Re: PATCH: prevent repeated reconnections with some chipsets

[Derek Atkins <warlord MIT EDU>]

Dan Williams <dcbw redhat com> writes:

>> 4. Check /why/ we are timing out after 8 seconds. Is it because the card
>> is busy scanning still? Then wait for at least 20 seconds before
>> starting a reconnect. 
>
> Can we get more test data?  Does this happen when the signal strength is
> marginal?  Does the driver send disconnection events?  You might try
> re-enabling the detailed debugging output in nm-device-802-11-wireless.c
> by adding adding the '-ddd' parameter to the wpa_supplicant command line
> in supplicant_exec(), then we'll be able to see from /var/log/messages
> if wpa_supplicant is really getting disconnection events from the
> driver.
>
> Again; I don't want to do sometihng like increase the timeout if we
> don't really know what the cause of the problem is.

I've seen this issue on FC5 with NM 0.6.4 and the madwifi-ng 0.9.2
driver.  It seems to happen particularly in locales with multiple APs
on the same ESSID.   If I connect purely through ifup/ifdown then I
NEVER have a problem, but NM seems very unstable.

I really don't have control over the environments so I can't really
test much, I'm afraid.  For example, two weeks ago at the IETF in
Prague NM didn't work at all.  It would never associate/connect to any
AP, but if I used ifup/ifdown directly it would.  From where I sit the
problems NM 0.6.5 still seem to be:

  - ESSIDs with multiple APs (problems connecting, roaming)
  - non-broadcasting ESSIDs  (can't connect)
  - two ESSIDs (a and b/g) on the same BSSID (can't properly differentiate)

I haven't tried this particular patch yet, but I was planning to the
next time I'm in an environment where I see this particular lossage.

-derek
-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord MIT EDU                        PGP key available