"sysctl -a" hangs multiuser boot

[Paul Mather]

I'm having problems with a kernel built today from a csup done today.  I'm using an Xserve G5 running FreeBSD/powerpc64 9-CURRENT.  The system will not boot up multiuser: it hangs at "Entropy harvesting: interrupts ethernet point_to_point" and gets no further.  I can't Ctrl-C to get the boot past that point.

It appears it is the "sysctl -a" in the entropy kickstart portion of /etc/rc.d/initrandom that is causing the problem.

If I do "sysctl -a" manually it will produce output so far and then no further.  Here are the last few lines before it stops outputting anything more:

[[...]]
hw.usb.ugen.debug: 0
hw.usb.power_timeout: 30
hw.usb.uhub.debug: 0
hw.usb.no_pf: 0
hw.usb.proc.debug: 0
hw.usb.pr_recovery_delay: 250
hw.usb.pr_poll_delay: 50
hw.usb.uhid.debug: 0
hw.usb.ukbd.no_leds: 0
hw.usb.ukbd.debug: 0
hw.usb.ums.debug: 0



The system doesn't hang, but console input ceases to work and I can't SSH into the system any more.  I can still ping the machine, but ping times vary wildly, e.g.,

27 packets transmitted, 26 packets received, 3.7% packet loss
round-trip min/avg/max/stddev = 57.076/708.818/1395.450/280.318 ms

Those ping times are way higher than normal.  Here is an example of pinging the machine when it is not under the influence of doing a "sysctl -a":

23 packets transmitted, 23 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 0.421/0.591/0.758/0.096 ms


Also, at occasional intervals, this will be output on the console after having done a "sysctl -a":

max66900: iicbus read failed


(That's the only console output I have seen since issuing "sysctl -a".)

The fans will occasionally race, but eventually be brought under control.

For now, I have removed the "sysctl -a" from /etc/rc.d/initrandom so my system can boot, but I don't think that is a desirable long-term fix.

Any ideas for a good long-term solution?

Cheers,

Paul.