[bug-libsigsegv] Re: sigsegv on s390 only giving start address of page in segv handler

[Heiko Carstens]

On Mon, Mar 14, 2011 at 11:50:37AM +0100, Bruno Haible wrote:
> Heiko Carstens wrote:
> > You would have to perform instruction decoding to figure out what was the
> > exact failing address. Not really a thankful job.
> 
> This is done in libsigsegv and Boehm GC for m68k, PowerPC, and Alpha processors
> (see [1][2][3][4]) already. Why not also for s390 and s390x?

You can certainly do that. Even if some parts might be a bit ugly (thinking of
e.g. the EXECUTE instruction).
But most of that should have been done in the s390 Valgrind port anyway (cc'ing
Christian).

> Or even better, why doesn't the kernel do this job before passing the address
> to userland?

Problem here is that old kernels on new hardware won't be able to decode new
instruction formats. What is the kernel supposed to do if it should decode an
unknown instriction format?
That doesn't work.