On Oct 18,  8:49pm, thorpej@shagadelic.org (Jason Thorpe) wrote:
-- Subject: Re: delivering faulted-upon address in trap frame to userland

| 
| On Oct 18, 2006, at 3:54 PM, Christos Zoulas wrote:
| 
| > Yes, such as referencing an invalid/non-existant register, but
| > usually something related to the internal functioning of the
| > cpu/co-processor and un-related to the memory/io subsystem.
| > SIGSEGV/SIGBUS are the signals used to indicate a problem outside
| > the cpu subsystem.
| 
| But I thought the issue here is that, with certain types of GPFs,  
| there isn't enough info to send a proper SIGSEGV or SIGBUS, hence the  
| need to send something else (like a SIGILL, which doesn't require as  
| much extra data).

The issue is that we don't know the faulting address. I think it is better
to return a magic cookie, than change the signal type.

christos