(added Christos to the cc list)

On Mon, May 10, 2004 at 01:32:52PM -0700, Matt Thomas wrote:

> There is a serious bug here if user-code can induce cache errors.  That
> shouldn't be possible.

Agreed, but it's happening.

I've narrowed the problem down a bit: csh invokes something in libc that will
panic a machine with a r4k-style cache (Manuel's response indicates that r5k
will just dump core).

-current as of 20040421 (and 2.0, all dates) have this problem; -current after
20040422 does not.  That corresponds to Christos' fgetstr() implementation
(not likely to have fixed the problem) and more statfs()->statvfs() changes.

I'll backport the 20040422 changes to 2.0 for testing.  If it fixes the problem,
would it be appropriate to request a pullup (which would address the symptom --
the true cause is probably deep in arch/mips/, and I'm not sure I'll have enough
time to figure it out before 2.0 ships)?  These changes resulted in a shlib
minor bump ...

-- Chris
	GPG key FEB9DE7F (91AF 4534 4529 4BCC 31A5  938E 023E EEFB FEB9 DE7F)