From: Gerald Britton (gbritton_at_doomcom.org)
Date: 2002-09-01 18:59:07 UTC
Hmmm.. I'm wondering if I've run into a bug in the PCI suspend/resume code i contributed a few months ago, was wondering if anybody else had run into it. It seems that during suspend the machine locks up if the network is heavily loaded. I've not been able to reliably reproduce the problem though. My best guess is that an interrupt is occuring after the device is disabled and powered down. The interrupt routine blindly accesses registers, so probably needs some form of protection should the card be powered down during a suspend.
... current idea is going to be disabling interrupts before shutting down the card. It looks like this is probably the right approach, but i notice that much of this is unprotected by spinlocks and i suspect it might lose under SMP. hmmmm...