>Well, i thought we were talking mostly about what's in user-land core
>dumps, not in kernel core dumps...

Oh :). 

Still, We really havent gotten very far on deciding how to *use*
nonvolatile ram, since we dont (yet) have drivers to handle it.

I could imagine mmap'ing space in Presto ram, in order to to
shift post-crash cleanup out of the kernel and into a user-space
daemon, with an ioctl to say that `things are clean'.

If we did that (which may be a dumb idea), then we'd want to dump
mmap'ed nvRAM as part of a process coredumps.