Comment 3 for bug 201037

after adding some bash -x jazz to both acpi-support and pm-utils' suspend paths (which I note are disturbingly different), it turns out that this is caused by the e1000 module not being unloaded during pm-util's suspend.
acpi-support unloads all networking modules, pm-support unloads none.

As such, this is either a bug in pm-utils for not removing e1000, or in e1000 for not suspending correctly. Possibly both.