On 22/05/2019 23:20, Jakub Kicinski wrote: > On Wed, 22 May 2019 22:37:16 +0100, Edward Cree wrote: >> * removed RFC tags > Why? There is still no upstream user for this Well, patch #2 updates drivers to the changed API, which is kinda an "upstream user" if you squint... admittedly patch #1 is a bit dubious in that regard; I was kinda hoping someone on one of the existing drivers would either drop in a patch or at least point me at the info needed to write one for their HW, but no luck :-( My defence re patch #1 is that I'm not really adding a 'new feature' to the kernel, just restoring something that used to be there. It's a grey area and I'm waiting for the maintainer(s) to say yea or nay (Jiri noped v1 but that had a much bigger unused interface (the TC_CLSFLOWER_STATS_BYINDEX callback) so I'm still hoping). > (my previous objections of this being only partially correct aside). I didn't realise you still had outstanding objections, sorry. Regarding RTM_GETACTION dumping, it should be possible to add an offload callback based on this, which would just pass the action cookie to the driver. Then the driver just has to look the cookie up in its stats tables to find the counter. That would deal with the 'stale stats' issue. But right now, that really _would_ be an API without a user; still, I might be able to do that as an RFC patch to prove it's possible, would that help this series to go in as-is?
-Ed
