 860P
differences, but I guess I missed something.  I am working on it,
and if you have one of these boards, please let me know.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Ralf HECKHAUSEN]

>I just placed the latest MPC860T FEC driver on ppc.kernel.org.
>This seems to work with the Embedded Planet CLLF.  It is a
>pretty bad hack, as the driver should really determine many
>of the things in #ifdef's dynamically.
>
>Another thing.  On the "back" side of the CLLF, at the edge
>near the four switch pack, you will find an unpopulated R08.
>Place a jumper wire across this, as it is the source of the
>PHY link change interrupt to the Port C pin.  Future CLLF boards
>will have this populated.
>
>And another thing.  This driver doesn't seem to work correctly
>on the 860P version of the boards.  I thought I found all 860P
>differences, but I guess I missed something.  I am working on it,
>and if you have one of these boards, please let me know.

Try to initialize PDDIR with 0x1fff instead of 0x1c58.

>	-- Dan

Ralf


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Dan Malek]

Ralf HECKHAUSEN wrote:


> Try to initialize PDDIR with 0x1fff instead of 0x1c58.

Right, now I remember....that doesn't work on pre Rev-D 860T parts.
Guess I need to add some silicon revision detection and set
accordingly.  Next version.



	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[duncanp]

On 19 Jan, Dan Malek wrote:
> I just placed the latest MPC860T FEC driver on ppc.kernel.org.

This works well for me - thanks. I had some troubles when
auto-negotiate was enabled - the PHY decided it was running at 100MBps,
but the link did not come up. Forcing it to 10MBps worked around it - i
think this may have something to do with the switch i had the board
plugged into.

> This seems to work with the Embedded Planet CLLF.  It is a
> pretty bad hack, as the driver should really determine many
> of the things in #ifdef's dynamically.
>

If you haven't done any more work on this, i'll be happy to have a look
at neatening things up.

Dunk.


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Embedded planet 860T CLLF HELP!

[Dan Malek]

duncanp@research.canon.com.au wrote:

> This works well for me - thanks.

Good.  Just remember, playing with two physical Ethernet ports
is sometimes a challenge....make sure you use separate network
numbers :-).

I had the chance to use a 66 MHz 860P CLLF.  I ran a TCP
source/sink set of programs between my iMac and the CLLF.  The
TCP/IP interface ran 30 Mbits/sec, so we are getting there.
I am assuming the CLLF was the slow end, I guess I will try
my G4 and make sure :-).


> If you haven't done any more work on this, i'll be happy to have a look
> at neatening things up.

I wouldn't spend any time right now.  The QS6612 has been
discontinued and Embedded Planet is changing the PHY.  We can
update the driver when the new PHY appears, and I will keep
you in mind :-).

Thanks.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/