Re: Process traceing

[Mark Kettenis <kettenis@wins.uva.nl>]

   From: lars brinkhoff <lars@nocrew.org>
   Date: 18 Feb 2000 08:50:30 +0100

   Mark Kettenis <kettenis@wins.uva.nl> writes:
   >    From: Mariusz Woloszyn <emsi@it.pl>
   >    How can I determine the context of cr2 special register when the proces
   >    caught SIGSEGV? The sigcontext structure defined in hurd in
   >    /include/bits/sigcontext.h file does not contain the cr2 register.
   > The Hurd uses the same signal handler conventions as BSD, which means
   > that you can declare you signal handler as:
   >    void handler (int sig, int code, struct sigcontext *scp)
   > For SIGSEGV, `code' is the faulting address.

   Also, how do you find out if it was a read or a write (or even exec)
   access?

I don't know if there is an easy way.  The `sc_error' member of the
sigcontext is set to the error code that is generated by the Mach
kernel in association with the fault.

The KERN_WRITE_PROTECTION_FAILURE error code looks promising.  The
comment in <mach/kern_return.h> suggests that you can set
VM_PROT_NOTIFY for the pages you're interested in such that this error
code will be returned when you write to those pages.  But I'm not sure
this really works.

Mark