8.1-RC2 - PCI fatal error or MCE triggered by USB/ehci on Sun X4100M2?

[John Baldwin]

On Monday, July 12, 2010 8:41:51 am Markus Gebert wrote:
> 
> On 10.07.2010, at 01:53, Markus Gebert wrote:
> 
> >> I'm curious if disabling USB legacy support in the BIOS causes it to 
still die 
> >> even with ehci not loaded.  If so, then the SMI# for the ehci controller 
must 
> >> somehow prevent the issue, perhaps by triggering frequently enough to 
slow the 
> >> rate of I/O requests down?
> > 
> > 
> > I disabled usb legacy support in the BIOS and booted a kernel with 
usb+ohci+ukbd+ums but without ehci. Unfortunately, I cannot reproduce the MCE.
> 
> 
> Well, the situation has changed. Machine died over the weekend running our 
test load with above kernel configuration. It seems that not having ehci in 
the kernel at boot just makes the MCE much more unlikely to occur, but it 
occurs. With ehci, I can panic the machine within a minute, without ehci it 
seems to take at least hours. Still, I don't get why not having the ehci 
driver in the kernel should have any effect, especially because nothing is 
attached to it.

Ok, so maybe the SMI# interrupts do play a role somehow, at least as far as 
altering the timing.

> Panic message:
> 
> ----
> MCA: Bank 4, Status 0xb400004000030c2b
> MCA: Global Cap 0x0000000000000105, Status 0x0000000000000007
> MCA: Vendor "AuthenticAMD", ID 0x40f13, APIC ID 2
> MCA: CPU 2 UNCOR BUSLG Observer WR I/O
> MCA: Address 0xfd00000000
> panic: blockable sleep lock (sleep mutex) 128 @ 
/usr/src/sys/vm/uma_core.c:1992
> cpuid = 2
> KDB: enter: panic
> [thread pid 12 tid 100039 ]
> Stopped at      kdb_enter+0x3d: movq    $0,0x69ccb0(%rip)
> ----
> 
> Don't know, why it's not a fatal trap 28 this time despite an MCE was 
detected. Seen this before though, also with kernels that have ehci and with 
usb legacy support, so seeing a different panic this time seems not related to 
the way the kernel was configured. Maybe a symptom? Or may it even be useful? 
If yes, what should I pull out of DDB?
> 
> In the meantime, I'll try harder to reproduce the MCE on current...

Well, it panic'd trying to malloc something in a non-safe place, because the 
machine check can happen at any time like an NMI.  The panic was caused by the 
MCE however.

-- 
John Baldwin