[zd1211-devs] dhclient causes the 1.6.0.0 driver to disconnect / zd1205_close weirdness

[Mayne <zd1211.sf.net@co...>]

On Wed, 24 Aug 2005, Mayne wrote:

> The 2.0.0.0 driver's stable so far apart from some weird association
> problems sometimes (which are probably due to the missing firmware
> download patch - I'm testing that currently).

The problem seems to be that dhclient does an ifconfig down/ifconfig up
sometimes. 1.6.0.0 [1] introduced a new "macp->bAssoc = 0" in line
4430/zd1205_close which seems to break everything once an ifconfig
down/ifconfig up was issued (the driver does not associate again when
setting an essid, it does not even seem to scan, too).

I think that's something severely broken in the driver. The kernel's
net_open and net_close methods seem not to be intended to do more or less
everything concerning driver initialization/deinitialization (see for
example zd1201.c for the ZD1201 chipset included in the 2.6.12 kernels:
There, net_stop does a netif_stop_queue, that's all). The zd1211 driver,
however, completely removes all hooks and data structures on an ifconfig
down (net_stop/zd1205_close).

If course, for the time being one can comment out the line mentioned
above. However, obviously this is quite a hack.

Regards

[1] http://zd1211.ath.cx/changeset/7

[zd1211-devs] Source Bundle Released

[Markus KARG <markus.karg@gm...>]

Hi all,

today I published the source bundle assembled by Roman KOROVKIN. Feel free to 
download from sourceforge and send email to the list or add items to the 
trackers. :-)

Have Fun
Markus

Re: [zd1211-devs] Source Bundle Released

[Roman K <roma_rik@ma...>]

> Hi all,
> 
> today I published the source bundle assembled by Roman KOROVKIN. Feel free to 

To Markus: In fact, Kurakin. Korovkin is the last name of Dmitry.
You were involved by mysterious 'K.' Anyway, thank you.

> download from sourceforge and send email to the list or add items to the 
> trackers. :-)

For all others. This tarball is the current state of CVS.
If you do not like to use CVS version, now it's time to
try what was done since last release and if it solve your
problems.
If you'll send a request to the trackers, don't forget
to mention version of the code you are using.

sagamore

> Have Fun
> Markus

[zd1211-devs] Re: Source Bundle Released

[Albert Rybalkin <albertr@ir...>]

Roman K writes: 

> For all others. This tarball is the current state of CVS.
> If you do not like to use CVS version, now it's time to
> try what was done since last release and if it solve your
> problems.
> If you'll send a request to the trackers, don't forget
> to mention version of the code you are using. 
> 

Guys, 

Not to rain on your parade, but this code is way obsolete. Everyone
must be using at least 1.2.x (prior versions had shared-key WEP auth
broken), and 2.0.0.0 was released recently as you all know.
The newer ZyDAS versions have many improvements and bug fixes, there's
no point to use the obsolete code from CVS. Why don't you concentrate
on merging 2.0.0.0 into your CVS tree instead?
 -albertr

Re: [zd1211-devs] Re: Source Bundle Released

[Mayne <zd1211.sf.net@co...>]

On Tue, 23 Aug 2005, Albert Rybalkin wrote:

> The newer ZyDAS versions have many improvements and bug fixes, there's
> no point to use the obsolete code from CVS. Why don't you concentrate
> on merging 2.0.0.0 into your CVS tree instead?

Regarding the subversion repository, I merged most non-endian patches. The
subversion repository is based on 2.0.0.0. I'll add further patches this
weekend. Trac's coming, too.

Regards

[zd1211-devs] Re[2]: Source Bundle Released

[Roman K <roma_rik@ma...>]

> Roman K writes: 
> 
> > For all others. This tarball is the current state of CVS.
> > If you do not like to use CVS version, now it's time to
> > try what was done since last release and if it solve your
> > problems.
> > If you'll send a request to the trackers, don't forget
> > to mention version of the code you are using. 
> > 
> 
> Guys, 
> 
> Not to rain on your parade, but this code is way obsolete. Everyone
> must be using at least 1.2.x (prior versions had shared-key WEP auth
> broken), and 2.0.0.0 was released recently as you all know.
> The newer ZyDAS versions have many improvements and bug fixes, there's
> no point to use the obsolete code from CVS. Why don't you concentrate
> on merging 2.0.0.0 into your CVS tree instead?

