Adjusted subject to more accurately describe the discussion.
Mike et. al, please let us know if this is going too far off
topic and we can take the discussion off-list.
<summary of first post by Samuel>
WISP on AAEON SBC with orinoco spewing:
Aug 1 15:34:50 10.50.1.10 kernel: netcs1: Tx error, status 1
messages. Client reporting slow connection.
</summary of first post>
[Homer Parker confirms same behavior]
<summary of my me-too report from Brad Fritz>
On my Bering RC2 firewall with Orinoco Silver card with nightly
syslog rotation:
# grep -c "eth5: Tx error, status 1 (FID=" /var/log/syslog
2187
My setup is:
Linux Kernel Card Services 3.1.22
Vadem VG-469 ISA-to-PCMCIA at port 0x3e2 ofs 0x00
pcmcia-cs 3.1.33
Lucent/Orinoco/Agere Silver card: manfid: 0x0156, 0x0002
orinoco.c, orinoco_cs.c v0.09b
Lucent/Agere firmware version 6.16 (if reported correctly,
might be 8.10)
Using WEP in Ad-Hoc Mode
Communicating with one identical card using firmware v8.10 and
orinoco.c 0.11b with kernel pcmcia drivers in 2.4.19-pre10
and a second SMC 2632W using the prism2_cs.o module in
wlan-ng 0.1.13.
</summary of my me-too report from Brad Fritz>
On Fri, 02 Aug 2002 13:43:50 PDT Brock wrote:
> What a strange coincidence! I've been running Bering RC2 for several
> months now with Lucent Silver cards without any problems. I recently
> built a similar box for a friend who has two wireless clients in his
> house and immediate neighborhood. The ones in the house have been
> working fine (the box has not been rebooted since it went in a month
> ago) and I finally got around to dealing with the more distant client
> last night, using external antennas at the Bering box and remote client.
>
> I established a path with my laptop and an antenna, then spent some time
> getting a Raylink PCI-PCMCIA adapter to go in the intended client. (if
> anyone cares, the windows driver in '98 is broken - you need the raylink
> version installed before the card goes in and Windows tries to install
> its own!) This card uses the TI chipset (1211?) and has been used in a
> Bering box previously without issue. The path is reasonable, given that
> I've used far worse on really long hops! Orinoco was reporting an
> 11mbit connection, sometimes dropping to 5.5 in one direction only.
>
> When I finally got the client up and running, the connection looked good
> - nice and fast to his msn.com home page (not my first choice for a home
> page, but who am I to judge ;-). But after that, the performance
> dropped through the floor. The same errors you guys have reported
> showed up on the Bering box. However, swapping the wireless card and
> using the same antenna, my laptop could surf just fine, using either
> Linux or Win98!
>
> I don't know how, or even if it's possible, but it seems like the client
> is causing the error. All the cards involved have been upgraded to
> 8.10. With two completely separate installs, I have never seen this
> error until last night, with this latest machine. It's an Athlon 1400
> running Win98SE, with the Raylink PCI-PCMCIA adapter. I also noticed
> that pinging the client from the Bering box (or the other way for that
> matter) resulted in packet losses of 10-15%. I didn't have time to try
> this with my laptop at the same location to compare (I'm kicking myself
> now!).
>
> I won't be able to go back and poke around again for about a week
> unfortunately. As it was an entirely unexpected problem, discovered
> when I should really have been asleep in bed, I'm afraid I didn't
> trouble shoot it in a really logical way!
>
> Does anything I've reported trigger anything with you guys?
Yes, this jives with my experience. I just verified the SMC
prism2 card is the one triggering the Tx errors on the Bering
firewall with orinoco.c 0.09b. I ran two flood pings:
orinoco -> orinoco on Bering:
ping -c 2000 -f 192.168.21.11
2171 packets transmitted, 2000 packets received, 7% packet loss
round-trip min/avg/max = 2.3/122.9/196.7 ms
SMC prism2 -> orinoco on Bering:
ping -c 2000 -f 192.168.21.11
4172 packets transmitted, 2000 packets received, 52% packet loss
round-trip min/avg/max = 4.4/4867.8/7678.9 ms
During the SMC flood ping, syslog on the firewall went crazy
with these messages:
Aug 2 21:09:56 firewall kernel: eth5: Tx error, status 1 (FID=03EA)
Aug 2 21:09:56 firewall kernel: eth5: Tx error, status 1 (FID=034B)
Aug 2 21:09:56 firewall kernel: eth5: Tx error, status 1 (FID=0519)
Here are a few relevent postings from the wireless list I cited
in my last message:
http://lists.samba.org/pipermail/wireless/2002-January/003648.html
says it's a known problem in 0.09b related to waking up the Tx
queue less aggressively than in pre 0.09 drivers. No mention of
a fix.
http://lists.samba.org/pipermail/wireless/2002-February/003814.html
starts a very technical thread that seems to overlap with the
symptoms we are seeing.
http://lists.samba.org/pipermail/wireless/2002-February/003826.html
starts a long thread focusing on StrongARM patches but also
touches on TX queue problems, including an explaination of
"status 1".
I didn't see anything about a resolution to the Tx queue problems,
but I only scanned the mbox archive quickly. It might be worth
trying v0.11b or v0.12 to see if that fixes the problem. I will
probably do so eventually, but I'm not using the SMC card much
so it's not a priority at the moment.
Homer and Samuel, can you tell if the Tx errors are caused by
a non-orinoco peer to the orinoco card?
If orinoco*.c v0.11b or v0.12 does not fix the problem, it might
be a good idea to collect our configuration information and submit
it to the wireless list. David Gibson or another one of the
driver hackers there might be able to use it to fix the Tx queue.
--Brad
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
------------------------------------------------------------------------
leaf-user mailing list: [EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/leaf-user
SR FAQ: http://leaf-project.org/pub/doc/docmanager/docid_1891.html
