Bugzilla – Bug 450
NetworkManager+ipw2200 cannot determine link quality
Last modified: 2005-10-04 14:27:16
You need to log in before you can comment on or make changes to this bug.
NetworkManager+ipw2200 cannot detect a signal from an access point. NetworkManager was released with FC3. It could end up being the premier app for managing network interfaces on Linux. See the current status of NetworkManager, waproamd, wpa_sullicant compatibility with ipw2200 at http://www.ces.clemson.edu/linux/auto_connect.shtml. In the NetworkManager mail list item http://mail.gnome.org/archives/networkmanager-list/2004-October/msg00020.html Dan Williams, the primary developer of Netmanager, is explaining what has to be done to correct this problem for the ipw2100 driver. Apparently, the problem was solved based on Dan's comments in http://verbum.org/blog/freesoftware/dan-williams-badass The ipw2200 has the same problem.
On the NetworkManager mailing list Dan Williams of Redhat claims that ipw2100 and ipw2200 scans to not report the strength of the signal from the AP. Another repondent says that problem has been fixed in the latest version of ipw2100.
Here is a discussion from the NetworkManager list. On Thu, 2004-12-09 at 10:08 +0100, Sebastien Estienne wrote: > On Thu, 09 Dec 2004 00:40:49 -0500, Dan Williams <dcbw redhat com> wrote: > > On Wed, 2004-12-08 at 23:20 -0500, Bill Moss wrote: > > > I have the Nov 17 CVS development distribution working with a fully > > > update installation of FC3 on a IBM T42 with version 0.16 of the ipw2200 > > > driver. So far it is working nicely on my home network which broadcasts > > > its SSID, except that the signal strength bar in the Info popup is > > > empty. I think this is an issue for the ipw2200 driver and I have > > > submitted a bug report. > > > > Yes, the ipw2200 drivers don't report _any_ signal strength for scanned > > access points, and the ipw2100 drivers don't either. What's worse, last > > I checked the ipw2100 drivers didn't do it for your current Access Point > > either. Driver bug. > > > Are you sure that it's still valid? because ipw2100 drivers are > version 1.0.1 since november 19. and the signal strength seems to work > for me for many months. > http://ipw2100.sourceforge.net/ Ok, I have this backwards and slightly wrong. The ipw2100 drivers _do_ support connected-AP-quality, but not scan quality. And the ipw2200 drivers have gotten connected-AP-quality in the last 2 months, but still don't have scan quality. In both cases: ipw2100_translate_scan(): #if 0 /* Add quality statistics */ /* TODO: Fix these values... */ iwe.cmd = IWEVQUAL; iwe.u.qual.qual = network->stats.signal; iwe.u.qual.level = network->stats.rssi; iwe.u.qual.noise = network->stats.noise; iwe.u.qual.updated = network->stats.mask & IEEE80211_STATMASK_WEMASK; start = iwe_stream_add_event(start, stop, &iwe, IW_EV_QUAL_LEN); #endif
Fixed in 0.20. We now use the IW_QUAL_LEVEL_INVALID, IW_QUAL_NOISE_INVALID, and IW_QUAL_QUAL_INVALID to mask out values that aren't being configured in the scan resutls. With newer wireless tools, this resolves the user problems that originally caused us to to not directly provide level and signal statistics in the standard way.
ipw2200 version 0.21 is providing NetworkManager with a scan signal strength/quality which is being displayed by the NetworkManager applet. The NetworkManager applet is displaying 72%. iwconfig is displaying 100%. If I run enough packets through the interface, iwconfig will momentarily drop to 90% in my current location, but at idle it aways reads 100%.
Sending back to scrub to see if more info testing is requested.
If you run some transfer tests (pings, scp, etc.) does your throughput indicate a quality of transfer closer to 90-100%, or to 72%? Do you have any insight from the NM folks on how it is computing the reported quality?
Hi Bill, Has the interoperability between recent versions of ipw2200 and NM been satisfactory?
I am running a fully updated FC4 with kernel-2.6.12-1.1398_FC4, ieee80211-1.0.3, ipw2200-1.0.6, and NetworkManager CVS HEAD 1.427, which is a few minor bug fixes past the version in Rawhide. All compatibility issues between NM and ipw2200 have been resolved. The principle NM developer, Dan Williams (dcbw@redhat.com), has published recommendations regarding WEXT (see http://www.ussg.iu.edu/hypermail/linux/kernel/0501.3/0667.html). I believe ipw2200 is currently inline with these recommendations. In my mind two issues remain that need to be considered by the wireless driver community at large. (1) A common model for converting signal level in dbm to signal quality as a percentage. In http://www.ces.clemson.edu/linux/dbm-rssi.shtml, I suggest letting 20% correspond to loss of association which would put the clean signal strength in the range of 1-10%. (2) A common method for a driver to report an AP that is hidden. NM was patched early on to handle the special way in which ipw2200 reports a hidden AP.
Issues seem to be resolved, per submitter. Marking as fixed.