vmspace changes to use sysref

[Matthew Dillon]

:Well, no panics yet, but I'm seeing some new badness with USB:
:
:...
:kernel: intr 10 at 50001 > 50000 hz, livelocked limit engaged!
:kernel: uscanner0: Canon CanoScan, rev 1.10/1.00, addr 2
:kernel: uhci2: <VIA 83C572 USB controller> port 0xdc00-0xdc1f
:irq 5 at device 16.2 on pci0
:kernel: usb2: <VIA 83C572 USB controller> on uhci2
:...
:...
:kernel: uscanner0: at uhub1 port 2 (addr 2) disconnected
:kernel: uscanner: ops removed
:kernel: uscanner0: detached
:kernel: uscanner0: Canon CanoScan, rev 1.10/1.00, addr 2
:kernel: intr 10 at 7389 < 20000 hz, livelock removed
:kernel: intr 3 at 7376 < 20000 hz, livelock removed
:
:The 5-minute hang goes away if I unplug the scanner before booting.
:
:I tried booting without ACPI, and partway thru the kernel messages I get
:an infinite number of these:
:sched_ithd : stray interrupt 7
:
:Any other data you might need?

    dmesg output from when you boot with an older kernel with the scanner
    plugged in (which presumably works?).

    There are a few things I would like to try.  First, adjust the livelock
    defaults in kern/kern_intr.c (livelock_limit and livelock_lowater).
    Try changing them to 5000 and 1000 respectively, just to see if that
    solves the 5-minute hang issue.

    If it doesn't (or even if it does), also try turning on the emergency
    interrupt polling feature, which can be done from the boot prompt or
    in /boot/loader.conf.

    kern.emergency_intr_enable="1"	in /boot/loader.conf
    set kern.emergency_intr_enable 1	from the boot prompt (I think).

    --

    When booting without ACPI, getting the stray interrupt messages. 
    It wasn't doing that with older kernels booted without ACPI?  or
    was it?

    What kind of machine is this?  duel-core?  AMD or Intel?

					-Matt
					Matthew Dillon 
					<dillon at backplane.com>