On Mon, 04 Dec 2023 20:14:10 +0100 Johannes Berg wrote: > Heh. But do I want to get blamed for the (perhaps inevitable?) > performance regression? I guess I'll try ... I'd happily bet that nobody will notice. Feel free to add: Suggested-by: Jakub Kicinski <kuba@xxxxxxxxxx> If that makes it better? > Actually I could even still combine this with the netif carrier up count > in the wireless events, so we only have to do the rtnl_getlink if we > haven't seen an event yet, and - in the likely common case - save the > extra roundtrip? Though I guess it's not a huge problem, it's once per > connection basically. No objections to merging your carrier count patches to wireless, if you prefer to keep them. But it'd be nice to also have a generic mechanism.