It is less obsolete than was released before. Only that.

About 2.0.0.0 I've heard from some users that it is more broken in some
cases. So if you do not need WEP our version is more stable.

rik

>  -albertr

[zd1211-devs] Re: Source Bundle Released

[Mayne <zd1211.sf.net@co...>]

On Wed, 24 Aug 2005, Roman K wrote:

> About 2.0.0.0 I've heard from some users that it is more broken in some
> cases. So if you do not need WEP our version is more stable.

I don't need WEP and the CVS driver definitely produced kernel panics here
every two or three days. I use the stick for my 24/7 internet connection.
The 2.0.0.0 driver's stable so far apart from some weird association
problems sometimes (which are probably due to the missing firmware
download patch - I'm testing that currently).

Regards

[zd1211-devs] Re: (buggy?) Source Bundle Released

[Mayne <zd1211.sf.net@co...>]

On Wed, 24 Aug 2005, Mayne wrote:

> I don't need WEP and the CVS driver definitely produced kernel panics here
> every two or three days.

Without patch 1244579 [1] (which was submitted a month ago but hasn't made
it to CVS) you get kernel panics due to a race condition every few hours.
This bug is known at ZyDAS and was fixed in version 1.2.0.0 of the ZyDAS
drivers (or perhaps even earlier, didn't check that).

May I ask what the checks done on the CVS driver are? Does anybody use the
CVS driver on a daily basis, perhaps running some stress tests? The kernel
panics mentioned above I get with only a ping running... (but with quite
heavy traffic on air by other machines connecting to the same AP)

Regards

[1]
http://sourceforge.net/tracker/index.php?func=detail&aid=1244579&group_id=129083&atid=713756

[zd1211-devs] Re: Source Bundle Released

[Albert Rybalkin <albertr@ir...>]

What is the eerpom version of your USB dongle? 

 -albertr 


Mayne writes: 

> On Wed, 24 Aug 2005, Roman K wrote: 
> 
>> About 2.0.0.0 I've heard from some users that it is more broken in some
>> cases. So if you do not need WEP our version is more stable.
> 
> I don't need WEP and the CVS driver definitely produced kernel panics here
> every two or three days. I use the stick for my 24/7 internet connection.
> The 2.0.0.0 driver's stable so far apart from some weird association
> problems sometimes (which are probably due to the missing firmware
> download patch - I'm testing that currently). 
> 
> Regards 
> 
> 
> -------------------------------------------------------
> SF.Net email is Sponsored by the Better Software Conference & EXPO
> September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
> Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
> Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
> _______________________________________________
> Zd1211-devs mailing list
> Zd1211-devs@...
> https://lists.sourceforge.net/lists/listinfo/zd1211-devs 
> 
 

[zd1211-devs] dhclient causes the 1.6.0.0 driver to disconnect / zd1205_close weirdness

[Mayne <zd1211.sf.net@co...>]

On Wed, 24 Aug 2005, Mayne wrote:

> The 2.0.0.0 driver's stable so far apart from some weird association
> problems sometimes (which are probably due to the missing firmware
> download patch - I'm testing that currently).

The problem seems to be that dhclient does an ifconfig down/ifconfig up
sometimes. 1.6.0.0 [1] introduced a new "macp->bAssoc = 0" in line
4430/zd1205_close which seems to break everything once an ifconfig
down/ifconfig up was issued (the driver does not associate again when
setting an essid, it does not even seem to scan, too).

I think that's something severely broken in the driver. The kernel's
net_open and net_close methods seem not to be intended to do more or less
everything concerning driver initialization/deinitialization (see for
example zd1201.c for the ZD1201 chipset included in the 2.6.12 kernels:
There, net_stop does a netif_stop_queue, that's all). The zd1211 driver,
however, completely removes all hooks and data structures on an ifconfig
down (net_stop/zd1205_close).

If course, for the time being one can comment out the line mentioned
above. However, obviously this is quite a hack.

Regards

[1] http://zd1211.ath.cx/changeset/7