Setting single rate in ath10k broken by "reject/clear user rate mask if not usable"

[Ben Greear]

On 10/18/2017 02:02 PM, Johannes Berg wrote:
> On Wed, 2017-10-18 at 13:51 -0700, Ben Greear wrote:
>> "
>> I guess you could implement this part? I.e. iterating the clients and
>> checking that they all support the rate that is set. However, then you
>> also need to implement that this gets reset when a new client that
>> doesn't support this rate connects.
>> "
>>
>> The first part seems OK, but the second seems like a pain.  Maybe just
>> keep a new client from being able to connect at all if it doesn't support
>> any available rates?
>
> I suppose if you reject the NEW_STATION command, then hostapd will
> reject the association though, so it's probably not hard to do.
> However, I'm not really sure why you'd want that? If you do want that
> then basically you're just implemented a very roundabout way of adding
> this rate to the basic rates, so you might as well just add it and work
> with the current basic rates check?

If a user specifies a specific rate or rate-set, then I do not think we
should quietly change it out from under them.  So, we could check at the
time the rate-set is applied (all current peers).  We can reject the change
at that point if one of the peers does not support any common rates.
And, whenever a peer tries to be associated, we can check that there is some common rateset
in place.  If there is no common rateset, then reject the association
one way or another.

> We'll need to be a little careful what happens with client-mode
> interfaces, which is where we actually observed hitting the WARN_ON()
> about not having any rates at all, but I think I already put a reset of
> the rates there anyway if they're not compatible with the AP. At least
> that's easier because there's only one client.

It would be easy to configure a station to do VHT MCS 8 only, and then
it would never be able to associate with an HT AP, for instance.  I don't
think this should be a WARN_ON event, just a failure.  It would be great
if we could get a useful error message back to the caller, but I am not
sure how feasible that is with the current netlink and mac80211 code.

If your main concern is hitting a WARN_ON, maybe just change it to a
quieter error message or remove it entirely and just return a failure
code?

Thanks,
Ben

>
> johannes
>


-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com