Skip to contentSkip to conversation

a MINIMAL405 submodel

DISCUSSION David Updegraff 4 participants 5 messages Jan 15, 2002 – Jan 18, 2002

lore
Share & tools
lore

David Updegraff

author
#0 lore raw

Greetings.

I have made a trivial little patch (and a few files..) to define a "minimally configured" 405[gp] board. Its basically a 405 with no KB, RTC, PCI, et.al. The idea being that if you define this submodel of CONFIG_4xx then you write loadable modules on your own to handle peripherals; this being just enough in the kernel to get you to net- or flash-boot.

Its like a "WALNUT" but without any peripheral bank setup, RTC or KB assumptions.. you get the idea.

In our case we bolt on lots of silly junk to the expansion banks and do DMA... but all inside of custom little drivers that set those banks up. Might we be able to limit the proliferation of submodel in the PPC 4xx world with the addition of my proposed MINIMAL405 submodel and then let vendors write drivers instead of kernels?

For any who care, have a peek at
http://www.toimi.com/linuxppc
for the .patch file and a .tar file of needed additional platform files.

Cheers.

--
Dave Updegraff / [email protected] / 218-525-1154

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

#1 lore raw
Quoted 3 lines · David Updegraff

... Might we be able to limit the proliferation of submodel in the PPC 4xx world with the addition of my proposed MINIMAL405 submodel and then let vendors write drivers instead of kernels?

The 8xx and 8260 are almost like this. What you will discover over time is a driver you write is 99% of what someone else wants. In your proposal, you would end up with a whole new driver that is almost identical to another, which isn't practical to maintain. Just follow the convention of adding a board descriptor for your platform. If it is truly identical to another, then the platforms can share the same files and this knowledge is implicit in the configuration. Now, if the current configurations make assumptions about platforms that use a 405GP (and it shouldn't), that is what needs to change.

Thanks.

	-- Dan

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

author
#2 lore raw
Quoted 13 lines · Dan Malek
Quoted 3 lines · David Updegraff

... Might we be able to limit the proliferation of submodel in the PPC 4xx world with the addition of my proposed MINIMAL405 submodel and then let vendors write drivers instead of kernels?

The 8xx and 8260 are almost like this. What you will discover over time is a driver you write is 99% of what someone else wants. In your proposal, you would end up with a whole new driver that is almost identical to another, which isn't practical to maintain. Just follow the convention

Quoted 5 lines

of adding a board descriptor for your platform. If it is truly identical to another, then the platforms can share the same files and this knowledge is implicit in the configuration. Now, if the current configurations make assumptions about platforms that use a 405GP (and it shouldn't), that is what needs to change.

I'm happy with that approach too: so we'd have a new board description of

a MINIMAL405. Or.. with some minor surgery to the WALNUT description, I could fit in there (additional CONFIG_* things maybe to adjust presence of peripherals and presumption of PPCBOOT vs. IBM-BIOS..). It boils down to almost same thing: a board defn. that can be configured down to a peripheral-less 405gp. I has assumed that my chances of lobbying for a "new" board description might be a bit better than lobbying for a bunch of changes to the WALNUT or EP405. Though either of those would work fine if they just consistently obeyed CONFIG_PCI, CONFIG_RTC, CONFIG_NVRAM, CONFIG_VT, and we had a mechaninsm for ppcboot cognizance of board_t. Infact, it looks to me like they could be merged in that case..

Please advise.

--
Dave Updegraff / [email protected] / 218-525-1154

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

#3 lore raw

Hi David,

what do you mean by a peripheral-less 405gp ? It is possible to deselect all OCP functions from the kernel config. You do not need a new plattform to do this.

I am using the current 2.4.x devel tree with the CPCI405 config on a custom 405CR board. The 405CR does not have PCI or Ethernet. So its nearly a peripheral-less GP. I only had to deselect PCI and Ethernet.

So why a new plattform ?

Matthias

--
-------------------------------------------------------------------------

                            _/_/_/_/   Matthias Fuchs
                           _/_/_/_/   Dipl.-Ing.
                          _/_/_/_/   [email protected]

      _/_/_/   _/_/_/_/_/_/_/      esd electronic system design gmbh
    _/   _/  _/             _/    Vahrenwalder Str. 207
   _/   _/    _/_/_/   _/   _/   D-30165 Hannover
   _/             _/  _/   _/   Phone: +49-511-37298-0
    _/_/_/_/_/_/_/   _/_/_/    Fax:   +49-511-37298-68

-------------------------------------------------------------------------

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

#4 lore raw

building gdbserver 5.0

In reply to David Updegraff

I've searched for info on building gdbserver from gdb5.0 w/o success -- any pointers?

I cd .../gdb-5.0/gdb/gdbserver,
./configure --target=powerpc-linux --host=i586-pc-linux-gnu make CC=powerpc-linux-gcc

3 objects are created:
remote-utils.o server.o utils.o

and the following are undefined @ link:
mywait, kill_inferior, mythread_alive, registers and a few others.

I suspect that some other source should be compiled in but am not seeing how to configure so that it's mentioned in the Makefile.

suggestions appreciated,
Mark

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