#21284: EA3500 (Audi Kirkwood) wifi unstable.
------------------------+---------------------------------------
  Reporter:  anonymous  |      Owner:  developers
      Type:  defect     |     Status:  new
  Priority:  normal     |  Milestone:  Designated Driver (Trunk)
 Component:  kernel     |    Version:  Trunk
Resolution:             |   Keywords:
------------------------+---------------------------------------

Comment (by thx1111):

 I am seeing a problem with the EA3500 which might be related or have the
 same underlying cause.  For what it's worth, I am looking at this from a
 different perspective, so might be useful.  I have several laptops that
 associate with the EA3500 which seem to work normally, but then I have
 one, a Toshiba Satellite running a Realtek RTL8188CE using the rtl8192ce
 and nl80211 drivers, which reacts badly.  I note that the dmesg log has
 [   11.638789] ieee80211 phy0: Selected rate control algorithm 'rtl_rc'
 and I'm not sure where that rate control is operating, in the firmware, or
 in some kernel module, or in userspace.

 The problem is easiest to see running ping from a machine with a wire
 connection to the access point, pinging the laptop.  Every four to eight
 seconds, the ping response hangs for about four seconds, and then all the
 ping replies are returned in a burst.  No pings are dropped.  On the
 access point, "iw wlan0 station dump" shows "tx retries:     0" and "tx
 failed:      0".  Running tcpdump on the laptop, it seems to respond
 immediately whenever it receives an echo request.  It just does not get
 the request during the "hangs".  But running tcpdump on the access point,
 the echo requests are seen passing through wlan0 at regular intervals.  I
 have not looked at the actual RF, so I don't know if the packets are
 really going over the air or not.  Still, it is as if there were a buffer
 of echo requests filling-up at the access point, and then "dumping" after
 four seconds or so.

 Now, it seems there is an effective and inexplicable workaround for this
 periodic "hang".  If I run ping on the *laptop*, and at the same time ping
 from the remote machine, the ping hangs at the remote machine completely
 disappear, and all the ping times average around 1ms.

 The EA3500 is running the Marvell 88W8366 Integrated MAC/Baseband MIMO SoC
 with the 88W8063 MIMO RF Transceiver, and the driver is the mwl8k, which
 then depends upon mac80211 and cfg80211, and then upon the compat module.
 The 88W8366 firmware does its own rate control.  It is hard to say if the
 "problem" is in the access point or in the laptop, since they mostly act
 badly together.  But I did try the same ping test using the older Belkin
 N1, "F5D8231-4 v2" radio, which also has a Marvel chip set, the 88W8361,
 but running their own driver.  There is a similar effect, except that the
 "hangs" are all on the order of about 900ms instead of 4000ms, and still
 completely disappear when running ping from the laptop.  Perhaps there is
 some common problem in the Marvell firmware?

 I hope all that might spark some ideas, since my workaround is such a
 hack.

--
Ticket URL: <https://dev.openwrt.org/ticket/21284#comment:41>
OpenWrt <http://openwrt.org>
Opensource Wireless Router Technology
_______________________________________________
openwrt-tickets mailing list
openwrt-tickets@lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-tickets

Reply via email to