mvme kernel patches [LONG]

Gabriel Paubert paubert at
Tue Apr 20 20:54:05 EST 1999

On Tue, 20 Apr 1999, David De Ridder wrote:

> Hi,
> here are some things I discovered whilst testing Gabriel Paubert's
> kernel patches ( on our MVME2306
> board :
> * using the old 2.2.4 patches (pre-16 April), ``make menuconfig'' would
>   only allow you to load or save an alternate config file - no other
>   kernel options were accessible.

Well, I only use make xconfig and make oldconfig (because xconfig has 
a tendency to screw up some things). Occasionally CONFIG_SERIAL_CONSOLE
vanishes, and I don't know why. 

> * Using the mvme2600.config.nfsroot with these patches never resulted
>   in a kernel that would boot (``Residual data located'' message, even
>   before the Linux kernel had said anything).

That's not normal. Once you have copied the to
.config, you should always do a make oldconfig to setup things properly.

> * With the new patches, ``make menuconfig'' works fine (only one error
>   during parsing). But now, the options to save or load are *unavailable*.
>   This is not such a problem, the old ``cp'' and ``mv'' commands can do
>   the trick ;-)


> * The new precompiled nfsroot-2.2.4 kernel works great (as did the
>   2.2.1). Even better : it does not give the ``openpic.c'' error anymore.

I'm up to 2.2.6, it will appear soon on the server. Lots of VME tests this

> I still have some questions :
> * Why is the Generic DECchip driver used instead of the DEC2x41x driver ?
>   Is this because of the problems with the DEC21143 ?

Arenn't you confused ? I think they are the sme driver under evolving
names. The source file is de4x5.c, tulip.c has never worked properly on
big endian machines. There are still problems with my source on 21143, but 
Matt Porter just posted agin the necessary patches. 

> * What is the use of this ``BC336'' ?

This is an example driver that uses the services exported by universe.c
to handle VME interrupts and access a board by only knowing that it is an 
A16 D08(O) dumb slave. 

This board performs actually precise synchronization between
different systems (with one board in master mode and several in slave,
you can build a dedicated time distribution network). It can also record
the time of synchronization pulses generated by other equipment and
generate a heartbeat at any frequency up to 2kHz. It has some other bells
and whistles but I don't record all them right now (like automatically
measuring the error rate of its internal timebase when it is connected to 
a reference and switching to flywheeling whenever the master source
is unavailable).

However, this is not and ad for BanComm, I had to write this because it
happens to be what we use at the telescope so I put this as an example.

BTW: I'm going to try to finish the doc of the universe driver this week
and then announce it officially. I no more touch the code...

> * (Gabriel) Is it possible to post the .config that resulted in the
>   current zImage-nfsroot-2.2.4 on your FTP site ? Or by mail to this
>   list ?

Yes, I thought they were uptodate, so I must have screwed up something.

> I'm currently trying to compile my own 2.2.4 with the latest patches.
> I'll keep everyone posted.

Ok, thanks for the report.


[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check ]]
[[ and for useful information before posting.   ]]

More information about the Linuxppc-dev mailing list