Trouble using bcm4318 compact flash with b43 driver

[Dylan Cristiani]

On Wed, 19 Jan 2011 14:03:28 +0100
dylan cristiani <d.cristiani at idem-tech.it> wrote:

> On Tue, 18 Jan 2011 08:34:46 -0600
> Larry Finger <Larry.Finger at lwfinger.net> wrote:
> 
> .....
> > > Hi Larry i've some news: i tryied with kernel 2.6.35 and it works!
> > > the module associates to AP and i can browse the net, ping
> > > external addresses and so on...; it seems that some drivers'
> > > changes made things go worste for my 'platform': do you have any
> > > suggestions? Note that all kernel settings are the same in
> > > 2.6.35, 2.6.36 and 2.6.37 except, obviously, for the new kernel's
> > > parameters of the latter two.
> > 
> > The usual advice offered in cases like yours is to install a git
> > tree and bisect the issue. As you have a problem with mainline
> > kernels, I definitely recommend the linux-2.6.git tree as it
> > bisects better than development trees like wireless-testing.git.
> > Your number of trials should not be too bad as you know it broke
> > between 2.6.35 and 2.6.36.
> > 
> > Larry
> Tks for hint, but today i had a brand new behaviour: before diving
> into git, i tryed to add som debug printk into ssb backplane to see
> if there were some difference between 2.6.35 and 2.6.36, and the
> magic is that the module worked also with 2.6.36 kernel; for sure i
> don't really think that my few printk made the miracle; the only
> difference was that i manually made 'insmod' of ssb.ko and b43.ko,
> then i 'ifdown wlan0' then 'ifup wlan0' and it worked...; as it seems
> to me there must be some alea somewhere, that in certain conditions
> makes the module to misbehave and the AP forbids to associate to it;
> but i cannot get the point also for the fact that if a restore the
> starting buggy situation ('insmod' is performed by a script launched
> by inittat just after init), the 'race' situation doesn't happen
> anymore (for now), and the wlan0 still keep working: don't know where
> to hit my head......(desk already hitten!)
> 
> dylan (swimming in deep dark....)
> 
probably i found a clue (maybe the point): when the b43 wlan interface
comes up its rate is, at the beginning, setted to the lowest possible:
1Mbit, as i can see with iwconfig, and it seems that, at this low rates,
some access points doesn't work, becuase of some handshake timeout
failure happening or other different timeouts issue, like they can't
'deal' with such communication being too slow; if i set higher rates
(i.e. 'iw set rate 24M') it associates to access point (WPA2
encryption), it gets dhcp address and it is possible to ping other net's
nodes, also if there are still some issue, like high percentage of
packet lost and so on; furthermore, trying to force setting of the rate
to highest possible 'iw set rate 54M', doesn't improve the performances
but, at the opposite, it leads again to association failures; is this a
normal behaviour? do you think that i'm doing something wrong? why is
the starting rate setted to 1M (it's a wlan module policy or a driver
choice?); which is the best way to solve this issue if any?

as usual thanks for you support and patience!!

dylan