On 28 October 2013 12:20, Ben Greear <gree...@candelatech.com> wrote: > > On 10/28/2013 07:34 AM, Джонатан Вашингтон wrote: > > On 2013-09-19 17:01, James Hogan: > >> I think I have stumbled on to the real issue. I'm sorry if I have been > >> wasting your time at all. Apparently, the network is having problems > >> with all wireless-n mode cards. > >> The network doesn't like how wireless-n is roaming to the different AP's > >> and is registering as me disconnecting when the wireless card tries to > >> connect to roam to the other AP. > >> It's not just ath9k though, Ra-Link and other cards are having these > >> issues as well. Is there a way to disable or to make it possible to > >> disable the wireless-n mode of the card and just > >> use wireless abg? On a note, I noticed when I passed to the command line > >> "iw event -r" that at about the time my card would scan, that is when I > >> would stop receiving service even though > >> my cards never registered as being disconnected. After the scan, the > >> card just continues to reissue scans. Sometimes however, it will try to > >> reconnect me to the network. I was talking to one of the IT employees > >> in my class today and he said that they have been having this issue with > >> all of the newer higher-end cards. > > > > I've been having a similar problem, potentially the same. My > > university updated the firmware in their access points in August, and > > while I can still associate with the APs with no trouble, I can only > > get traffic through for short periods of time. As in James's case, I > > seem to lose service around when my card scans. > > > > I have been told by my university's wireless team that they are aware > > that this issue affects most linux users on campus, but don't > > currently have any solution. Their suggested work-around is to > > disable 11n speeds, which apparently "solves" the problem for most > > users. However, while iwconfig can be used to set various rates when > > connected to my 11g router at home, the speed seems to be locked at > > 65Mbps when I'm associated with the router on campus. That is, "sudo > > iwconfig wlan0 rate 54M" and similar directives seem to have no > > effect. > > > > For reference, I have an AR9285 card and so far have been using the > > ath9k driver that comes with my 3.2.0 debian stock kernel (which might > > technically be 3.2.41?). I have noticed nothing unusual about using > > any other networks, including 11n ones elsewhere. > > > > On 2013-09-21 05:40, Ben Greear wrote: > >> If you are using ath9k rate control, please try disabling that and use > > the minstrel-ht rate control instead. > >> > >> With a modern wpa_supplicant and kernel you can just disable HT in > > the supplicant config, by the way. > >> > >> > >> # disable_ht: Whether HT (802.11n) should be disabled. > >> # 0 = HT enabled (if AP supports it) > >> # 1 = HT disabled > >> # > >> # disable_ht40: Whether HT-40 (802.11n) should be disabled. > >> # 0 = HT-40 enabled (if AP supports it) > >> # 1 = HT-40 disabled > > > > I found I had an older version of wpa_supplicant (v1.0, which is what > > appears to be current in debian unstable) that did not support > > disable_ht and disable_ht40, so I downloaded a newer version (v2.0) > > and compiled it with CONFIG_HT_OVERRIDES enabled. Using that, I was > > able to add disable_ht=1 and disable_ht40=1 to the network block of my > > wpa_supplicant configuration file. > > > > When I run wpa_supplicant now in debug mode, it seems to parse these > > directives, and even outputs lines like "wlan0: set_disable_ht40: 1" > > where it seems to be telling the driver to disable the mode. However, > > iwconfig still indicates that the speed is 65Mbps, and no number of > > times manually telling it otherwise seems to have any impact on it. > > > > I don't know whether this limitation is due to an issue with the ath9k > > driver, with wpa_supplicant, or with something else, or is simply my > > misperception of how things should work. Either way, I'm tempted to > > apply the patch to my ath9k source that James posted so that I can > > disable 11n modes altogether, but I'm more than willing to try other > > suggestions. > > I think your kernel is too old to support that feature, it is relatively > recent. > > And, while I know it works with ath9k, I'm not sure it works with other > drivers. > > Thanks, > Ben > > -- > Ben Greear <gree...@candelatech.com> > Candela Technologies Inc http://www.candelatech.com >
On 28 October 2013 11:47, Sujith Manoharan <suj...@msujith.org> wrote: > Джонатан Вашингтон wrote: >> For reference, I have an AR9285 card and so far have been using the >> ath9k driver that comes with my 3.2.0 debian stock kernel (which might >> technically be 3.2.41?). I have noticed nothing unusual about using >> any other networks, including 11n ones elsewhere. > > 3.2 is really old... > >> I don't know whether this limitation is due to an issue with the ath9k >> driver, with wpa_supplicant, or with something else, or is simply my >> misperception of how things should work. Either way, I'm tempted to >> apply the patch to my ath9k source that James posted so that I can >> disable 11n modes altogether, but I'm more than willing to try other >> suggestions. > > Can you try the latest backports release, using just wpa_supplicant and > Network Manager disabled ? > > https://lkml.org/lkml/2013/10/27/149 > > Sujith With the driver from backports-3.10.17-1 (using my current kernel, which turns out to be from debian testing, not debian unstable), I'm able to adjust the speed manually, and disable 11n speeds in wpa_supplicant config files. However, as I found from other users who've encountered the same limitation at my university, I must actually entirely disable 11n speeds at the driver level. It has something to do with speeds presented as available during authentication, I guess? Anyway, the driver I compiled from backports doesn't seem to support this. Would the patch presented by James in September be a viable solution, or is there some reason for me to avoid it? -- Jonathan _______________________________________________ ath9k-devel mailing list ath9k-devel@lists.ath9k.org https://lists.ath9k.org/mailman/listinfo/ath9k-devel