Current state of MPC8260 Linux

Current state of MPC8260 Linux

[diekema_jon]

	* My observations as to the current state of the MPC8260
	version of Linux:

	{

	 If anyone has additional wisdom on this topic, please send it to
	 the mailing list <linuxppc-embedded@lists.linuxppc.org> or
	 Jon Diekema <diekema_jon@si.com>

	}

	- MontaVista Software, http://www.mvista.com/, and at least one other
	  company contracted with Dan Malek <dan@netx4.com> to perform the
	  MPC8260 Linux port.  Dan Malek was heavily involved with the MPC8xx
	  Linux port, so he was a natural choice to do the MPC8260 port.

	- The Linux kernel is covered by the GPL, GNU Public License, version
	  2.  See the COPYING from the Linux kernel distribution.

	  From the GPL version 2:

	  "2. You may modify your copy or copies of the Program or any portion
	      of it, thus forming a work based on the Program, and copy and
	      distribute such modifications or work under the terms of Section 1
	      above, provided that you also meet all of these conditions:

	      ...

	      b) You must cause any work that you distribute or publish, that in
	      whole or in part contains or is derived from the Program or any
	      part thereof, to be licensed as a whole at no charge to all third
	      parties under the terms of this License."

	  Therefore, MontaVista software is compelled to release their MPC8260
	  changes back to the Open Source community, however the timing
	  isn't defined.  I would expect MontaVista to release these changes
	  either before or shortly after their MPC8260 work is released.

	- Paul Staudacher <paul_staudacher@mvista.com>, Sales,
	  indicated that MontaVista will be releasing the MPC8260 Linux changes,
	  but it is not currently scheduled.

	- Jim Lewis <jlewis@mvista.com>, Sr. Field Applications Engineer, told me
	  that a production MPC8260 Linux release for the EST SBC8260 board,
	  http://www.estc.com/products/SBC/SBC8260_8240.html, is targeted for
	  May/June 2000 timeframe.

	  I asked Jim Lewis what features of the SBC8260 will be supported.
	  Here is his response:

	  From Jim Lewis <jlewis@mvista.com>:

	  Right now, we have data cache turned off because Cache Snooping is
	  not working. We are using 10 Mbit ethernet on an SCC rather than
	  100MBit on an FCC. The first release will have those two things fixed
	  at a minimum. So here is a minimum list of features:

	         NFS or Initial Ram Disk Root FS
	         Async serial using SMC's
	          Ethernet on SCC
	         10/100 Ethernet on FCC
	         Proper cache operation
	         x86 Linux hosted C/C++ Cross Dev environment
	         Remote GDB KErnel debug over serial
	         Remote GDB application debug over Serial/Ethernet

	- MontaVista makes its money through annual support contracts
	  per contribution engineer.  A contributing engineer is defined to
	  be anyone making a software contribution to the project.  The
	  support costs aren't inexpensive, however you usually get what
	  you pay for.  For the details on the MontaVista pricing, you
	  should contact Paul Staudacher <paul_staudacher@mvista.com>.

	- The PPC Linux development is being done with the BitKeeper tools,
	  http://www.bitkeeper.com/.  In order to get the BitKeeper tools,
	  you need to fill out an NDA.  The BitKeeper developers are scrambling
	  to make the tools work for the PPC Linux developement people.

	  I have filled out a NDA, and both FAXed and US mailed the NDA to BitKeeper.
	  I am waiting for a download login and password, but haven't received
	  it yet.  I called BitKeeper and they returned my call on the same day :).
	  They explained they were dragging their feet on adding more people to
	  the Beta program because of an antipicated file format change that
	  was going to happen in the next week.  This change would force everyone
	  to use a new release of the BitKeeper software.

	  BitKeeper wanted me to avoid the hassle of performing the upgrade.

	  It has been week since the telephone conversation with BitKeeper,
	  and I am still waiting to get the software.  I suspect they are
	  under the gun in getting the bugs worked out.

