Comment 31 for bug 21574

I have tracked it down! (Still don't know the mechanism, but there is a fix.)

The bug was fixed in commit 30b35399ceb2398d05837863476dcb12f12f3a82

It was broken again in commit 19bfafb2ed1a59efb979b6725ab7626a94f7e078

The latter only removes one line, a call to pci_set_power_state() in intel8x0_suspend() (a line that had been added in the first commit).

Simply unapplying that commit fixes the problem. The bizarre thing is that the latter commit was designated as correcting the speaker problem on suspend to RAM. At least for my laptop (HP nw8240) this patch actually *causes* the problem.