On Sun, 2009-05-31 at 17:31 +0100, Alan Cox wrote:
> The only source of signals during a dump should be external ones. Far
> better would be to set some kind of defined signal mask during the dump
> (say SIGPIPE, SIGINT, SIGQUIT) ? I agree with Paul's patch in the sense
> we don't want spurious SIGIO events or similar spoiling a dump.

Something similar to this is what Andi Kleen first proposed.  The issue
is that it modifies the signal mask for the process before the core was
dumped, so that the mask saved in the core was not the real mask used by
the process when it took the exception.

Andi mentioned this problem and suggested we'd need to keep the original
mask around to be saved in the core.  That would involve deeper hacking
in the ELF format but maybe this is the right answer.