Hi all,
Is anyone working on this feature enhancement(i mean wireless roaming)?
I saw it on the to-do list, and that Mircea had proposed it for GSoC. I
thought i shall dig into it, but just wanted to know if anyone's
already being involved and actively working on it.
Thanks
Balaji R
Dan Williams <[email protected]> writes:
>> > 2) a way of telling userspace whether or not it has to handle
>> > roaming, now that new cards are showing up (iwm) that can handle it
>> > themselves
>>
>> What about older fullmac cards, do they support roaming in hardware?
>
> Yes. airo at least, and I think also atmel, orinoco, but not Libertas.
> Most of the fullmac cards do handle roaming between BSSIDs in the same
> SSID transparently. Libertas on the other hand pushes that control to
> the driver.
Can you tell anything about the configuration parameters available? For
example, can the driver configure an rssi threshold for enabling
background scanning[1]?
[1] I'm sure there are N+1 different names for this
--
Kalle Valo
On Mon, 2009-06-22 at 17:51 +0300, Kalle Valo wrote:
> Dan Williams <[email protected]> writes:
>
> >> > 2) a way of telling userspace whether or not it has to handle
> >> > roaming, now that new cards are showing up (iwm) that can handle it
> >> > themselves
> >>
> >> What about older fullmac cards, do they support roaming in hardware?
> >
> > Yes. airo at least, and I think also atmel, orinoco, but not Libertas.
> > Most of the fullmac cards do handle roaming between BSSIDs in the same
> > SSID transparently. Libertas on the other hand pushes that control to
> > the driver.
>
> Can you tell anything about the configuration parameters available? For
> example, can the driver configure an rssi threshold for enabling
> background scanning[1]?
On airo maybe, there is for example:
__le16 linkLossDelay;
__le16 maxBeaconLostTime;
__le16 refreshInterval;
which might have some effect on the firmware jumping to a new AP. Or it
might just be the threshold at which the driver gives up and signals
disassoc.
TBH, I'm actually not sure about atmel, I forgot that their firmware is
somewhat lighter than airo and orinoco. Later orinoco firmware versions
might give up more control to the driver.
Dan
On Thu, 2009-06-18 at 15:32 +0300, Kalle Valo wrote:
> Johannes Berg <[email protected]> writes:
>
> > On Wed, 2009-06-17 at 09:20 +0300, Kalle Valo wrote:
> >> Balaji Ravindran <[email protected]> writes:
> >>
> >> > Is anyone working on this feature enhancement(i mean wireless roaming)?
> >> > I saw it on the to-do list, and that Mircea had proposed it for GSoC. I
> >> > thought i shall dig into it, but just wanted to know if anyone's
> >> > already being involved and actively working on it.
> >>
> >> Nobody is currently working on it AFAIK. We are planning to talk about
> >> roaming at Berlin next week and I will have a brief overview to start
> >> the dicussion, but that's about it.
> >
> > Yeah, I don't think anybody is working on it.
>
> That's my understanding as well.
>
> > FWIW, what we need in the kernel/drivers is, I think
> >
> > 1) signals for significant signal change (offloaded to firmware for
> > beacons while no traffic is flowing)
>
> I agree. Also I think we need to have an user space interface to
> configure the threshold value.
>
> > 2) a way of telling userspace whether or not it has to handle
> > roaming, now that new cards are showing up (iwm) that can handle it
> > themselves
>
> What about older fullmac cards, do they support roaming in hardware?
Yes. airo at least, and I think also atmel, orinoco, but not Libertas.
Most of the fullmac cards do handle roaming between BSSIDs in the same
SSID transparently. Libertas on the other hand pushes that control to
the driver.
Dan
> > and probably some more. I don't suppose you will be in Berlin, but it
> > would probably be helpful for you to wait for that discussion to have
> > taken place before working on it, because then we'll have at least a
> > plan of where we want to see this go.
>
> I agree. I'm planning to write a small report of our roaming discussion.
> Maybe we can implement this finally :)
>
Balaji Ravindran <[email protected]> writes:
> Is anyone working on this feature enhancement(i mean wireless roaming)?
> I saw it on the to-do list, and that Mircea had proposed it for GSoC. I
> thought i shall dig into it, but just wanted to know if anyone's
> already being involved and actively working on it.
