> The problem is the same as the original post: something bad happens, swap
> gets used or over-used, and the machine locks.
AFAIK this is not a common problem. There's a known problem in ZFS that
exhibits this behavior, and IIRC there could be similar problems in the
past if you tried to swap over NFS (not sure if that problem was fixed
since; it was many years ago), but I don't know of such problems for the
"normal setup".
> Without even a warning message. BSD-derived OS's running on the very
> same commodity Intel hardware dont have that problem. Why does linux?
AFAIK it's usually just a bug, tho sometimes it's one that's not easy to
fix because it's not result of bad code or bad design but of
incompatible designs in two different parts, so fixing it implies
a redesign of at least one of the sides. I can't remember what was the
underlying issue in the swap-over-NFS case but it was non-trivial to fix
satisfactorily, IIRC (and I seem to remember that the problem was
fundamental enough that it likely affects all possible implementations,
so all a "fix" could hope to do was to reduce the incidence rate and
improve the debug-log to help diagnose the problem). In the ZFS case
the problem is likely that the ZFS code was not designed with the right
assumptions for use in the Linux kernel.
Stefan
