thorpej@shagadelic.org wrote:

> > I added some debug printf()s on C table shortage, and then
> > more than a half of tables look still unused on the panic.
> > So it looks the TAILQ structure corruption.
> > Actually, the panic sometime happened during bootstrap,
> > right after ntpd was started. (mlockall() might trigger it?)
> >
> > I wonder if we have to protect these TAILQ ops against A, B and C
> > tables by splvm()/splx() pairs... Comments?
> 
> If those tables can be allocated as a side-effect of calling malloc 
> (9), then yes.

These TAILQs are modified in pmap_enter(), pmap_unwire() and
pmap_remove(). In this case, should these functions be protected
by splvm()?
(3/80 is too slow to confirm how patches work... :-)
---
Izumi Tsutsui