On Sun, Jun 22, 2025 at 08:47:03PM -0500, Mario Limonciello wrote: > Actually I came up with the idea to forbid runtime PM on the service when it > doesn't allow d3 at probe which I believe means no need to check again on > remove. > > This works cleanly for me. LMK what you think of this. User space can override the allow/forbid setting through the sysfs "control" attribute, hence this doesn't seem like a suitable way to somehow cache the return value of pci_bridge_d3_possible() and thus avoid its re-execution on remove. However we shouldn't have to cache the return value anyway as the underlying assumption is that the function always returns the same value. Thanks, Lukas
