On Thu, Oct 20, 2005 at 14:41:25 +0000, Christos Zoulas wrote:

> I would suspect an i-cache issue.  Can you reproduce the problem if
> you flush the cache on each syscall?

I'll try that, but the bug is very hard to reproduce reliably w.r.t
different kernel configs.  E.g. I can reliably reproduce it with this
kernel with gmake config.status.  If I e.g. drop few options (INET6,
some others) from the kernel, the bug is no longer reproducible.
Ktracing or running under gdb make it go away too.  So if flushing the
i-cache makes it go away, it doesn't prove much.

SY, Uwe
-- 
uwe@ptc.spbu.ru                         |       Zu Grunde kommen
http://www.ptc.spbu.ru/~uwe/            |       Ist zu Grunde gehen