[ath9k-devel] Questions about virtual managed interfaces.

[Ben Greear <greearb@candelatech.com>]

On 07/01/2010 06:14 AM, Rodolphe Marques wrote:
> Hi Ben,
>
> I've been doing some work with ath9k virtual interfaces.
> Well first the MAC is different. It usually differs in the first and
> last byte.

How are you reading this MAC?  They seem the same everywhere I look.

[root at atom ~]# uname -a
Linux atom 2.6.35-rc3-wl+ #1 SMP Mon Jul 5 11:36:08 PDT 2010 i686 i686 i386 GNU/Linux

[root at atom ~]# ifconfig wlan1
wlan1     Link encap:Ethernet  HWaddr 00:0E:8E:20:09:4E
           UP BROADCAST MULTICAST  MTU:1500  Metric:1
           RX packets:0 errors:0 dropped:0 overruns:0 frame:0
           TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
           collisions:0 txqueuelen:1000
           RX bytes:0 (0.0 b)  TX bytes:0 (0.0 b)

[root at atom ~]# ifconfig wlan0
wlan0     Link encap:Ethernet  HWaddr 00:0E:8E:20:09:4E
           UP BROADCAST MULTICAST  MTU:1500  Metric:1
           RX packets:0 errors:0 dropped:0 overruns:0 frame:0
           TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
           collisions:0 txqueuelen:1000
           RX bytes:0 (0.0 b)  TX bytes:0 (0.0 b)

[root at atom ~]# cat /debug/ath9k/phy0/wiphy
primary: phy0 (ACTIVE chan=0 ht=0)
secondary: phy1 (ACTIVE chan=0 ht=0)
secondary: phy2 (INACTIVE chan=0 ht=0)
addr: ef:be:ad:de:ef:be
addrmask: ef:be:ad:de:ef:be

[root at atom ~]# cat /sys/class/ieee80211/phy0/macaddress
00:0e:8e:20:09:4e
[root at atom ~]# cat /sys/class/ieee80211/phy1/macaddress
00:0e:8e:20:09:4e
[root at atom ~]# cat /sys/class/ieee80211/phy2/macaddress
00:0e:8e:20:09:4e

> As for having 128 virtual interfaces I think that it is just impossible.
> For instance the default value for the scheduler is 500ms, which means
> that if you have 128 virtual interfaces each one will be served for
> 500ms each 64 seconds. At least I think this is how they work.
>
> I have been using only three virtual interfaces in managed mode and it
> works but the virtual interfaces get dissassociated a lot.

Ok, it would appear to me that the virtual phys are not really so
useful for my purposes then.  If you have a machine that is easy to
power cycle, I'm curious if your system also crashes if you add
a virtual phy and then rmmod ath9k (mine crashes every time).

Using a somewhat hacked ath5k in 2.6.31 kernel, we were able to support
128+ virtual stations (on a single physical NIC).  We basically put it into
promisc mode and then added some logic to deal with lots of stations
trying to scan and associate at once.  (We let the first one win, and
after it was associated, the rest of the stations could only scan that
one channel).  We used software encryption since the chip can't handle
more than one set of keys at a time (as far as we could tell).

Virtual APs were handled as well, with some logic to deal with lots
of beacons.

If we can't use lots of virtual phys, then we'll probably need to
take the same approach and just use lots of virtual stations on
ath9k too.

Thanks,
Ben



>
> Best Regards,
>       Rodolphe Marques
>
> On 07/01/2010 06:39 AM, Ben Greear wrote:
>> I'm trying to understand the features and limitations of the ath9k
>> virtual phy and virtual station support.  I'm using the 2.6.34 kernel.
>>
>> First, I was able to create new virtual phys with the
>> echo add>   /debug/ath9k/phy1/wiphy
>> logic.
>>
>> That seems to work OK, and a new wlanX interface is created.
>>
>> The first potential problem I see is that the wlanX interfaces
>> have the same MAC.  A previous post to this list made it appear
>> that they would be XORd with phy-id or similar to get a unique
>> MAC?
>>
>> It does appear that I can set the MAC address using 'ip link set ...', but I'm
>> not sure if that is the right approach.
>>
>> I also tried adding new managed interfaces using
>> iw phy phy1 interface add wlan1-1 type managed
>>
>> This also seems to work in as far as a new interface is created.
>>
>> Now, I am curious about configuring these things.
>>
>> *  If I have two managed wlans on the same phy, I assume they must
>>       both be on the same channel?  Is there any support in the driver
>>       to make sure that once one interface is associated on a particular
>>       channel that another interface cannot start scanning on different channels
>>       (and thus disrupt the first one)?
>>
>>       In some previous hackings of virtual station interfaces on ath5k, we
>>       had all sorts of issues getting a bunch of interfaces associated
>>       when using wpa-supplicant because one would start scanning and that would
>>       dis-associate the others..over and over.
>>
>> *  If I want, say, 128 virtual station interfaces, does it make sense to try
>>       to use 128 virtual phys, or should I aim at multiple stations per phy?
>>
>>       If anyone has tested the upper limits of how many virtual phys and/or
>>       stations can be supported, I'm interested to know!
>>
>> Thanks,
>> Ben
>>
>>
>
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com