On 12/05/2016 09:28 AM, Felix Fietkau wrote:
On 2016-12-05 16:01, Ben Greear wrote:
On 12/05/2016 06:53 AM, Johannes Berg wrote:
Unless I screwed up, this patch also returns an average.
Oops, sorry. I missed the whole mac_div() indirection thing.
I'm not super convinced anyway though - all of this data already is
available in a much more reliable fashion, even trackable when stations
are removed (all data gets sent in the DEL_STATION notification), so
adding a crippled way to get the same data seems a bit strange?
Ethtool stats are easy to program against, and I am already making the
get-stats call to get other things, so it was a quick tweak to return
these additional values. I agree the stats are of somewhat limited worth,
but the cost to get them is also pretty small code wise :)
You're still adding bloat for everybody for the sake of a little
convenience on your side. I think that's a bad trade-off.
nl80211 is not *that* hard to program against...
I don't really see much point in duplicating an arbitrary subset of
nl80211 information in ethtool.
I don't care that much either way. I already have local features
that cannot go upstream (per radio station hashes, ability to
set an advertised rateset, VHT on 2.4Ghz, 4.9Ghz support,
forked ath10k driver, etc), so I need to run my own kernel regardless.
I'll repost with the comment removed since I said I would. Johannes
can apply it if he wishes.
I'm more interested in getting things like the 'iterate' logic fixed
and the ath10k crash bugs resolved.
Thanks,
Ben
--
Ben Greear <greearb@xxxxxxxxxxxxxxx>
Candela Technologies Inc http://www.candelatech.com
