PDC console problem.

Hi,

I am trying to boot PA-Linux off a PA-RISC simulator. I am facing a problem with the PDC console. On my first attempt, I saw the message asking me to switch the console. So I turned the PDC console config option on ( the serial mux option was turned off). I added the parameter 'console=ttyB0' in 'palo.conf'. I created the device file '/dev/ttyB0' with major/minor = 11/0 on my ramdisk. With this, the kernel boots up on the simulator all the way upto the sash ( Stand-alone shell) prompt. I can even see the prompt. But, the problem is that this console doesnt seem to be accepting input from the keyboard. It is kind of like in a locked state.

I was going through some of the FAQ and the mailing list archive and understand that the PDC console is not a full blown console and can only display output. Is is true that the PDC console cannot accept keyboard input? Or am I missing something?

I also see that there are 2 versions of the PDC console driver :
'arch/parisc/kernel/pdc_cons.c' and 'drivers/char/pdc_console.c'. Both use the device ttyB. From what I understand, the former is used before any of the serial drivers are initialized during boot, and has only write capability. The latter is the full-blown console driver (though I am not sure if it has write capability). Is my understanding correct?

I am using the source tarball 'palo-latest.tar.gz' from the parisc-linux cvs respository and cross-compiling this on an x86 box.

Regards,
Naresh.

PS: My "init" process is 'sash' and I also have a '/dev/console' device ( major=5 minor=1) on my ramdisk. The Serial driver config option is on and I can see it being initialized in the boot path after the PDC console.

Hmm.. The pdc_console has both read and write capabilities. We used the PDC console for K/E class systems before the SerialMux driver was written. I am not familiar with the PDC simulator, but I would be willing to test the pdc console again on K-Class.

Which kernel are you using?

Thanks,

- Ryan

Ryan Bradetich <[email protected]>

A small correction to the original mail: the kernel source that was compiled was the one thats listed as latest ( cvs/linux_latest.tar.gz), which I guess is 2.4.20-pa22. I have booted this very same kernel off an L-Class machine with the PDC console option off.

Regards
Naresh.

Hello !

The PDC driver supports both directions, IN and OUT. The PDC console driver was ported from the full serial driver interface, some services not needed for poor console support are dummy routines.

So the serial console cannot be used for MODEM and other full-blown serial devices, but is a fully functional console device including all UNIX like features (login is working like on all other serial devices).

He Ryan, by the way:
Why is my driver (PDC console) NOT in the mainstream kernel 2.5 ? Even your driver covers a couple of machines, I do not see in the MUX driver the replacement !
I agree, that the MUX driver should be used on machines with that MUX interface. But the idea of my PDC driver was to support a fallback driver using the ROM console usable on all machines, independent of special setups.

I many architectures such a fallback driver is present. I remeber the case, there was an failure in the HW-detection on my SUN Sparc-2, and the kernel used a ROM console driver instead of the frame buffer driver for screen/keyboard. This driver was awfully slow, and I do not recommand it, but such a fallback driver should part of a kernel delivery !!

So we should think about adding this driver also, independent of al the other drivers included for console (and although the other drivers should alwaysb be preferred).

Christoph

-------------------------------------------------------
private:	[email protected]
company:	[email protected]

Hello again!

There are two drivers, but they depend.

Int arch/parisc/kernel driver implements the kernel "printk()" (the internal kernel console). This driver has NO device name, as it cannot be access as /dev/ttyB0. BUT: This driver is register under the name "ttyB0" to be selected as preferred device on the kernel paramter line with "console=ttyB0".

The drivers/char driver is the full PDC serial console driver as a TTY driver (like all the other serial drivers). This driver is registers as MAJOR 11, MINOR 0 driver in the TTY interface and supports fully TTY functionality (althogh some routines are dummy, e.g. you cannot change the BAUD RATE, etc...).
The low lever routione (printing out a character, polling the keyboard) are implemented in the first mentioned driver..

The third driver play a role is the "/dev/console" tty driver, which assosiates the two drivers above as console driver. So the kernel paramter line also assigns the "/dev/ttyB0" as the driver behind the "/dev/console". And this driver is running bidirectional.

So commig to your simulator. I do not know this, but perhaps you have simple a problem with the timing here. My PDC console drive is in opposite to other typical TTY drivers running in POLLING mode. The PDC interface does not support a interrupt-like handling. So the PDC driver polls all 30ms (as far as I remember) if a key was pressed on the PDC console. My be, the time progress has other properties on the simulator ....

Christoph

-------------------------------------------------------
private:	[email protected]
company:	[email protected]