On Po 16-01-06 00:25:17, Matthew Garrett wrote: > In older kernels, power/state for PCI devices took a PCI power state as > an argument and so "3" was an entirely sensible thing to echo into it. > In current kernels, it hits a BUG() in pci_choose_state and things blow > up. > > While I realise that the former interface was broken and wrong, would it > be possible to move to a new one without breaking existing code? I had patch to fix this breakage, but it was rejected. What depends on that code? Pavel -- Thanks, Sharp!