Jonathan Cameron wrote on Mon, Apr 01, 2024 at 05:47:56PM +0100: > Good luck. If you have time it might be good to hear what you end up > with! Just a quick follow-up since you asked -- given we manage our own kernel that already has its share of patches and it's not something user-visible we'll stick with the aliases approach for this kernel to make identifiers static. (and I'm adding labels for meticulous users, but not expecting it to be used in practice, it'll mostly be used in automated testing to make sure the number doesn't change on our end) The rationale was as per my previous mails that paths in /sys/devices/platform have changed in the past so we'd rather not rely on these being set in stone, and while a new symlink would have been workable it's a user-noticeable change so we've prefered just pinning the device numbers. I'm always reluctant to take in more "in house" patches in our tree but in this case it's "simple enough" (death by thousands paper cut?), and we'll rediscuss this if/when another upstream solution shows up. Thanks a lot for your time thinking it through and discussing it though, that was appreciated! (Jonathan and everyone else involved) -- Dominique
