Trouble with SR71-A on embedded device

[eschoeller]

I was able to pickup an SR71-A a few days ago.

I installed it into my Asus WL500-W router with one of the recent versions of openwrt. I had lots of trouble getting it to work, and tried several different builds, and eventually built openwrt myself - but still had no luck with the card.

I decided to pop it into my laptop running Kubuntu 8.10. The card worked just fine (it showed up in lspci, and ifconfig)

On my Asus WL500w "lspci" fails when the card is installed (blank output). When I install either a BCM43xx card or an Intel ipw2200 lspci works, and I am able to see the card.

With an older version of openwrt (2.4 series kernel) lspci works, and I can see the card in the output, however there is no ath9k driver for openwrt 2.4 so I am unable to use the card.

This may be strictly an openwrt issue, but has anyone else had this experience using the SR71-A on an embedded linux device based on the BCM947xx/953xx ?

[UBNT-Mike.Ford]

Hello,

I do apologize that I cannot support OpenWRT here, but hopefully some of the other linux guru's can jump in.

Thanks,

Mike

[key4078]

We also have looked for drivers for SR71-A so far, and the following two drivers are the only ones available for us now.

(1) ath9k (probably you already know about it)
- It doesn't support AP mode
- It supports only recent kernels

(2) Atheros driver
- It is for SparkLan's WMIA-199N/268N.
(I guess it is originally for Atheros' MB82 reference design)
- works only with 2.6.21.6 kernel and ixp4xx boards

[eschoeller]

Yes, I have heard about the options for drivers. The problem is that I can't even get far enough with it to start trying different driver options. When installed, the SR71-A doesn't show up at all - not even in an "lspci" output. Running "lspci" shows absolutely nothing, almost as if the entire PCI bus is gone. After removing the SR71-A PCI devices are visible again.

Using the card in the ubuntu system, lspci works - and ath9k picks it up fine. So, it would seem like this is an issue with the openwrt 2.6 series kernel, but I posted here hoping someone else might have a similar problem - that their entire PCI bus seems to dissappear - after installing this card.

[eschoeller]

I am bumping this, anyone have success with SR-71A and openwrt?

[iddione]

Hi,

I put the SR71A on a mikrotik routerboard 433. I built openwrt two days ago including ath9k module. You must set some values in the hostapd config files in order to have it support the 802.11n protocol.
I can confirm that intel proset utility is reporting a G/N network (on channel 6), but I couldn't reach more than 24 mbits throughput with a simple file transfer between my laptop featuring a 4965 AGN card and a wired pc.
The Intel web site reports that throughputs better than 56 mbits can only be achieved with no encryption or aes wpa2, and they advice to configure the AP on the 5 ghz frequency for channel bonding to be effective. Also, I have connected to the sr71a a set of 3 5db antennas operating only on the 2.4 frequency and I can't test this last configuration for the moment.

Any advice is welcome,

greetings

simone

[iddione]

I'm going on with the tests. I couldn't reach throughput higher than the one reported because the proset utility does not connect in 11n mode but only in G mode. It reports an error "security mismatch with 11n", which looks like a placeholder for something else. Such an error surely is not related to security settings, as it is given also when no security is configured (open access point with no encryption). I'm using now WPA2 AES.
I also tried to connect under debian, after having installed a snapshot 2.6.28 kernel, which seems to provide an enanched version of the iwl4965 driver, now called iwlagn, but also in this case I could connect only in G mode. Also, I have a bunch of lines in the output of iwlist which are clearly buggy, as for example

IE: Unknow: [hex]

It looks like none of this driver is ready yet to work in 11n mode. By the way, AP mode works but just in G mode.

[eschoeller]

Ah, glad to hear you have it working .. somewhat. I need to take another stab at it. What specific settings did you add to the hostapd config?

[iddione]

Hi,

to give it a try, I suggest to compile the system from a snv checkout. I did it today and I noticed they updated the kernel version for ar71xx to the 2.6.28-5. The important steps when compiling should be:

1 - Select ath9k and remove madwifi.
2 - Select the package CRDA if you want to use the 5 ghz spectrum.

When you log into the system, after setting the basic configuration, you can edit the file /lib/wifi/hostapds.sh adding the following two lines almost at the end
ieee80211n=1
ht_capab=[HT40-][SHORT-GI-40][DSSS_CCK-40]

This is my /etc/config/wireless

root@OpenWrt:~# cat /etc/config/wireless

config 'wifi-device' 'wlan0'
option 'type' 'mac80211'
option 'disabled' '0'
option 'channel' '40'

config 'wifi-iface'
option 'device' 'wlan0'
option 'network' 'lan'
option 'mode' 'ap'
option 'ssid' 'OpenWRT'
option 'encryption' 'psk2'
option 'key' 'secretpassword1'


My problem is that I have a intel 4965 agn card as client. I could connect in G and A mode from debian (running 2.6.28-1 kernel and last version of compat-wireless) and from windows XP and Vista. Only under XP I have a small evidence that N is running, as it is reported from the proset utility (but not used due to a quite strange error). I found on the mailing list of the wireless.kernel.org a post where a guy says that aggregation in iwlagn is not working (so no N for the moment).
If someone has a card listed here (http://wireless.kernel.org/en/users/Drivers/ath9k#supportedchipsets), would be interesting to know if he/she can reach N throughput using ath9k also on the client side.

UPDATE: following the advice of a guy in the mailing list of openwrt I can now connect in N mode from vista and XP but not linux with my intel client. The connection is suffering packet dropping as reported also from other users and the quality is really bad (it's almost unusable). It looks like that Broadcom clients can connect succesfully and with good quality with other atheros chipset, so it's reasonable to assume that should be also the case with the SR71 cards. Maybe someone which have a broadcom card could help.
The N connection was achieved adding the following lines

wme_enabled=1
#
# Low priority / AC_BK = background
wme_ac_bk_cwmin=4
wme_ac_bk_cwmax=10
wme_ac_bk_aifs=7
wme_ac_bk_txop_limit=0
wme_ac_bk_acm=0
# Note: for IEEE 802.11b mode: cWmin=5 cWmax=10
#
# Normal priority / AC_BE = best effort
wme_ac_be_aifs=3
wme_ac_be_cwmin=4
wme_ac_be_cwmax=10
wme_ac_be_txop_limit=0
wme_ac_be_acm=0
# Note: for IEEE 802.11b mode: cWmin=5 cWmax=7
#
# High priority / AC_VI = video
wme_ac_vi_aifs=2
wme_ac_vi_cwmin=3
wme_ac_vi_cwmax=4
wme_ac_vi_txop_limit=94
wme_ac_vi_acm=0
# Note: for IEEE 802.11b mode: cWmin=4 cWmax=5 txop_limit=188
#
# Highest priority / AC_VO = voice
wme_ac_vo_aifs=2
wme_ac_vo_cwmin=2
wme_ac_vo_cwmax=3
wme_ac_vo_txop_limit=47
wme_ac_vo_acm=0
# Note: for IEEE 802.11b mode: cWmin=3 cWmax=4 burst=102
ieee80211n=1
ht_capab=[HT40-][SHORT-GI-40]


the topic is at this address http://forum.openwrt.org/viewtopic.php?id=18577&p=2.

[iddione]

Hi,

today I made a test with a mac book pro. I could reach 44 mb/s on a large file transfer. The connection is still not perfect as I experienced lags during the transfer. By the way I could surf the web without problem. It looks like the developer are doing a very good job.