All, I have debug captures (with various dbg options) of 3 different instances. As near as I can tell, ANI, calibration, etc. look to all be reactive to the root cause. The cycle seems to be rapid resets and calibrations, then suddenly a chain is deaf. It's like AGC or similar suddenly dropped way down. Then ANI OFDM errors drop down, OFDM level is stair-stepped down from 8/9 to 2 in ~8 seconds. Everything then appears great, except we can't hear some of the signals (2nd chain only can hear) and nothing changes, it's stuck.
Does this behavior ring a bell for anyone? https://drive.google.com/folderview?id=0B2bEy75HhwWhYk5ZWEdhakVtZzA&usp=sharing Regards, Joe On Mon, Oct 12, 2015 at 9:22 PM, Joe Ayers <j...@ayerscasa.com> wrote: > Adrian, > > Updated debug data. I think I captured ANI debug data over the exact > time period when the receive chain went deaf. > > At the following link are 3 snapshots (note syslog history going back ~10 > minutes so some overlap): > > ani-2015-10-12-20-41-42.tar.gz: period of time when it is going deaf > ani-2015-10-12-20-45-01.tar.gz: it is deaf (neighbor-station dropped out) > ani-2015-10-12-20-52-21.tar.gz: after an "iw wlan0 scan" reset and links > had been restored. > > > https://drive.google.com/folderview?id=0B2bEy75HhwWhYk5ZWEdhakVtZzA&usp=sharing > > See station_dump.txt of last data snapshot to see what the RSSI of the > neighbor nodes should be. > > Regards, > Joe AE6XE > > > On Mon, Oct 12, 2015 at 10:20 AM, Joe Ayers <j...@ayerscasa.com> wrote: > >> Adrian, >> >> Here's a link to the debug data below. Still a little green to do this >> debugging, so hopefully didn't do something silly. I did the following >> thinking it correctly turns on the right debug option needed: "echo >> '0x00000040' > debug" in the ANI debugfs file. If additional debug options >> would be helpful, please advise. >> >> >> https://drive.google.com/file/d/0B2bEy75HhwWhM3JYS2N2WXlUMDg/view?usp=sharing >> >> There are several snapshots generally 5 minutes apart of the ANI debugfs >> files leading up to the point that the H-pol chain is deaf. Refer to >> "scan.log" which has: >> >> 1) 'iw wlan0 station dump' when deaf state showing RSSI >> 2) 'iw wlan0 scan' to trigger reset >> 3) 'iw wlan0 station dump' instantly back to normal RSSI >> >> Timestamps inserted to correlate with ANI debugfs files. >> >> Joe AE6XE >> http://www.aredn.org >> >> >> On Sat, Oct 10, 2015 at 6:07 PM, Adrian Chadd <adr...@freebsd.org> wrote: >> >>> Hi, >>> >>> Can you turn on ANI debugging in ath9k and capture what it's doing? >>> >>> I've seen this happen on some other outdoor stuff with AR9380 but I >>> think their setup is more "the interference is terrible". But yes, I'd >>> like to debug ANI more. :) >>> >>> >>> >>> -adrian >>> >>> >>> On 8 October 2015 at 21:13, Joe Ayers <j...@ayerscasa.com> wrote: >>> > We (http://www.aredn.org ) are diagnosing a problem with >>> > compat-wireless-2014-05-22 for ath9k (under OpenWRT BB) that has >>> appeared, >>> > but does not exist in compat-wireless-2012-09-07 (under OpenWRT AA). >>> > >>> > While we dig through the commits and learn more about wireless drivers >>> to >>> > isolate the issue, would appreciate if someone might recognize the >>> issue >>> > and/or suggest a better path. >>> > >>> > Here's a summary of the config and symptoms: >>> > >>> > * Ubiquiti Rocket M5 XM with AR9280 on ch 139, 171 (we've taken out >>> our code >>> > to extend up to ch 184 to isolate :) ). >>> > * Reproducable on 5Mhz, 10Mhz, and 20Mhz channel widths >>> > * Ad-hoc mode with long distance links and OLSR on top. >>> > * In one configuration, the receive becomes deaf on one chain where all >>> > signals (if still showing) are attenuated down to near noise floor. >>> > * Only a strong (+10dB above the others) MIMO dual polarity incoming >>> signal >>> > stays working on 2nd chain, which is not attenuated from normal. >>> > * It's stuck. running an 'iw wlan0 scan' causes reset and all the >>> stations >>> > instantly connect back to normal operation. >>> > * Can occur multiple times per day. Readily reproducible and have >>> seen in >>> > multiple locations. >>> > * One example yesterday at a location where all stations dropped out (4 >>> > stations within ~2dB signal strengths of each other). Scan restored >>> back >>> > to normal. >>> > * distances are typically 5 to 15 miles links. >>> > >>> > We're speculating that the immunity level is getting stuck at high >>> threshold >>> > and digging around in the commits accordingly for the moment. >>> > >>> > Regards, >>> > Joe AE6XE >>> > >>> > _______________________________________________ >>> > ath9k-devel mailing list >>> > ath9k-devel@lists.ath9k.org >>> > https://lists.ath9k.org/mailman/listinfo/ath9k-devel >>> > >>> >> >> >
_______________________________________________ ath9k-devel mailing list ath9k-devel@lists.ath9k.org https://lists.ath9k.org/mailman/listinfo/ath9k-devel
