Comment 8 for bug 675330

I've observed something similar, along with a serious regression between the 2.6.35-22-generic kernel and 2.6.35-27-generic.

When I boot with 2.6.35-22-generic, the wireless is mostly usable, and I can often fix it simply by turning off and then back on wireless networking in Network Manager.

With 2.6.35-27-generic, however, it is much worse. After a fresh boot, the network stops working when simply opening my web browser (which typically is restoring 30-40 tabs). And then toggling wireless doesn't seem to fix it, either.

When I have problems in 2.6.35-22-generic,I'd see trace like this in the system log:
Mar 11 12:23:26 purcell kernel: [ 465.663870] ===>rtl8192se_link_change():ieee->iw_mode is 2
Mar 11 12:23:26 purcell kernel: [ 465.720728] =====>rtl8192_set_chan()====ch:1
Mar 11 12:23:26 purcell kernel: [ 465.839173] =====>rtl8192_set_chan()====ch:2
Mar 11 12:23:26 purcell kernel: [ 465.950197] =====>rtl8192_set_chan()====ch:3
Mar 11 12:23:26 purcell kernel: [ 466.069926] =====>rtl8192_set_chan()====ch:4
Mar 11 12:23:26 purcell kernel: [ 466.188502] =====>rtl8192_set_chan()====ch:5
Mar 11 12:23:26 purcell kernel: [ 466.299549] =====>rtl8192_set_chan()====ch:6
Mar 11 12:23:26 purcell kernel: [ 466.418044] =====>rtl8192_set_chan()====ch:7
Mar 11 12:23:26 purcell kernel: [ 466.527829] =====>rtl8192_set_chan()====ch:8
Mar 11 12:23:26 purcell kernel: [ 466.637588] =====>rtl8192_set_chan()====ch:9
Mar 11 12:23:27 purcell kernel: [ 466.748674] =====>rtl8192_set_chan()====ch:10
Mar 11 12:23:27 purcell kernel: [ 466.867169] =====>rtl8192_set_chan()====ch:11
Mar 11 12:23:27 purcell kernel: [ 466.978247] =====>rtl8192_set_chan()====ch:12

but often the problem resolved itself reasonably quickly, so it was mostly usable.

With 2.6.35-27-generic, however, I also see lots of lines like this:
Mar 11 12:12:01 purcell kernel: [ 255.633584] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:02 purcell kernel: [ 256.628638] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:03 purcell kernel: [ 257.628675] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:04 purcell kernel: [ 258.636168] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:05 purcell kernel: [ 259.633734] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:06 purcell kernel: [ 260.631212] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:07 purcell kernel: [ 261.637029] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0

Sometimes the set_chan messages might be interspersed with the above like this:
Mar 11 12:11:59 purcell kernel: [ 253.637275] rtl819xSE:No more TX desc@1, ring->idx = 0,idx = 0, skblen = 0x44 queuelen=0
Mar 11 12:12:00 purcell kernel: [ 253.975369] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x18 queuelen=0
Mar 11 12:12:00 purcell kernel: [ 253.975380] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x18 queuelen=0
Mar 11 12:12:00 purcell kernel: [ 253.975389] ===>rtl8192se_link_change():ieee->iw_mode is 2
Mar 11 12:12:00 purcell kernel: [ 254.037687] =====>rtl8192_set_chan()====ch:1
Mar 11 12:12:00 purcell kernel: [ 254.047766] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.047774] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.154752] =====>rtl8192_set_chan()====ch:2
Mar 11 12:12:00 purcell kernel: [ 254.164822] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.164829] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.266940] =====>rtl8192_set_chan()====ch:3
Mar 11 12:12:00 purcell kernel: [ 254.277012] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.277022] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.384154] =====>rtl8192_set_chan()====ch:4
Mar 11 12:12:00 purcell kernel: [ 254.394227] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0
Mar 11 12:12:00 purcell kernel: [ 254.394235] rtl819xSE:No more TX desc@6, ring->idx = 0,idx = 0, skblen = 0x3e queuelen=0

At one point I tried getting and building a driver from the realtek site with -27, but I saw the same errors. Reverting to -22 is *much* better. The connection still drops sometimes (and I've not detected a pattern to it yet), but it is mostly usable. On -27, I find it unusable.

I'm running Maverick 10.10 on a Thinkpad T510.