weird swi4:clock misbehavior after fire saver goes in

[Ricardo Nabinger Sanchez]

Hello,

After upgrading a somewhat low-end machine to -CURRENT, I noticed a
misbehavior of a system thread, "swi4: clock".

What happens: after some time, its cpu usage went from ~0.1% to ~62%,
without reducing afterwards (in case it was just a peak usage) until
rebooting.  In a first moment, I thought this was due to something
related to "make buildworld" or "portupgrade -a" (which could last for
4 days if there are *many* fat ports to update, like firefox), but
this morning the system thread gone crazy without anything happening
in the system -- except for the fire saver.  When in 5-STABLE or
6-STABLE it went "only" up to ~20% of cpu usage.

The machine is up for 9 hours, I did lots of compilations and plenty
of network traffic, and the 'swi4: clock' thread is behaving nicely
(0.00 to 0.2% of cpu usage).  I configured the machine to not enable
the screensaver.  I'm not completely sure, but I don't remember this
behavior happening if there was a X session running with the terminal
"pointing" to it, instead of a tty.

Details about the machine:
rnsanchez at camelo:~% uname -a
FreeBSD camelo.lan.box 7.0-CURRENT FreeBSD 7.0-CURRENT #1: Fri May 26
17:29:30 BRT 2006
root at camelo.lan.box:/home/cvsup/obj/home/cvsup/src/sys/CAMELO  i386

kernconf <http://rnsanchez.wait4.org/FreeBSD/CAMELO.txt>
pciconf -lv <http://rnsanchez.wait4.org/FreeBSD/pciconf.txt>
dmesg <http://rnsanchez.wait4.org/FreeBSD/dmesg.txt>
sysctl -a <http://rnsanchez.wait4.org/FreeBSD/sysctl.txt>

I'd be glad to provide additional information, just tell me what you need :)

Regards, and thanks in advance.

-- 
Ricardo Nabinger Sanchez
GNU/Linux #140696 [http://counter.li.org]
Slackware Linux + FreeBSD