Hello all,

The following patch to cache.c rearranges the logic in cpu_cache_init() 
somewhat to make the detection of the processor via the 
current_cpu_data.cputype take precedence over the cpu_has_4ktlb test.  
This seems to make sense, as the cputype test is significantly more 
specific than the 4ktlb test.  It also arranges for the panic() to 
happen if the 4ktlb test fails as well (the old version would fail 
silently).

It also happens to fix things so that the SB1/swarm starts booting 
again....  ;)

Please consider this for inclusion.

Thanks,
Mark