Re: ACS and Hostapd

Re: ACS and Hostapd

[Luis R. Rodriguez]

On Fri, Nov 23, 2012 at 01:44:51PM +0530, Yogesh Ashok Powar wrote:
> Hi Luis,
> 
> I was looking at your patches on the linux-wireless mailing list
> regarding ACS feature addition in hostapd.
> 
> It is seen that you went till v8 patches and still the patches are not
> seen upstream?

Correct.

> Is there any specific reason for this patches not being accepted?

It didn't get good test coverage and by the time I got results from Jouni
on his test I already lost interest on the code. If anyone wants to work
on it and enhance it feel free to do so.

> I am refering http://marc.info/?l=linux-wireless&m=131257363303693 seems
> like all in one patch. But I have trouble finding correct git base for
> it; since patch doens't apply cleanly even when I take hostap.git back
> to Jul end of 2011. For v7, 1 through 3 apply cleanly but not 4th.
> 
> Or I am I looking at some wrong hostapd git?

It should have been the same date / tag around when I submitted it. Sorry
I do not recall the exact commit.

  Luis

Re: ACS and Hostapd

[Jonathan Bither]

On 12/11/2012 02:42 PM, Luis R. Rodriguez wrote:
> On Fri, Nov 23, 2012 at 01:44:51PM +0530, Yogesh Ashok Powar wrote:
>> Hi Luis,
>>
>> I was looking at your patches on the linux-wireless mailing list
>> regarding ACS feature addition in hostapd.
>>
>> It is seen that you went till v8 patches and still the patches are not
>> seen upstream?
> Correct.
>
>> Is there any specific reason for this patches not being accepted?
> It didn't get good test coverage and by the time I got results from Jouni
> on his test I already lost interest on the code. If anyone wants to work
> on it and enhance it feel free to do so.
>
>> I am refering http://marc.info/?l=linux-wireless&m=131257363303693 seems
>> like all in one patch. But I have trouble finding correct git base for
>> it; since patch doens't apply cleanly even when I take hostap.git back
>> to Jul end of 2011. For v7, 1 through 3 apply cleanly but not 4th.
>>
>> Or I am I looking at some wrong hostapd git?
> It should have been the same date / tag around when I submitted it. Sorry
> I do not recall the exact commit.
>
>    Luis
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hello Yogesh,
     I myself was curious about the status of ACS. Thank you for 
following up about it. I have interest in this functionality and I would 
be happy to help test and assist with the implementation. I have devices 
which use ath5k, ath9k, and some misc broadcom chips.

Thanks,
Jonathan Bither

RE: ACS and Hostapd

[Yogesh Powar]

>On 13/12/2012 8:57 PM, Jonathan Bither wrote:
>>On 12/11/2012 02:42 PM, Luis R. Rodriguez wrote:
>> On Fri, Nov 23, 2012 at 01:44:51PM +0530, Yogesh Ashok Powar wrote:
>>> Hi Luis,
>>>
>>> I was looking at your patches on the linux-wireless mailing list
>>> regarding ACS feature addition in hostapd.
>>>
>>> It is seen that you went till v8 patches and still the patches are not
>>> seen upstream?
>> Correct.
>>
>>> Is there any specific reason for this patches not being accepted?
>> It didn't get good test coverage and by the time I got results from Jouni
>> on his test I already lost interest on the code. If anyone wants to work
>> on it and enhance it feel free to do so.
>>
>>> I am refering http://marc.info/?l=linux-wireless&m=131257363303693 seems
>>> like all in one patch. But I have trouble finding correct git base for
>>> it; since patch doens't apply cleanly even when I take hostap.git back
>>> to Jul end of 2011. For v7, 1 through 3 apply cleanly but not 4th.
>>>
>>> Or I am I looking at some wrong hostapd git?
>> It should have been the same date / tag around when I submitted it. Sorry
>> I do not recall the exact commit.
>>
>>    Luis
>
>Hello Yogesh,
>     I myself was curious about the status of ACS. Thank you for
>following up about it. I have interest in this functionality and I would
>be happy to help test and assist with the implementation. I have devices
>which use ath5k, ath9k, and some misc broadcom chips.
>
>Thanks,
>Jonathan Bither

Hi Jouni/Luis,

It seems that, we have some resources that can be utilized to test this
feature of ACS for hostapd.

Before we begin on this work of Luis and optimize/enhance it further, would you
like to recommend some immediate changes from the last test results?

Thanks
Yogesh

Re: ACS and Hostapd

[Luis R. Rodriguez]

On Thu, Dec 13, 2012 at 10:54 AM, Yogesh Powar <yogeshp@marvell.com> wrote:
> Hi Jouni/Luis,
>
> It seems that, we have some resources that can be utilized to test this
> feature of ACS for hostapd.
>
> Before we begin on this work of Luis and optimize/enhance it further, would you
> like to recommend some immediate changes from the last test results?

By my part:

The algorithm I wrote was pretty complex and although I did review it
with some folks who liked it and it made sense the algorithm can be
simplified by taking the input elements and using a multiplier for
each one and letting the user configure the multiplier. In other words
all the log stuff I did and linearizing it some way was not all that
necessary except for mental masturbation.

Apart from that its just a matter of testing and proving it works.

  Luis

Re: ACS and Hostapd

[Yogesh Ashok Powar]

On Thu, Dec 13, 2012 at 12:33:06PM -0800, Luis R. Rodriguez wrote:
> On Thu, Dec 13, 2012 at 10:54 AM, Yogesh Powar <yogeshp@marvell.com> wrote:
> > Hi Jouni/Luis,
> >
> > It seems that, we have some resources that can be utilized to test this
> > feature of ACS for hostapd.
> >
> > Before we begin on this work of Luis and optimize/enhance it further, would you
> > like to recommend some immediate changes from the last test results?
> 
> By my part:
> 
> The algorithm I wrote was pretty complex and although I did review it
> with some folks who liked it and it made sense the algorithm can be
> simplified by taking the input elements and using a multiplier for
> each one and letting the user configure the multiplier. In other words
> all the log stuff I did and linearizing it some way was not all that
> necessary except for mental masturbation.
> 
> Apart from that its just a matter of testing and proving it works.
Hi Luis,
   We tried your code of ACS from hostapd mailing list archieve, after
resolving some merge conflicts, on latest hostapd. Following are some
of our observations

a) The code can work for the drivers that dont support offchannel-tx.
   For such drivers, setting channel_time_tx to zero will make
   interference factor to depend only on busy time from passive
   listener's point of view.
b) HT 40 Mhz might have some issues with current ACS. When primary
   channel with least interference factor is selected secondary channel
   is set to offset from the primary instead of actual secondary
   channel.

Apart from the existing complex logic, explained here
http://wireless.kernel.org/en/users/Documentation/acs, can simple scan
based ACS algorithm feasible?

Lets say one just scans all the channel and gathers the beacons. Based
on the information based on collected beacons one can find out which
channel has least number of APs residing. One can even look into
detail IE informations in the beacons for detail study of wireless
environment.

This could help drivers that do not have access to parameters needed for
interference factor calculation.

Thanks
Yogesh