Felix, in case anyone else has the same problem, your suggestion was correct. I installed the 2.6.32 kernel backport for Debian Lenny (from backports.org), built the latest daily "bleeding edge" compat-wireless from wireless.kernel.org and built the development version of hostapd (0.7.2). It wasn't as scary as I expected; the only significant complication was with nl80211 when building hostapd, and this was resolved by removing CONFIG_DRIVER_NL80211=y from hostapd.config. Everything seems to be completely stable, including with eth0 and wlan0 bridged. Thanks for your help! - Norm > On Sat, Jul 10, 2010 at 21:21, Felix Fietkau <nbd@xxxxxxxxxxx> wrote: >> >> On 2010-07-10 9:40 PM, Norman Henderson wrote: >> > Thanks very much Felix for the quick comment. I'll follow up on it >> > over the next few days as I get time to move to "bleeding edge" >> > (something I've sworn from my youth to never do...). >> > >> > However I guess I was more concerned about the regression from an >> > environment that worked (with disconnections) to one that doesn't >> > connect at all - is that just the luck of the draw or did I do >> > something stupid? (more precisely, what was it that I did stupid, >> > because that's inevitably true :~ >> No idea if it was a configuration error or if it broke somewhere. To be >> honest, I stopped caring about AP mode in old kernels completely, >> because of the large number of known bugs that are fixed in newer versions. >> >> - Felix > > ================================== > Original message: > > Hi, > > I'm a bit new to all of this - here's the short story. I would > appreciate any help available! > > I was building a WAP using a PC Engines ALIX board (AMD Geode LX800), > DNMA92 wireless card (Atheros 92xx) and Voyage Linux (based on Debian > Lenny, 2.6.30 kernel with hostapd 0.6.9 and libnl1.1). It was all > working but, the wireless card would drop connectivity for no apparent > reason after an undetermined time (minutes to hours). Stations (Win or > cellphone) would no longer see it at all, as if it was no longer > transmitting. > > The Voyage maintainer advised that ath9k isn't that stable at 2.6.30. > So, I upgraded the kernel to 2.6.32 using the Debian Lenny backport. I > then installed the "stable" compat-wireless, 2.6.32.16. > > Hostapd appears to start OK. However, the wireless SSID is only > visible to the Windows clients for a short time - seconds/minutes - > and they aren't able to connect. There were no errors in the hostapd > startup and it does its key rotations periodically. All that's logged > from hostapd -dd during the actual time that the Win clients can see > it and are trying to connect is: > 1278785975.350851: STA 00:13:02:be:7b:f9 sent probe request for broadcast SSID > 1278785975.353034: MGMT (TX callback) ACK > 1278785975.353073: mgmt::proberesp cb > (repeated a few times) > > The Win7 station used to connect automatically but now doesn't and > when I click "connect" it fails and shortly thereafter the SSID > disappears from the list. > > Nothing in /var/log. iw phy phy0 info and iw dev wlan0 info appear to > report sensible info as does iwconfig. > > During boot I do get messages about "VIA Padlock Hash Engine not detected" and > modprobe: FATAL: Error inserting padlock_sha > (/lib/modules/2.6.32-bpo.5- > 686/kern...): No such device > From googling it's not clear to me at least whether this is relevant. > > Suggestions? What further diagnostics could I do? > > Thanks, > Norm Henderson -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
