Hi Ben,
On 10/4/24 12:18 PM, Ben Greear wrote:
On 10/4/24 11:46, James Prestwood wrote:
Hi,
I've noticed that when setting the colocated flag much of the scan
results are not appearing when they should be. This was seen in a
large warehouse where our network/SSID was using only the 6ghz band.
There were other networks on 2.4/5ghz though, which maybe is having
some effect, but those networks are not broadcasting on 6ghz.
This first came to my attention when the majority of our clients were
experiencing horrible signal quality. Upon looking at the logs and
nl80211 messages to the kernel IWD was issuing both limited 6ghz
scans as well as full spectrum scans and only showing a few 6ghz
results, like 1-2 BSS's which should not have been the case. I was
able to scan manually using "iw scan" and I saw all the expected
BSS's. I isolated it to the fact that IWD was setting the
SCAN_FLAG_COLOCATED_6GHZ and removing that flag resulted in all the
BSS's showing up in scan results. Note, that with or without the flag
all the 2.4/5ghz BSS's were showing up without a problem, it seems
completely isolated to the 6ghz band.
The NEW_SCAN_RESULTS event was indicating the kernel had scanned all
the expected frequencies we were asking to scan, but the results were
mostly empty for 6ghz. Only ever the connected BSS would show up and
_sometimes_ another BSS, but generally not.
I'm trying to read through the code associated with the COLOCATED
flag, but its not exactly intuitive (to me at least). I'm not sure if
this is something with mac80211 or at the driver level. I really
don't have anything to compare it to as ath11k is the only hardware
available at this warehouse with many APs around.
I'm looking for some guidance on if this is expected behavior with
the colocated flag, and if there is any concern with including it
unconditionally on every scan. If it has unintended consequences like
this, generally, across multiple drivers I'd like to re-think its use
in IWD.
I had some similar issues using hostap, and hacked it to do every 4th
or so scan with that
flag set differently. Hopefully that lets the 'fast' scan work when
it can, but still lets one
scan the hard way...
Possibly there are driver concerns too, I was using Intel be200 when
debugging this.
Thanks for the extra info. Seems like a mac80211 quirk if its also on an
Intel card. Long term I may just parse RNR elements in userspace and
issue extra scans if needed.
Thanks,
James
