Re: [PATCH] mac80211: Fix off-channel problem in work task.

[Stanislaw Gruszka <sgruszka@redhat.com>]

On Thu, Oct 20, 2011 at 08:22:55AM -0700, Ben Greear wrote:
> On 10/20/2011 07:58 AM, Stanislaw Gruszka wrote:
> >On Wed, Oct 19, 2011 at 11:44:36AM -0700, greearb@candelatech.com wrote:
> >>From: Ben Greear<greearb@candelatech.com>
> >>
> >>The ieee80211_cfg_on_oper_channel method compared the
> >>current hardware config as well as the desired hardware
> >>config.  In most cases, this is proper, but when deciding
> >>whether to go back on-channel, if the hardware is not
> >>configured on-channel, but logically it *should* be
> >>on-channel, then we must go on-channel.
> >>
> >>This patch adds a flag to the ieee80211_cfg_on_oper_channel
> >>logic to disable comparing the actual hardware so we do not
> >>have to create another tricky method with similar logic.
> >>
> >>Reported-by: Eliad Peller<eliad@wizery.com>
> >>Signed-off-by: Ben Greear<greearb@candelatech.com>
> >
> >I much more prefer previous one-line patch from Eliad
> >http://news.gmane.org/find-root.php?message_id=%3c1311607763%2d12603%2d3%2dgit%2dsend%2demail%2deliad%40wizery.com%3e
> >
> >this one seems to provide unneeded code complexity, but
> >behaviour i.e. number of channel switches in hardware
> >is the same.
> 
> I believe it's possible to set tmp_channel to NULL and not actually
> change the on/off channel configuration, and my patch should allow us to
> skip a hardware config in that case.
> 
> However, this might only be possible if we are in this code while
> scanning, and currently, I don't believe we can be in this work
> code while scanning.

So you want to optimize a case that is not even possible at present.
Do you know that "premature optimization is the root of all evil" ? :-)

I'm not quite against by that change, but I would like to have -stable
fixes, since bug we are talking about not only annoys people by warning,
but also disallow to associate to wireless network. I prefer that we
apply Eliad patches and cc -stable and do possible further optimization
on top of that (ideally if some measurement will be available that show
optimization really works).

Thanks
Stanislaw