> > I just found a bug in the floppy driver. Andy disabled the support for
> > the second floppy driver. Unfortunately there is a bug that then
> > results in random writes into the memory.
> Hmmm. Is this a bug in the standard driver, or is this my fault?
It is a bug in the standard driver that was activated by your changes.
Just look into drivers/block/floppy.c, into function set_fdc(). This
functions calls set_dor() twice. Once for controller fdc and once
for controller 1-fdc. The later call results in erroneous memory
accesses when N_FDC (in <asm/floppy.h>) has been defined to 1, not two.
Just put #if N_FDC > 1 ... #endif around the second call and everything
Finding this bug was a bit of luck - I wondered about certain "unplaned"
memory accesses when debugging something else. The consequences however
are fatal - every process that is just running when set_fdc is being
called gets "tuned" ...