Re: regression on BWU613SB

Hi,

On Thu, 8 Apr 2004, The Anarcat wrote:

> I've noticed a regression in the support for the chipset inside my
> BWU613SB USB wifi. I got it to work in January and the modifications I
> suggested got committed into the main tree by around then, however, I
> never checked to see if the CVS code was actually working.
>
> One thing is for sure now, it's not working anymore. The symptom is that
> when plugging the device in, gkrellm crashes, ifconfig segfaults, as cat
> /proc/net/wireless. I appended a typical crash dump below.
>
> I've tried to pinpoint the date at which the regression got into the
> tree, so I tried the following dates, in order:
>
> 1 feb: works
> 1 mar: crash
> 15 feb: works
> 25 feb: crash
> 20 feb: works

> 23 feb: crash while compiling (!)
IMHO hard to blame the driver for this (unless it was loaded before
compilation time).

There were only a few updates to at76c503.c during that time
(from "cvs log at76c503.c"):

----------------------------
revision 1.44
date: 2004/03/16 20:15:44;  author: jal2;  state: Exp;  lines: +17 -3
- treat ' ' as a cloaked SSID, too
- rx_mgmt_beacon: write \0 after copying the SSID
----------------------------
revision 1.43
date: 2004/02/20 22:14:42;  author: jal2;  state: Exp;  lines: +2 -10
branches:  1.43.2;
0.12beta7: un-done the workaround for 2.4.x kernels and usb-uhci calling
disconn
ect during resetting the device (it didn't work)
----------------------------
revision 1.42
date: 2004/01/10 20:31:17;  author: jal2;  state: Exp;  lines: +24 -10
version 0.12beta5:
- fixed bug in at76c503.c:at76c503_delete_device(), which called
unregister_netdev() even if netdev wasn't registered before
- usb-uhci of 2.4.23/24 calls the disconnect() after we resetted the
device
  try to work around this
----------------------------

version 1.43 removed a work-around for usb-uhci on 2.4.x problems which
didn't work here (but I guess it worked for you :-|

> By then the system got a little unstable, I got a little nervous and
> stopped testing. :) So it's probably around the end of february.
>
> However, there might be a deeper problem... When trying to get a known
> version to work, I can't find my base station, and what's worse, there's
> 2 wlan extensions listed in iwconfig (!!!). This is really unusual, and
> I never saw this behavior in january. However, I re-installed the
> machine in between, using the knoppix installer (yes, I know, pure evil)
> so there might be something a little wrong on this box.

Even your syslog below shows a second wlan1 interface.

> I'm a little hesitant to try to debug this on an old version, so I'd
> rather wait for input on how to solve the first problem with the current
> tree (segfaulting ifconfig on connection) and the attack the second
> problem rather than trying to fix problems with the old code.

I'd suggest trying the latest cvs version and take uhci instead of
usb-uhci (if possible). Maybe your re-installing of the system swapped
uhci for usb-uhci?
The current cvs has a problem with 2.4.x and usb-uhci due to different
behaviour of the USB subsys at USB device reset calls (but it works with
uhci, ohci and ehci).

/Jörg