[mpeg2] Re: RESOLVED: Hard lockups when loading the kfir module

[Andreas Trottmann <andreas.trottmann@werft22.com>]

On Wed, Apr 25, 2001 at 11:47:46PM +0200, Frank Rosengart wrote:

> > There are the following messages from the driver:
> > 
> > After loading it (with "debug=1" as the only parameter):
> > 
> > kfir: Kfir board revision 5e
> > Unable to load 'micro/altera_keb_plx_5e.rbf'.
> > loaded microcode with length 32385
> 
> !! this is wrong
> 
> > kfir: done
> > kfir: Reset Kfir and PLL ... done
> > Unable to load 'micro/MPEG1_6.HLD'.
> 
> this too.


Hm, I feel I need to slap myself with a baseball bat or so...

First of all:

HUGE THANKS TO EVERYBODY! WE CAN NOW SUCCESSFULLY ENCODE MPEG-2 VIDEO!


Now, here's what we've done:

Again, the machine where we compiled the driver is a different machine
than where we installed the board (bascially, this was because i compile
all kernels on one machine, and thus don't have any compilers installed
on the production ones... and the compiling machine does not have any
free PCI slots, so we didn't install the kfir there).

Now, this lead to the first problem, that kfor.o crashed the machine
when it couldn't find the firmware. At this point, I read the code and
saw


#ifdef __KFIR_PACK__
                    !(microlen=firmread("micro/altera_keb_plx_5e.rbf",
                                        (char**) &micro)) &&
#endif
                    !(microlen=firmread(KFIR_MICRO_PATH "altera_keb_plx_5e.rbf", 
                                        (char**) &micro)))

which let me beleive that the code first tries to load the microcode in 
micro/, and then in KFIR_MICRO_PATH, which is defined as /lib/kfir. Because
I didn't want to have a relative path, I put the microcode in /lib/kfir on 
the production machine.

This turned out to be wrong... if the microcode is in /lib/kfir, I get
the behaviour described earlier -- however, if the microcode is in
micro/ (relative to the cwd of the insmod process), it simply works.

I first couldn't really believe this, but apparently it really is 
the case.


So, thanks A LOT for your help!

-- 
Andreas Trottmann <andreas.trottmann@werft22.com>