What is i386 thread.trapno?

[Jeff Dike]

With a normal segfault, the handler gets a trapno == 14 in the sigcontext.  
With UML, I can make a process infinitely segfault with trapno == 1.  The page 
being accessed is correctly mapped in according to /proc/<pid>/maps, so the 
odd trapno is the only clue that I can see that something is different.

The i386 page fault handler sets trap_no to 14, so the fault isn't coming from 
there, but I can't see where a SIGSEGV is being delivered to a process with 
thread.trap_no == 1.

So:
	What do these trap numbers mean?
	Where can I read about them?
and
	Where's this segfault coming from?

				Jeff