Re: [ath9k-devel] [PATCH] mac80211: disassociate when switching bands and not scanning

[reinette chatre <reinette.chatre@intel.com>]

Hi Luis,

On Thu, 2009-06-11 at 12:04 -0700, Luis R. Rodriguez wrote:
> On Wed, Jun 10, 2009 at 3:44 PM, Luis R.
> Rodriguez<lrodriguez@atheros.com> wrote:
> > On Wed, Jun 10, 2009 at 3:42 PM, Luis R.
> > Rodriguez<lrodriguez@atheros.com> wrote:
> >> On Wed, Jun 10, 2009 at 2:07 PM, Johannes Berg<johannes@sipsolutions.net> wrote:
> >>> On Wed, 2009-06-10 at 14:04 -0700, Luis R. Rodriguez wrote:
> >>>> On Wed, Jun 10, 2009 at 2:03 PM, reinette
> >>>> chatre<reinette.chatre@intel.com> wrote:
> >>>>
> >>>> >> > This touches code targeted for removal in recent "mac80211: disconnect
> >>>> >> > when user changes channel" patch.
> >>>> >>
> >>>> >> Heh I don't see that patch yet but if it exists that should fix this
> >>>> >> too unless that patch didn't deauth/disassoc. Did you try with it
> >>>> >> applied by any chance? I can't seem to find that patch on the list.
> >>>> >
> >>>> > That patch does seem to have same intention as yours. It is still RFC so
> >>>> > I am not sure about its status. I did not test with it, but will do so.
> >>>> > The patch can be found at
> >>>> > http://marc.info/?l=linux-wireless&m=124458448631186&w=2
> >>>>
> >>>> Thanks yes, I see, yup I think that should fix it, we need a fix for
> >>>> stable though. Not sure how John or Johannes want to handle this.
> >>>
> >>> *scratches head*
> >>> are you advocating those rate cleanups for stable?
> >>
> >> No, a few fixes are needed for stable (now 2.6.30) though. For example
> >> the assert downgrade patch can be used but most likely that'll just be
> >> doing something completely wrong, so we do need to ensure the dynamic
> >> ps stuff does not trigger when scanning and I do consider that a
> >> proper fix for stable. I forget when ps stuff when in.
> >>
> >> The issue here is this *is* a bug not only for iwlwifi. iwlwifi
> >> currently gets no warning because of a work around in the driver's
> >> rate control algorithm as it is in stable. I haven't tried running
> >> this band change with ath9k
> >
> > Just to clarify -- I did run this with the patches I posted and I did
> > hit the warning, but I did not try this without the patches I posted.
> > Chances that will cause the kernel to commit suicide and oops.
> 
> As expected I was able to reproduce the warning with atht5k and ath9k
> on 2.6.30. I did not get a crash with ath9k though, which is good. But
> the warning is still present.
> 
> Do we want to fix this for stable? If so is the patch I posted reasonable?

I ran the steps to reproduce on a kernel that is the latest
wireless-testing + "mac80211: disconnect when user changes channel" +
"iwlwifi: remove rs_get_rate workaround". Unfortunately I still get the
rs_get_rate warning. 

Reinette