Re: HEADSUP: Sleep queues added to kernel, so be careful.

On Sat, 28 Feb 2004, Vincent Poy wrote:

> > > >   Switch the sleep/wakeup and condition variable implementations to use the
> > > >   sleep queue interface:
> > >
> > > There should be little user-visible effect from this change.
> >
> > Except for the kernel breakage, perhaps ;-) I guess John is away until
> > Monday though.
> 
> 	What kind of kernel breakage?

I think Kris is referring to the tinderbox messages -- the midway driver
appears to be having problems compiling because it relied on the
cv_waitq_empty() primitive, which is no longer present following the sleep
queue change.

> What seems to happen for me is that I was originally on a September 23,
> 2003 -CURRENT and then applied all /etc updates and then according to
> src/UPDATING - I did a make buildworld, followed by a make buildkernel,
> make installkernel and rebooted and things were still working.  Then I
> did a make installworld and then recvsupped with the latest tree which
> had src/sbin/savecore/savecore.c 1.68 as the latest thing, then repeated
> the make buildworld, make buildkernel, make installkernel, make
> installworld but this time as soon as I rebooted, the network seems to
> not work at all.  I can't get past the machine.  Is this something
> broken with ipfw since I can't ping the public IP but I can ping the
> local 192.168.0.1 address but not anything beyond that and is connected
> at 100Mbps Full Duplex.  Tried switching both NIC cards and cables just
> in case they decided to fail but no go either. 

Could you confirm that your userspace and userland are really 100% in
sync?  If you run without IPFW, do things work properly?

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert_at_fledge.watson.org      Senior Research Scientist, McAfee Research