	- The PPC Linux development bits are available with rsync,
	  http://rsync.samba.org/rsync/, from daily snapshots of the
	  BitKeeper repository.

	  rsync -avz --delete hq.fsmlabs.com::linuxppc_2_3 linuxppc_2_3
	  rsync -avz --delete hq.fsmlabs.com::linuxppc_2_2 linuxppc_2_2

	  By looking at what files are changing, you get a sense of what the
	  developers are working on.  I would rather have access to the
	  BitKeeper tools first hand, so I could see who is doing what.

	- The PPC Linux develeopment BitKeeper repository is being hosted
	  by Finite State Machine Labs, FSMLabs, http://hq.fsmlabs.com/.
	  FSMLabs is a service of VJY Associates, LLC.

	  Who is FSMLabs?

	  From http://hq.fsmlabs.com/:

	  "Created by the developers of RealTime Linux (RTLinux[tm]) and one of
	  the main developers of PowerPC Linux, FSMLabs provides operating systems
	  support to commercial vendors of products based on Linux. Providing core
	  kernel expertise and development, we consult on general systems issues
	  with a focus on real-time and embedded systems used in industry and
	  research. In addition, we have cooperative arrangements with companies
	  that provide direct customer support, systems integration and
	  application development."

	  FSMLabs are people behind RTLinux.

	  Cort Dougan is involved with FSMLabs, and Cort has his name on
	  the http://www.ppc.kernel.org/ web site.

	- Hints on how to use BitKeeper, BK:

	  http://lists.linuxppc.org/listarcs/linuxppc-dev/200002/msg00193.html
	  http://lists.linuxppc.org/listarcs/linuxppc-dev/200002/msg00220.html
	  http://lists.linuxppc.org/listarcs/linuxppc-dev/200002/msg00221.html

------------------\\----------------------\\----------------------------
Jon Diekema        |                       | Smiths Industries
(616) 241-8310     |                       | 3290 Patterson Avenue, SE
diekema_jon@si.com \\                      \\  Grand Rapids, MI 49512-1991

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

Re: Current state of MPC8260 Linux

[Dan Malek]

diekema_jon wrote:


>          If anyone has additional wisdom on this topic,


There are people that profusely contribute Open Source software
without complaint, providing clever designs and elegant software
style.  These people have my greatest respect and deepest
gratitude.  I am striving to be like them.

Some of us contribute to the best of our ability, fortunate to
have others realize our value to their success.  To pay our
bills in life, we have agreed to provide production quality
software for their products.  Ultimately, indirectly, albeit
more slowly, they too contribute Open Source software.

Others believe the Open Source train isn't stopping at their
station, failing to realize bleeding edge technology requires
resource investments commensurate with their expected rewards.
Their product development will fail due to the lack of exercising
options that reduce development risk and ensure success.  They
are unlikely to benefit from or contribute Open Source software.


	-- Dan

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

PPC403 bootup problem

[Tom Shaver]

I am trying to boot my IBM 403 eval board using Grant's 403 kernels. First
question is, I don't know what eval bord it is. It doesn't say "OAK" or
anything like that on it. 2nd, the board came with a 403GA processor, in a
socket, which we have replaced with a 403GCX. The ROM monitor eprom routines
startup fine with the GCX installed, recognizing the processor and speed
correctly. It goes through what looks like a normal startup, does a bootp,
loads the linux kernel via tftp, tries to boot the kernel, and hangs
immediately. Here's the transcript of what happens (from the serial port).
Any input appreciated.

=======================================================================

403GCX 8.2 ROM Monitor (5/12/97)

 ------- System Info --------
 Processor speed = 66 MHz
 Bus speed       = 33 MHz
 Amount of DRAM  = 16 MBytes
 ----------------------------

 --- Device Configuration ---
 Power-On Test Devices:
   000  Enabled   System Memory [RAM]
   001  Enabled   Ethernet [ENET]
   004  Disabled  Serial Port 2 [S2]
 ----------------------------
 Boot Sources:
   001  Enabled   Ethernet [ENET]
                  local=206.185.34.211  remote=206.185.34.5
hwaddr=0004ace30b95

   004  Disabled  Serial Port 2 [S2]
                  local=8.1.1.4  remote=8.1.1.1  hwaddr=ffffffffffff
   005  Disabled  Serial Port 1 [S1]
                  Baud = 115200
 ----------------------------
 Debugger: Disabled
 ----------------------------
Press any key to get configuration menu
Checking system RAM
Bank 4 (DRAM): 04 MB   Addr: 0x00c00000
Bank 5 (DRAM): 04 MB   Addr: 0x00400000
Bank 6 (DRAM): 04 MB   Addr: 0x00800000
Bank 7 (DRAM): 04 MB   Addr: 0x00000000
Installed DRAM: 16 MB
Checking DRAM between 0x80000000 and 0x81000000

  System RAM check complete
Booting from [ENET] Ethernet...
Sending bootp request ...


