Re: HEAD/6.1_RC2 DOM0 problems?

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Sat, Mar 23, 2013 at 08:15:50AM -0700, Jeff Rizzo wrote:
> I just tried a 6.1_RC2 dom0 kernel on a dual opteron box which was
> working fine with a pre-6.0 dom0 kernel, and which works fine with a
> 6.1_RC2 GENERIC.  Things seem to start booting ok, then
> really slow down once they get to wd1, and eventually hang
> altogether.  I've seen it get as far as:
> 
> root file system type: ffs
> 
> ... but I've also seen it hang slightly earlier (after detecting
> wd1, somewhere in the USB).
> 
> I tried a 6.0.1 dom0 kernel, and it works fine.  A -current dom0
> from 20130322, though, hangs *even earlier*:
> 
> NetBSD 6.99.18 (XEN3_DOM0) #0: Fri Mar 22 12:55:23 UTC 2013
> builds%b8.netbsd.org@localhost:/home/builds/ab/HEAD/amd64/201303221130Z-obj/home/builds/ab/HEAD/src/sys/arch/amd64/compile/XEN3_DOM0
> total memory = 256 MB
> avail memory = 232 MB
> mainbus0 (root)
> cpu0 at mainbus0 apid 0
> cpu1 at mainbus0 apid 2
> cpu2 at mainbus0 apid 1
> cpu3 at mainbus0 apid 3
> ioapic0 at mainbus0 apid 4
> (XEN) ioapic_guest_write: apic=0, pin=0, irq=0
> (XEN) ioapic_guest_write: new_entry=000109f0
> (XEN) ioapic_guest_write: old_entry=00010000 pirq=0
> (XEN) ioapic_guest_write: Attempt to modify IO-APIC pin for in-use IRQ!
> (XEN) ioapic_guest_write: apic=0, pin=2, irq=0
> (XEN) ioapic_guest_write: new_entry=000109f0
> (XEN) ioapic_guest_write: old_entry=000009f0 pirq=0
> (XEN) ioapic_guest_write: Attempt to modify IO-APIC pin for in-use IRQ!
> (XEN) irq.c:1514: dom0: pirq 0 or irq 3 already mapped
> (XEN) ioapic_guest_write: apic=0, pin=4, irq=4
> (XEN) ioapic_guest_write: new_entry=000109f1
> (XEN) ioapic_guest_write: old_entry=000009f1 pirq=0
> (XEN) ioapic_guest_write: Attempt to modify IO-APIC pin for in-use IRQ!
> (XEN) irq.c:1514: dom0: pirq 0 or irq 5 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 6 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 7 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 8 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 9 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 10 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 11 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 12 already mapped
> (XEN) irq.c:1514: dom0: pirq 0 or irq 13 already mapped
> ioapic0: can't remap to apid 4
> ioapic1 at mainbus0 apid 5
> ioapic1: can't remap to apid 5
> ioapic2 at mainbus0 apid 6
> 
> <hang>
> 
> 
> ...does anyone know what might be happening?  This seems like
> something we should try to fix before release.  :)

The 6.1 problem has been reported by someone else already, also on a amd
system.  I suspect the recent ACPI changes, which cause an interrupt storm,
but I don't have more details. There's no difference in the dmesg.
I can't reproduce this on my Intel systems.

I was not aware of the HEAD issue, maybe we should try to not remap ioapic
which already has the right apid ...

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--