[rt2x00-users] [RFC/RFT 0/8] rt2x00: Various clean-ups and RT3572 fixes.

[Gertjan van Wingerde]

On Wed, Feb 1, 2012 at 3:36 PM, Andreas Hartmann
<andihartmann at 01019freenet.de> wrote:
>>
>> I have to refine this.
>>
>> kernel 3.1 runs on another machine (Intel Core i5 - quadcore, 64bit) -
>> kernel 2.6.37.6 on an Intel CeleronM (single core, 32bit).
>>
>> Now, I put kernel 3.0.3 on the CeleronM machine and the Core i5. Source
>> code (compat-wireless) is the same as mentioned above. On the CeleronM
>> machine, the test runs as expected. On the Core i5 machine, I'm getting
>> hard stalls after a few seconds or few minutes (DEBUG is switched off).
>> But I didn't saw the above usb error.
>>
>> I think there is an issue with smp or 64bit machines (my guess: usb
>> handling issues on fast machines :-)).
>
>
> Refine part 2:
>
> Tested now on 64bit smp machine with 2.6.37.6 (same kernel which runs
> the tests on 32bit CeleronM single core mostly fine): It stalled after
> few seconds.
>
> Conclusion: seems to be unusable on 64bit smp machines, eventually just
> on smp machines only.
>
>

OK. That could be the case as my tests were on a netbook with a single core
Atom N270 CPU.

My guess it is more related to SMP rather than 64-bit, as we've had more issues
around that in the past.

Now to find a way to test and reproduce in my own environment (problem is that
my big quad-core i7 server is connected via wired ethernet and I cannot disrupt
the network connectivity to run a Wifi performance test on that one) :-(

And to find the time to do all this :-)

---
Gertjan