Re: [B.A.T.M.A.N.] ibss.c backtrace when batman-adv adds wireless interface

2016-02-03 Thread Sven Eckelmann 
On Wednesday 03 February 2016 18:09:40 Sven Eckelmann wrote:
> > AFAICT this should be a driver (or perhaps mac80211) issue, but I don't
> > see any information about the driver used.
> 
> It doesn't really look like it is batman-adv specific. batman-adv also doesn't
> be active because the message "batman_adv: bat0: Adding interface: " is
> missing.

Maybe I should clarify a little bit. I meant that that message is missing
before the problem happens. The interface message appears after the problem
was detected.

And maybe I should also inform you that many batman-adv refcounting
patches [1] for some long standing problems are getting queued up by the brave
linux-stable people. But there are also two patches [2,3] which Antonio hasn't
submitted yet.

Kind regards,
Sven

[1] https://kernel.org/pub/linux/kernel/v4.x/stable-review/patch-4.3.5-rc1.gz
[2] https://lists.open-mesh.org/pipermail/b.a.t.m.a.n/2016-January/014304.html
[3] https://lists.open-mesh.org/pipermail/b.a.t.m.a.n/2016-January/014305.html

signature.asc
Description: This is a digitally signed message part.

Re: [B.A.T.M.A.N.] ibss.c backtrace when batman-adv adds wireless interface

2016-02-03 Thread Sven Eckelmann 
On Wednesday 03 February 2016 16:41:30 Johannes Berg wrote:
> On Wed, 2016-02-03 at 10:26 -0500, Josh Boyer wrote:
> > On Wed, Feb 3, 2016 at 10:24 AM, Josh Boyer  > g> wrote:
> > > Hi All,
> > > 
> > > We've had a user report the backtrace below when loading batman-adv 
> > > on
> > > his machine.  It looks like the cfg80211 layer is complaining about
> > > a
> > > null bss returned, but I cannot tell if the rtlwifi driver or
> > > batman-adv is in error here.
> > > 
> > > Thoughts?
> > 
> > Sorry, forgot to include the link to the actual bug report:
> > 
> > https://bugzilla.redhat.com/show_bug.cgi?id=1304428
> > 
> > Reporter says this is new with 4.3.y and did not happen on e.g.
> > 4.2.8.
> > 
> 
> AFAICT this should be a driver (or perhaps mac80211) issue, but I don't
> see any information about the driver used.

It doesn't really look like it is batman-adv specific. batman-adv also doesn't
be active because the message "batman_adv: bat0: Adding interface: " is
missing.

batman-adv is also not responsible for the underlying device. So it will not
join IBSS and thus should not trigger the "wlp0s26u1u4: Selected IBSS BSSID"
seen in the log.

But that said, I already saw weirder things happen because of memory
corruptions. So I will not say that batman-adv cannot be the culprit but right
now I don't see how.

The user can also try (when he thinks that batman-adv is to blame) to bisect
the batman-adv versions via the git of the out-of-tree module [1]. The version 
[2]
from 4.1 should be roughly v2015.0 (b6ff92aa334083dee8a1f5f0c5aa0f7c4c2584cf),
v4.2 should be v2015.1 (95b64b1cef38340f48f4eba99689e0fb1feddcd0) and v4.3 
should
not be newer than v2015.2 (e25bd38b0f8b4aeca85ef24884fa5eadd6270709). So 
bisecting
between b6ff92aa334083dee8a1f5f0c5aa0f7c4c2584cf and
e25bd38b0f8b4aeca85ef24884fa5eadd6270709 (~6 steps) should cover it. I 
personally
would recommend to start on 4.2.8 with v2015.2
(e25bd38b0f8b4aeca85ef24884fa5eadd6270709)

Kind regards,
Sven

[1] https://git.open-mesh.org/batman-adv.git
[2] 
https://www.open-mesh.org/projects/open-mesh/wiki/Download#batman-adv-in-the-Linux-tree


signature.asc
Description: This is a digitally signed message part.

Re: [B.A.T.M.A.N.] ibss.c backtrace when batman-adv adds wireless interface

2016-02-03 Thread Sven Eckelmann 
On Wednesday 03 February 2016 10:24:43 Josh Boyer wrote:
> Hi All,
> 
> We've had a user report the backtrace below when loading batman-adv on
> his machine.  It looks like the cfg80211 layer is complaining about a
> null bss returned, but I cannot tell if the rtlwifi driver or
> batman-adv is in error here.
> 
> Thoughts?

About the attachment "batman-adv oopses with kernel-4.3.4-300+debug". Sorry,
this was my fault and caused by an copy+paster error while adding some
lockdep_asserts. This caused too many lockdep reports. It was fixed a while
ago upstream in 008a37448707 ("batman-adv: Fix lockdep annotation of
batadv_tlv_container_remove").

Kind regards,
Sven

signature.asc
Description: This is a digitally signed message part.