Loading file "/LINUXPPC.IMG" ...
Sending tftp boot request ...
Transfer Complete ...
Loaded successfully ...
Entry point at 0x200000 ...


================================================================


Tom Shaver
Electrical Engineer
Planning Systems Inc.
21294 Johnson Rd
Long Beach, MS 39560
228-863-0007
228-863-0281 Fax


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

Re: Current state of MPC8260 Linux

[Gilbert Coville]

At 6:50 PM -0800 2/22/00, diekema_jon wrote:
>	* My observations as to the current state of the MPC8260
>	version of Linux:
>[...]
>	- MontaVista Software, http://www.mvista.com/, and at least one other
>	  company contracted with Dan Malek <dan@netx4.com> to perform the
>	  MPC8260 Linux port.  Dan Malek was heavily involved with the MPC8xx
>	  Linux port, so he was a natural choice to do the MPC8260 port.
>
>[GPL snippet]
>
>	  Therefore, MontaVista software is compelled to release their MPC8260
>	  changes back to the Open Source community, however the timing
>	  isn't defined.  I would expect MontaVista to release these changes
>	  either before or shortly after their MPC8260 work is released.

MontaVista is 100% committed to the open-source philosophy.  In every case
we have released all sources and/or source changes simultaneously with our
product releases.  In some cases, we have distributed early snapshots of work
before actual release.  We intend to continue to do this.  I do not know the
specific schedule for the 8260 work, but I am quite certain that we will be
releasing its source to the open-source community at large.

We also work closely with Dan Malek.  In some cases we've contracted with
him for work, but in most cases we have leveraged his already extensive
efforts.  This has worked well for us, and I believe it has worked well for
Dan.  With Dan's help, we will be making a better effort to get kernel
changes back into the bitkeeper base sooner, rather than later.

However, resources are always limited.  It takes quite a bit of effort to put
all these things together.

This thread comes at an interesting time, as we have recently completed
the 1.0 version of our 8xx cross-development kit.  In fact, the entire
thing is up on our FTP site at:
	ftp://ftp.mvista.com/pub/hhl-ppc_8xx/

I intend to make a more widespread announcement once our website has an
appropriate link to it, but the bits are there now.

Share and enjoy,
Gilbert

--
Gilbert Coville			MontaVista Software Inc.
gilbert@mvista.com		+1 408-328-9208


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

Re: PPC403 bootup problem

[Grant Erickson]

On Wed, 23 Feb 2000, Tom Shaver wrote:
> I am trying to boot my IBM 403 eval board using Grant's 403 kernels. First
> question is, I don't know what eval board it is. It doesn't say "OAK" or
> anything like that on it.

Judging by the output from the PROM output you included, I can't tell if
it's an "Oak" board or not. It looks like your board has four DRAM SIMM
slots, whereas mine only has two. That issue aside, my guess is that the
boards are otherwise identical. The Linux kernel shouldn't much care
either, as long as the devices are located in the same places in physical
address space:

	DRAM:	0x00000000 - DRAM Size
	ROM:	0xFFF00000 - 0xFFFFFFFF
	S2:	0x7E000000 - 0x7E000007
	S1:	0x40000000 - 0x40000007
	ENET:	0xF4000000 - 0xF400003F

> It goes through what looks like a normal startup, does a bootp, loads
> the linux kernel via tftp, tries to boot the kernel, and hangs
> immediately.

It appears to be hanging, but in reality, it's probably not. At present,
there is no serial console support, so once it makes the jump to 0x200000,
text output to your console ceases. But, rest assured, it's probably still
working in silence.

