Bjorn Helgaas wrote: [..] > v1/rfc: https://lore.kernel.org/all/20250304135108.2599-1-ilpo.jarvinen@xxxxxxxxxxxxxxx/ > > I suppose it's quixotic to hope for anything better than quirks that > have to be added or updated for every new processor that comes along. > > ACPI _HPX might be a possible way for the platform to tell us what to > do here. ACPI r6.5, sec 6.2.9 says it's for hot-added devices and > "Functions not configured by the platform firmware during initial > system boot" (how are we supposed to determine that?) In any case, > Linux does evaluate _HPX for every device in pci_configure_device(). > > I'm not sure _HPX really works; it's very general, and I would expect > to see reports of problems if firmware really tried to use it. > > Or, I guess a _DSM function would be a possible way to communicate > this. Ok, I am reading this as "Maintainer asserts the quirk is unsavory, please make this be something that Linux can ask the platform firmware if it needs to apply, or make existing _HPX just work. Either of those is preferable to a new entry in the quirk table."
