parisc,

After some banter on IRC about how one should set the stride
for a cache flush using the 'fic' instruction, I noted the 
following possible problems our trap handler:

a. Non-access instruction TLB miss fault (16)
	- No implemented case statement in traps.c
	- No problem, falls through to do_page_fault.
	- Suggest forcing a sigsegv on the process.

b. Non-access data TLB miss fault (17)
	- Calls parisc_terminate and gutters the cpu.
	- This should not happen.
	- Suggest forcing a sigsegv on the process.

c. Conditional Trap (13)
	- No implemented case statment in traps.c
	- Suggest forcing SIGFPE and setting si_code to zero,
	  and allowing the userspace app the decode the 
	  situation from the signal andler (HPUX seems to 
	  this?) via si_addr.

c. All paths must either return or set the fault_address
   and fault_space values? One place does not, I put a 
   FIXME there. Needs review. Those two values are needed
   by the check after the switch statement in handle_interruption.
   Is setting them to zero, as is done now, good enough?

Patch attached :)

c.