Nobody is currently working on it AFAIK. We are planning to talk about
roaming at Berlin next week and I will have a brief overview to start
the dicussion, but that's about it.
--
Kalle Valo
Johannes Berg <[email protected]> writes:
> On Wed, 2009-06-17 at 09:20 +0300, Kalle Valo wrote:
>> Balaji Ravindran <[email protected]> writes:
>>
>> > Is anyone working on this feature enhancement(i mean wireless roaming)?
>> > I saw it on the to-do list, and that Mircea had proposed it for GSoC. I
>> > thought i shall dig into it, but just wanted to know if anyone's
>> > already being involved and actively working on it.
>>
>> Nobody is currently working on it AFAIK. We are planning to talk about
>> roaming at Berlin next week and I will have a brief overview to start
>> the dicussion, but that's about it.
>
> Yeah, I don't think anybody is working on it.
That's my understanding as well.
> FWIW, what we need in the kernel/drivers is, I think
>
> 1) signals for significant signal change (offloaded to firmware for
> beacons while no traffic is flowing)
I agree. Also I think we need to have an user space interface to
configure the threshold value.
> 2) a way of telling userspace whether or not it has to handle
> roaming, now that new cards are showing up (iwm) that can handle it
> themselves
What about older fullmac cards, do they support roaming in hardware?
> and probably some more. I don't suppose you will be in Berlin, but it
> would probably be helpful for you to wait for that discussion to have
> taken place before working on it, because then we'll have at least a
> plan of where we want to see this go.
I agree. I'm planning to write a small report of our roaming discussion.
Maybe we can implement this finally :)
--
Kalle Valo
> Can you tell anything about the configuration parameters
> available? For example, can the driver configure an rssi
> threshold for enabling background scanning[1]?
Orinoco and WLAGS Hermes1 and Hermes2 has some "sensitivy"
setting, from 1..3:
$ iwconfig eth1 sens 3
What exactly this does is a bit opaque and handled by the
firmware.
Madwifi does roaming OK, here we have many iwpriv settings that
can be used to fine-tune this:
$ iwpriv eth1 bgscan 0
$ pre-up iwpriv eth1 rssi11g 26
$ pre-up iwpriv eth1 rssi11b 26
But this isn't exactly the driver, but their own 802.11
implementation.
On Wed, 2009-06-17 at 09:20 +0300, Kalle Valo wrote:
> Balaji Ravindran <[email protected]> writes:
>
> > Is anyone working on this feature enhancement(i mean wireless roaming)?
> > I saw it on the to-do list, and that Mircea had proposed it for GSoC. I
> > thought i shall dig into it, but just wanted to know if anyone's
> > already being involved and actively working on it.
>
> Nobody is currently working on it AFAIK. We are planning to talk about
> roaming at Berlin next week and I will have a brief overview to start
> the dicussion, but that's about it.
Yeah, I don't think anybody is working on it. FWIW, what we need in the
kernel/drivers is, I think
1) signals for significant signal change (offloaded to firmware for
beacons while no traffic is flowing)
2) a way of telling userspace whether or not it has to handle roaming,
now that new cards are showing up (iwm) that can handle it
themselves
and probably some more. I don't suppose you will be in Berlin, but it
would probably be helpful for you to wait for that discussion to have
taken place before working on it, because then we'll have at least a
plan of where we want to see this go.
johannes
Holger Schurig wrote:
>> Can you tell anything about the configuration parameters
>> available? For example, can the driver configure an rssi
>> threshold for enabling background scanning[1]?
>
> Orinoco and WLAGS Hermes1 and Hermes2 has some "sensitivy"
> setting, from 1..3:
>
> $ iwconfig eth1 sens 3
>
> What exactly this does is a bit opaque and handled by the
> firmware.
I wasn't paying too much attention to this thread. However as part of a
separate thread* I had a look into orinoco's sensitivity setting.
Anyway, see my e-mail earlier today for a description of what the values
do. The description should be valid for Intersil and Agere firmware
(including most if not all old versions), and applies to orinoco and
WL_LKM drivers.
Regards,
Dave.
* the thread is titled "[RFC 00/11] cfg80211: connect API + wireless
extension move"