Vincent <10.50@free.fr> writes:

> Bill Studenmund a écrit :
> 
> > All I can think is to add printf()s after that message and keep
> > trying. You should be able to find out which call is crashing (since
> > all the subsequent printf()s won't print), then figure out what is
> > being passed in.
> 
> Good theoretical idea, bad practical one. I lose at least 3 minutes
> each time in FSCK'ing the disks... But if there is no other way...

You shouldn't need that.  If the kernel crashes before mounting / rw,
there is no need to fsck.

Similary, if a userspace program triggers the crash, and you know what
program it is, you could try running that program from single-user.
You might even use mfs if such a program really needs to write
somewhere.

        Steinar