Console output after GPU handover?

[teh_orph]

Hi there,

long time lurker, first time poster.

I have a lot of embedded and ARMv5/ARMv7 experience and I'm tempted once I get my Pi to bring up an OS on it (that's not Linux).

However my #1 question would be: is there any kind of (character-based) display that is brought up by the GPU before the loaded OS is booted? eg some kind of MMIO region where you can echo characters to? I can't seem to find this info yet on these forums.

If I'm to attempt this I ideally need some kind of debug output. I see there are LEDs - are these programmable? If there's no display and no programmable LED then I guess that leaves the GPIO to work with - which is a little bit beyond my electronics ability...

Cheers!

[teh_orph]

Surely someone has turned one of these things on before with a screen plugged in...

[Gert van Loo]

I am not a sw expert but I would think any Linux bring up to use the UART for the kprintf output. Because you still have to debug the ARM software which interacts with whatever the GPU has prepared. That UART is definitely available from the start.

After you are sure that you can output your code trough the GPU area you can put the text on your screen (But I would not like that because on a screen I can only see the last number of lines. In a UART log I can go back hundreds of lines)

[XAPBob]

Isn't that what a video recorder is for

[tufty]

As far as I"m aware, no, there is no early boot console provided by the GPU. As has been mentioned, you do have access to a pl011 UART, which is probably the easiest way to get output early on (if nothing else, you probably have tested driver code for it)

GPU interaction and fb startup is currently only documented in the linux source.

Simon

[teh_orph]

Cheers guys.

Yeah I figure that framebuffer stuff would be a no-go at an early stage, that's why I was hoping for some kind of already-on character display!

Looking through the Rpi wiki page which discusses additional hardware it mentions that console output comes out via the UART by default at 115200 bps (ttyAMA0)...so at least someone's done that before! Looking through the Rpi kernel check-in there's code for programmable LEDs too, but it's not clear if that's just Linux support for LEDs attached to the GPIO.

I'm not savvy with the electronics side - would it be easy to hook up the eight GPIO pins to pick up this console output and turn it into something I can plug into a serial port? (to read on PC via a terminal program)

[tufty]

You only need 3 of the gpio pins - rx, tx and gnd, hooked up to a (I believe) 3.3v serial port - either a 3v3 ttl usb/serial adaptor or a 3v3->RS232 level converter, depending on your PC setup.

Simon

[teh_orph]

Sweet, thanks for the info!

Now that I know what I'm looking for I see there's existing threads about this UART/serial stuff. Thanks!

[TonyD]

There's this thread for a small RS232 breakout PCB (sorry about blowing my own trumpet)

[zippy]

This will be simple to implement in ARM code once we know the hardware addresses for setting up the "dumb frame buffer", or even the addresses of the basic setup routines in the GPU "blob" if we can't access the hardware directly. Hopefully this info will become available soon after the launch of the boards, one way or another.