I think I probably see the same thing on RT2860, but you've got further tracking down what's happening than me (my debugging is hampered by the AP being about 2 hour's drive away..)
In gmane.os.openbsd.misc, you wrote: > Hi, > > I 'm running "OpenBSD 4.4-current (RALDBG) #0: Fri Oct 10 16:56:50 CEST > 2008", which is GENERIC with RAL_DEBUG, but I've seen this problem with > previous kernels and without RAL_DEBUG, too. > > # dmesg | grep ral > ral0 at pci0 dev 14 function 0 "Ralink RT2860" rev 0x00: irq 10EEPROM > rev=1, FAE=1 > ral0: MAC/BBP RT2860 (rev 0x0101), RF RT2820 (2T3R) > > This is a pci Edimax EW-7728IN, which I believe is the same card that was > donated to damien@ (?) and that led to 28xx support. > > After an unfixed amount of time, from a few minutes up to a few days, the > interface simply stops respoding to probe requests: > > # tcpdump -nvvveeees 1000 -i ral0 -y IEEE802_11_RADIO not subtype beacon > 14:17:40.761912 CLI1-MAC > ff:ff:ff:ff:ff:ff, bssid ff:ff:ff:ff:ff:ff (seq > 16): 802.11: probe request, <radiotap v0, 1Mbit/s, chan 6, 11g, sig -19dBm, > antenna 2, signal 17dB> > 14:17:40.963338 CLI1-MAC > ff:ff:ff:ff:ff:ff, bssid ff:ff:ff:ff:ff:ff (seq > 32): 802.11: probe request, <radiotap v0, 1Mbit/s, chan 6, 11g, sig -17dBm, > antenna 2, signal 15dB> > 14:21:03.860025 CLI2-MAC > ff:ff:ff:ff:ff:ff, bssid ff:ff:ff:ff:ff:ff (seq > 1120): 802.11: probe request, <radiotap v0, 1Mbit/s, chan 6, 11g, sig > -27dBm, antenna 1, signal 25dB> > 14:21:04.306901 CLI2-MAC > ff:ff:ff:ff:ff:ff, bssid ff:ff:ff:ff:ff:ff (seq > 1520): 802.11: probe request, <radiotap v0, 1Mbit/s, chan 6, 11g, sig > -23dBm, antenna 1, signal 21dB> > > Whereas normally you'd see the probe req, probe resp, auth req, auth resp, > assoc req, assoc resp, wpa dance. > > # tcpdump -nvvveeees 1000 -i ral0 -y IEEE802_11_RADIO | grep beacon | grep > AP-MAC > Shows that it stops sending beacon frames. It's still picking up the > beacons from the 5 other wlans it can see, so rx seems to work fine. > > # ifconfig ral0 down && ifconfig ral0 up > Fixes everything, until it happens again after a seemingly random interval. > The kernel doesn't log anything unusual even with RAL_DEBUG. > > I suppose I should sendbug, but I think lots of people have these cards so > I'd like to know if anyone else is seeing this. Any ideas? > > Thanks and please cc, > > bbee