I had forgotten to add the "THE DRIVER" to the subject, so you may
oversee
the driver's code, sent as attachment in my pre-last mail.

To be sure, here the attachments again !


---- original mail ----

Hallo Thomas,
hallo Grant,

finally the CVS was working again.

I have no login for your CVS so I will send the `cvs diff -u' and the
newly
added file `drivers/char/pdc_drv.c'.

Some notes here:

* In my console drivers (serial and PDC) I always added code, that the
machine
can be resetted via `^\' (Ctrl-\) for development (especially for
machines 
without reset button as the E55 server). I removed this code for sending
the
CVS diffs.

* There is no CONFIG_ stuff for the PDC console, the console is
initialised
        in `arch/kernel/pdc_cons.c' (EARLY_BOOTUP_DEBUG) and the tty
driver
        has a __init constructor routine.

  ATTENTION: It is very important to have the init code in the
pdc_cons.c
        executed before the ttyBn driver is used, otherwise the first
access
        to the PDC driver (echo hallo > /dev/ttyB0) leads to a blocking
        machine ....

  The driver can also be used on machines running true serial console,
        by `echo hallo > /dev/ttyB0'. But the mix on the serial console
        of ttyS0 driver and ttyB0 driver is not recommended !!

* The driver as console is only tested on the server E55 (model
9000/856)
        and the HP apollo 700 series (9000/720/50), in both cases the
        machine was configured to use the serial line.
        Further I tested the kernel on a 712/60 machine, having PDC
console
        on STI device. Here the PDC console is used correctly, but then
        the console switches to the colored frame buffer (and the text
is
        completly BLUE (!) and is overwriting the poor TUX !
        I think here is a bug in the console handling not staying on the
        configured device (has nothing to do with my stuff) or it is a
        feature ti have full STI support if it possible ... 
        But /dev/ttyB0 is never used together with an STI able machine
.. :-)

* I added the `CP:' to my modules or `XXX CP:' if there is something to
do ...

* As I cannot checkin, I have no way to give change logs, so I will do
here
a short list:

- pdc_drv.c: Added file, PDC console tty driver for linux, /dev/ttyBn, 
        n=0 (only)

- pdc_cons.c: Added keyboard polling routine, added an function entry to
get
        an device entry for the console driver to "connet" the pdc_cons 
        driver to the PDC tty driver

- pdc.h: Added prototypes used for console and LEDs

- firmware.c: Added reading access to the PDC console (was former used
for 
        waiting on a keypress), readded the pdc_chassis_disp () (was
removed 
        in the meantime, but at the current state I use it for the E55
LEDs).

- led.c: Added a "trial" version of a LED controlling. Only RUN and
ATTENTION
        LED is supported. RUN/ATTENTION stated derived from state of the 
        HEART-BEAT LED (flashing->RUN, off->ATTENTION), and so the LEDs
are 
        also accessable via PROC-FS. The LED activation of my routines
is
        derived from the model name (hard coded 9000/856) for my tests.
        How to deal with this in the future ?

- major.h: Added free major 30 (character device) for PDC console
(ttyBn).

- drivers/char/Makefile: pdc_drv.o added (per hack) to arch PARISC
dependent code.

The code builds warning-free.

Again to test:

1. Compile and have a look that the pdc_drv.o is included in
drivers/char/Makefile
2. dev entry /dev/ttyB0 c 30 0
3. /etc/inittab, have a login at ttyB0
4. Added the `ttyB0' device to /etc/securetty, if you want login as
`root' on the
        console
5. kernel line parameter with `console=ttyB0'

(ad 1): In the future (if somebody can help me) we want to have a config
item in
the kernel configuration for the PDC console. 


I attch two files:
        - linux-cvs.diff-20011119.log
                Diffs displayed by `cvs -z9 diff -u .'
        - pdc_drv.c
                The new file (simple displayed as `?
drivers/char/pdc_drv.c')
                The file has to be checked in at `drivers/char'

I hope you can use it and will add it to the mainstream kernel ...

Bye
Christoph P.






-- 
-------------------------------------------------------
private:	christoph.plattner@gmx.at
company:	christoph.plattner@alcatel.at