From: Jeff Squyres (jsquyres_at_lam-mpi.org)
Date: Mon Jun 03 2002 - 13:03:06 PDT
On Mon, 3 Jun 2002, Jason Duell wrote:
> > It looks like the signal handler context is going to be very limiting.
>
> If the signal handler context is limiting just because 1) malloc is not
> signal safe and 2) pthread locks around data structures aren't safe
> either, it's worth noting that you can make both of these safe by using
> 'signal safe locks' (a mutex that you guarantee is never held when a
> signal handler is run), plus a separate malloc library for lam's
> internal use.
Good points.
I think there's more than just malloc/free, though -- we'd have to audit
the code to know exactly what calls we're going to have to be worried
about (I think you guys started to look through the code last week...?).
I think the preliminary list included the following:
malloc, free, accept, connect, gm_*
(I realize that we're not talking about the GM library yet, but I'm just
trying to take the long view)
Other ones that come to my mind offhand (I don't know if these are signal
safe or not):
shm*, ioctl, fcntl
{+} Jeff Squyres
{+} [email protected]
{+} http://www.lam-mpi.org/