If you had a hardware emulator/debugger (like RISCWatch), you could probe
the memory location at 'log_buf' and see the actual console output sitting
in it's memory buffer.

I'll try to throw Ben Bodley's NS16550A serial console patches up on the
web site later today. The ideal situation would be to get a 403GC[X] SCC
serial console driver going. Someone might well be working on this at
present.


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

Linux Planet Module Support

[Jo-Ellen F. Mathews]

Hi all,

I've been working on porting our 11Mb wireless LAN device driver to
Embedded Planet's nifty RPX-Lite (MPC823) board.  I've been using Embedded
Planet's wonderful, new Linux Planet development kit which includes
MontaVista's Hard Had Linux 1.0.  I'm loving this kit!

I'm using the cross compiler provided on MontaVista's cd, and I've been
very successful at compiling the default kernel and a few, simple
programs.

The most direct approach (okay, and quickest/easiest) is to keep the
wireless lan driver as a loadable module dependent on pcmcia card
services.  Before attempting to use Magnus Damm's pcmcia patches and
kernel modifications to support pcmcia, I first need to recompile my
kernel for module support.

With CONFIG_MODULES enabled (regardless of what CONFIG_MODVERSIONS or
CONFIG_KMOD are set to), I get the following errors:

ppc_ksyms.c:72: `isa_io_base' undeclared here (not in a function)
ppc_ksyms.c:72: initializer element is not constant
ppc_ksyms.c:72: (near initialization for `__ksymtab_isa_io_base.value')
ppc_ksyms.c:74: `pci_dram_offset' undeclared here (not in a function)
ppc_ksyms.c:74: initializer element is not constant
ppc_ksyms.c:74: (near initialization for
`__ksymtab_pci_dram_offset.value')


When I comment out these two lines, the compiler errors continue to get
worse.

Can anyone recommend any changes I should make to fix these errors?
Additionally, could you give me a rundown on any issues with the
MPC8xx kernel supporting modules?

Previously, we ported our 2Mb wireless lan driver to the RPX-Lite by hard
coding all the pcmcia "stuff" in the driver itself.  For this port, we'd
like to use pcmcia services so the embedded driver and our "non-embedded"
driver use the same code (easier to maintain source!).

Many, many, many thanks!
Jo-Ellen Mathews

---------------------------------------------------------
Jo-Ellen F. Mathews

AbsoluteValue Software     Web:    http://www.absoval.com
P.O. Box 941149            e-mail: joellen@absoval.com
Maitland, FL 32794-1149    Phone:  407.644.8582
USA                        Fax:    407.539.1294


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

Re: Linux Planet Module Support

[Joe Green]

On Thu, 16 Mar 2000, Jo-Ellen F. Mathews wrote:
> I've been working on porting our 11Mb wireless LAN device driver to
> Embedded Planet's nifty RPX-Lite (MPC823) board.  I've been using Embedded
> Planet's wonderful, new Linux Planet development kit which includes
> MontaVista's Hard Had Linux 1.0.  I'm loving this kit!
>
> I'm using the cross compiler provided on MontaVista's cd, and I've been
> very successful at compiling the default kernel and a few, simple
> programs.
>
> The most direct approach (okay, and quickest/easiest) is to keep the
> wireless lan driver as a loadable module dependent on pcmcia card
> services.  Before attempting to use Magnus Damm's pcmcia patches and
> kernel modifications to support pcmcia, I first need to recompile my
> kernel for module support.
>
> With CONFIG_MODULES enabled (regardless of what CONFIG_MODVERSIONS or
> CONFIG_KMOD are set to), I get the following errors:

If you can wait, we'll be fixing this in an upcoming kernel package.
It's mostly a matter of adding additional configuration information
(e.g., some symbols are only present if CONFIG_PCI).

--
Joe Green <jgreen@mvista.com>
MontaVista Software, Inc.


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

Re: Linux Planet Module Support

[Pavel Roskin]

Hello!

> When I comment out these two lines, the compiler errors continue to get
> worse.

Anyway, go ahead and remove all the symbols that cannot be exported.
The kernel will eventually compile and work.

IMHO, ppc_ksyms.c should become a generated file some day. What's the
point in exporting functions that the modules don't use anyway?

Pavel Roskin

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