Regression in 2.6.26-gitX: failure of wireless due to commit 175f9c1bba9b825d22b142

Regression in 2.6.26-gitX: failure of wireless due to commit 175f9c1bba9b825d22b142

[Larry Finger]

Wireless communication has failed for recent 2.6.26-gitX kernels - a definite 
regression. The logs show sequences such as:

eth1: authenticate with AP 00:1a:70:46:ba:b1
eth1: authenticate with AP 00:1a:70:46:ba:b1
eth1: authenticated
eth1: associate with AP 00:1a:70:46:ba:b1
eth1: associate with AP 00:1a:70:46:ba:b1
eth1: associate with AP 00:1a:70:46:ba:b1
eth1: association with AP 00:1a:70:46:ba:b1 timed out

I have this problem with b43 and p54usb, and there is at least one report of the 
same problem with zd1211rw. In my case, the encryption method is WPA-PSK/TKIP. I 
have not tested with WEP or unencrypted networks.

Bisection yields the following:

175f9c1bba9b825d22b142d183c9e175488b260c is first bad commit
commit 175f9c1bba9b825d22b142d183c9e175488b260c
Author: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Date:   Sun Jul 20 00:08:47 2008 -0700

     net_sched: Add size table for qdiscs

     Add size table functions for qdiscs and calculate packet size in
     qdisc_enqueue().

     Based on patch by Patrick McHardy
      http://marc.info/?l=linux-netdev&m=115201979221729&w=2

     Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
     Signed-off-by: David S. Miller <davem@davemloft.net>

========================================================

I tried to reverse apply the patch; however, that does not happen cleanly, but I 
am confident that the problem arises here. I don't understand enough about this 
qdisc stuff to even guess at what might be wrong.

As always, I am willing to test any trial patches, etc.

Thanks,

Larry

Re: Regression in 2.6.26-gitX: failure of wireless due to commit 175f9c1bba9b825d22b142

[David Miller]

From: Larry Finger <Larry.Finger@lwfinger.net>
Date: Sat, 26 Jul 2008 00:15:32 -0500

> Wireless communication has failed for recent 2.6.26-gitX kernels - a definite 
> regression. The logs show sequences such as:
...
>      net_sched: Add size table for qdiscs
> 

Yes we know, it exposes a larger problem that's been
present in the wireless stack for a long time.  It
assumes that the SKB control block can survive across
the SKB going through the qdisc, but that has never
been true.

We're thinking about how to solve this but have no
easy answers yet.

Re: Regression in 2.6.26-gitX: failure of wireless due to commit 175f9c1bba9b825d22b142

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 521 bytes --]

On Sat, 2008-07-26 at 00:33 -0700, David Miller wrote:

> Yes we know, it exposes a larger problem that's been
> present in the wireless stack for a long time.  It
> assumes that the SKB control block can survive across
> the SKB going through the qdisc, but that has never
> been true.
> 
> We're thinking about how to solve this but have no
> easy answers yet.

There are three places afaik, and I have a plan on how to fix the worst
two of them, the third one is in the disabled AP mode code.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]