Page MenuHomeFreeBSD

net80211: add TX rate control completion fields for HT, VHT, channel info
Needs RevisionPublic

Authored by adrian on Sun, Dec 15, 10:23 PM.
Referenced Files
Unknown Object (File)
Mon, Jan 13, 3:48 AM
Unknown Object (File)
Thu, Dec 26, 1:01 PM
Unknown Object (File)
Thu, Dec 26, 3:16 AM
Unknown Object (File)
Wed, Dec 25, 8:56 AM
Unknown Object (File)
Wed, Dec 25, 8:09 AM
Unknown Object (File)
Tue, Dec 24, 9:59 PM
Unknown Object (File)
Tue, Dec 24, 9:07 PM
Unknown Object (File)
Tue, Dec 24, 2:48 PM

Details

Reviewers
bz
Group Reviewers
wireless
Summary

Add frequency, NSS (number of spatial streams), MCS, PHY type and
channel width to the rate control TX completion flags.

This isn't an exhaustive list of useful information - I've also
thought about LDPC, STBC, MU-MIMO groups, etc - but they aren't
required to just get improved HT and some VHT awareness in the
rate control code.

It's not used by anything just yet, but since rtwn(4) is about to
grow some more VHT awareness it would be good to lay groundwork.

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Passed
Unit
No Test Coverage
Build Status
Buildable 61184
Build 58068: arc lint + arc unit

Event Timeline

fuz added inline comments.
sys/net80211/ieee80211_ratectl.h
78

Shouldn't you remove some of the spare bytes to account for the new fields?

Please also check if you can sort the new fields so there is no padding.

bz requested changes to this revision.Fri, Dec 27, 7:14 AM
bz added a subscriber: bz.
bz added inline comments.
sys/net80211/ieee80211_ratectl.h
76

So this gets added in D48099 but is used nowhere? So currently noise. Please show the usage for the new flags before committing.

78

@fuz: spares we only remove when MFCing usually or if they were reserved for a specific case and been implemented.

This revision now requires changes to proceed.Fri, Dec 27, 7:14 AM
sys/net80211/ieee80211_ratectl.h
76

I have plans for this stuff to get used in HT and VHT aware rate control stuff, but I don't want to start on that until the diff stack is landed.
This however at least establishes what we think is good information to populate.

emaste added inline comments.
sys/net80211/ieee80211_ratectl.h
78

Good point about sorting new fields. Also 15 bytes of spare might not make the most sense now (and maybe didn't before, either).