Re: [Acx100-devel] [PATCH 00/20] another batch

[Jim C. <jim...@gm...>]

On Tue, May 15, 2012 at 11:36 AM, Oliver Winker <oli...@ol...> wrote:
> Hi Jim, Larry,
>
> Alright, Thanks again for the new batches ;)!
>
> I was held up by an problem encountered while checking the latest
> OpenWRT trunk, which is related to a change in the mac80211 hw-queue
> handling in the current wireless-testing (and compat-wireless). I was
> checking different possibilities, but I think it's concluded to one now.
>
> So as next step I'll merge then in all the batches and do the usual
> testing with the devices here.
>
> Maybe I get this done tomorrow (maybe), latest Thu then (there is an
> public holiday). But I'm on it!
>
> Best Regards, Oliver
>


FYI,  several things


I recently rebuilt my kernel to add ieee80211 debugfs support,
and (I think) afterwards started seeing these periodically.


ieee80211 phy0: wlan0: No probe response from AP 00:14:d1:e8:65:0a
after 500ms, try 1/5
ieee80211 phy0: wlan0: No probe response from AP 00:14:d1:e8:65:0a
after 500ms, try 1/5
ieee80211 phy0: wlan0: No probe response from AP 00:14:d1:e8:65:0a
after 500ms, try 1/5
ieee80211 phy0: wlan0: No probe response from AP 00:14:d1:e8:65:0a
after 500ms, try 1/5
ieee80211 phy9: wlan1: No probe response from AP 00:14:d1:e8:65:0a
after 500ms, try 1/5
ieee80211 phy0: wlan0: No probe response from AP 00:14:d1:e8:65:0a
after 500ms, try 1/5

Im getting them for both wlan0 (rtl8180) and wlan1 (acx-mac80211)
which pretty much excuses the driver.

More suspiciously, Im getting tail errors:


acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=2, tail=2, head=9
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=11,
tail=11, head=1
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=10,
tail=10, head=2
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=15,
tail=15, head=4
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=14,
tail=14, head=5
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=0, tail=0, head=9
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=7,
tail=7, head=14
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=0,
tail=0, head=12
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=12,
tail=12, head=1
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=7, tail=7, head=8
acx.acx_tx_clean_txdesc: clean_txdesc: tail isn't free. finger=8,
tail=8, head=13


In all cases (I think), finger==tail, which suggests an off-by-one error,
probably due to merge, but I havent looked yet.

Iperf testing works, as does scp of a large tarball, checksums match,
so its affecting thruput at worst.


there are a bunch of retries though:

root@voyage:~# iw dev wlan1 station dump
Failed to read classid file: Object not found
Station 00:14:d1:e8:65:0a (on wlan1)
	inactive time:	14740 ms
	rx bytes:	61152680
	rx packets:	276428
	tx bytes:	72244666
	tx packets:	65471
	tx retries:	18447
	tx failed:	2
	tx bitrate:	54.0 MBit/s


root@voyage:~# more /dbg/acx_mac80211/phy10/info
acx.acx_dbgfs_open: opening filename=info fmode=35 fidx=0 adev=c0bd2fc0
acx driver version:		v0.6.0-glarry-512-19-g0fde744 (git: larry-512-19
-g0fde744)
Wireless extension version:	22
chip name:			TNETW1130 (0x03010101)
radio type:			0x16
form factor:			0x01
EEPROM version:			0x05
firmware version:		Rev 1.2.1.34 (0x01020134)

larry-512 is the (mis-dated) tag I applied to Larrys 15 on 509


Im attaching a test script Ive been using to bang on the driver
Dot it into your shell, and go from there.

root@voyage:~# . wifi-probe.rc
wifi-help for info
root@voyage:~# wifi-help

provides functions to probe, test wifi drivers using iw-tools
   wifi-probe:	run the probes/tests
   mon-up:	create monitor ifaces, do tcpdump captures on them
   mon-down:	close those monitor interfaces
   wifi-save:	save the files to to driver-version dir
   wifi-net:	run some connectivity tests (ifdowns eth0 wlan0)
vars: # set to change behavior
   devs:	wlan0 wlan1
   drvname:	acx
   apmac:	00:14:d1:e8:65:0a		# not used now
   chan:	11
   macs:	00:18:DE:06:B2:0C 00:30:BD:4D:E2:47 00:18:4D:24:B3:FA
   iperfsvrs:	192.168.10.127 192.168.10.128	# iperf server addresses
   iperfargs:	-t30 -i5
   others:	eth0 wlan0		# ifdown these in wifi-ent


Please suggests mods, enhancements useful to you.
2 Ive just thought of:
- snapshot of /dbg/acx-mac80211/phyX
- snapshots of /dbg/ieee80211/phyX

Maybe commit this to the tree ?