On Thursday 27 May 2010, Alan Stern wrote: > On Thu, 27 May 2010, Matthew Garrett wrote: > > > On Wed, May 26, 2010 at 06:14:52PM -0400, Alan Stern wrote: > > > > > We don't have to worry about that. If the hardware is broken and the > > > kernel doesn't realize it, the user will just have to tell it not to > > > allow runtime suspend for those devices. > > > > It's worse than that - Windows simply doesn't use legacy PCI PMEs for > > anything, so an increasing number of modern machines don't have them > > wired up. Defaulting to runtime PM simply isn't practical, since the > > behaviour will simply be that the devices stop working without any > > indication that they've stopped working. > > Tell me about it! The situation in USB is at least as bad... > > > The "easiest" workaround would > > be a thread that polls for PCI devices with set PME bits, along with a > > set of heuristics for determining whether the device can really wake up. > > ... and no such workaround is possible with USB. > > > Devices in the core logic (like USB generally is) will be fine in any > > case. It's discrete PCI devices that are the problem. > > In any case, I assume we'll continue to initialize devices with > "forbidden" status and require userspace (or drivers that have special > information) to explicitly allow runtime PM. Yeah. I'd leave that entirely to user space, though. Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
