Machine check on suspend

[Justin Hibbits]

Obviously this is a different case than most.  I'm working on
suspend/resume for PowerPC (PowerBooks, to be precise), and part of
that involves testing the cardbus.

When suspending, it suspends children first (obviously), followed by
parents, etc.  In this case, I have a ar5416 card in a cardbus slot.
The ath0 suspends, followed by cbb0, then a machine check:

Suspending ath0, child of cardbus0
Suspending cardbus0, child of cbb0
pci1:1:0:0: Resource still owned, oops. (type=3, rid=16, addr=88000000)
pci1:1:0:0: Resource still owned, oops. (type=1, rid=0, addr=3a)

fatal kernel trap:

   exception       = 0x200 (machine check)
   srr0            = 0x5d5530
   srr1            = 0x149030
   lr              = 0xd20896f8
   curthread       = 0x4c6e620
          pid = 12, comm = irq58: cbb0 ath0

[ thread pid 12 tid 100108 ]
Stopped at      0x5d5530:       sync
db> bt
Tracing pid 12 tid 100108 td 0x4c6e620
0xe4770a20: at +0x226530(326cbc)
0xe4770a50: at ar5416IsInterruptPending+0x3c(d20ca368)
0xe4770a70: at ath_intr+0xf0(d2064ef4)
0xe4770ab0: at cbb_func_intr+0x40(d1fd71b4)
0xe4770ad0: at +0x2074d4(307c60)
0xe4770b00: at +0x209034(3097c0)
0xe4770b50: at +0x203e78(304604)
0xe4770b80: at +0x4e4b70(5e52fc)

I'm using the projects/pmac_pmu branch, for anyone who may want to test
(synced with head a couple weeks ago).

Does anything need to be done before actually suspending the devices
(kill wpa_supplicant, or anything else)?

- Justin