>On the other hand, Sun's xnews server used to mess up things thoroughly, in
>my case by fouling the le0 driver causing an endless stream of
> "RINT: no buffer" messages.

This is a bug in the vm/pmap code.  I sent Theo a fix ages ago.
(Basically, the problem is that the pmap code turns off the `no
cache' bit on the Lance memory after xnews double-maps it.  Yes,
the Sun XNews server peeks into kernel data structures.  Why?  To
build the magic cookie, I think.  Yech.  But it is a generic
bug in /dev/mem / pmap.c and needs fixing anyway.  The fix is
a gross kludge, but works: never turn off PG_NC on any virtual
address in ``dvma'' space.  This address space should be removed,
but for now...)

Anyway, once you fix that, you will find that xnews demands a
/dev/win0 driver (!).  I am not sure how much it wants from it.

Chris

------------------------------------------------------------------------------