Comment # 16 on bug 1200919 from Matthias Brugger

(In reply to Hillwood Yang from comment #11)
> (In reply to Takashi Iwai from comment #10)
> > Try to trigger the crash via "echo c > /proc/sysrq-trigger", and confirm
> > that kdump is working (you get the crash dump).  If it doesn't work, it
> > means that kdump setup is likely insufficient.  Try to increase the reserve
> > memory size, for example.
> 
> Surprise! I confirm I enable kdump in Yast2, and set crashkernel as 1024M.
> But kdump still does not work, /var/crash is allways empty.

I think you have to define an offset for the crashkernel to make that work.
Chester do you remember?