ndiswrapper

On Sun, 2010-01-24 at 10:09 +0100, sacarde wrote:
> hi,
>   I have a usb-key wireless USrobotics 5421
> 
> in archlinux 32bit following this:  
> http://sourceforge.net/apps/mediawiki/ndiswrapper/index.php?title=US_Robotics_USR805421
> all works OK
> 
> I have problem installing it on archlinux 64bit
> I use this: 
> http://www.usr.com/support/5421a/5421a-files/5421-v6.01.19.10x64.exe
> 
> but I find only .inf file
> and no .sys

That's a Vista driver (NDIS 6).  ndiswrapper only works with Windows XP
and older drivers.

> can you help me?

No.  I suggest that you use the rndis_wlan driver from the Linux kernel.

-- 
Regards,
Pavel Roskin

On Tue, 2010-01-19 at 13:24 -0600, Raymond Norton wrote:
> I have Ubuntu running on a G3 Powerpc, but cannot get the wireless to 
> work. Will Ndiswrapper compile and install on a Powerpc?

No.  Only i386 and x86_64 architectures are supported.

-- 
Regards,
Pavel Roskin

On Wed, 2009-10-21 at 14:24 +0800, bobl wrote:
> Hi, All,
> 
> In the "main page", i see follow line:
> -------------------------------------------------------
> other devices are known to work: e.g., ethernet cards,
> -------------------------------------------------------
> 
> I want to know, how many type's of such cards support current time, i
> try to find, only find:
> 
> 
>   3Com 3C410

I you find that some other card is working, please feel free to report.

-- 
Regards,
Pavel Roskin

Pavel Roskin 写道:
> On Wed, 2009-10-21 at 14:24 +0800, bobl wrote:
>   
>> Hi, All,
>>
>> In the "main page", i see follow line:
>> -------------------------------------------------------
>> other devices are known to work: e.g., ethernet cards,
>> -------------------------------------------------------
>>
>> I want to know, how many type's of such cards support current time, i
>> try to find, only find:
>>
>>
>>   3Com 3C410
>>     
>
> I you find that some other card is working, please feel free to report.
>
>   
Thanks first!

But my means is want to confirm whether ndiswrapper can support normal
ethernet cards(not wireless)?

We want to setup a team to focus on this issue, i just do some
pre-resarch for this, so i send this mail to the maillist.

Can you give me a confirm?

Thanks!

Bob

On Thu, 2009-10-22 at 10:05 +0800, bobl wrote:

> But my means is want to confirm whether ndiswrapper can support normal
> ethernet cards(not wireless)?

I haven't seen any reports in the list that it does or that it doesn't.

> We want to setup a team to focus on this issue, i just do some
> pre-resarch for this, so i send this mail to the maillist.
> 
> Can you give me a confirm?

I have no reason to distrust the web page.

-- 
Regards,
Pavel Roskin

Pavel Roskin 写道:
> On Thu, 2009-10-22 at 10:05 +0800, bobl wrote:
>
>   
>> But my means is want to confirm whether ndiswrapper can support normal
>> ethernet cards(not wireless)?
>>     
>
> I haven't seen any reports in the list that it does or that it doesn't.
>
>   
>> We want to setup a team to focus on this issue, i just do some
>> pre-resarch for this, so i send this mail to the maillist.
>>
>> Can you give me a confirm?
>>     
>
> I have no reason to distrust the web page.
>
>   
Thanks! We just test Broadcom BCM5751 under ndiswrapper, "ping" is ok,
but can't use ssh and firefox...

Anyway, thanks.

Bob

--- On Fri, 23/10/09, bobl <bobl@...> wrote:

> Thanks! We just test Broadcom BCM5751 under ndiswrapper,
> "ping" is ok,
> but can't use ssh and firefox...

ping is icmp echo. ssh is tcp (and so is http - I assume that's what you meant by firefox). but the icmp vs tcp difference isn't done in hardware/driver, is it? (I think it is in the rest of the kernel, and regardless of driver/hardware). If ping works and ssh doesn't it is firewall issues, not hardware?

Hin-Tak Leung 写道:
> --- On Fri, 23/10/09, bobl <bobl@...> wrote:
>
>   
>> Thanks! We just test Broadcom BCM5751 under ndiswrapper,
>> "ping" is ok,
>> but can't use ssh and firefox...
>>     
>
> ping is icmp echo. ssh is tcp (and so is http - I assume that's what you meant by firefox). but the icmp vs tcp difference isn't done in hardware/driver, is it? (I think it is in the rest of the kernel, and regardless of driver/hardware). If ping works and ssh doesn't it is firewall issues, not hardware?
>
>
>       
>
>   
Thanks for your reply frist.

In fact, i only want to confirm ndiswrapper can support normal ethernal 
card... so i just do a test...

In my linux system, the BCM5751 work well use tg3 kernel driver. then i 
rmmod the tg3 driver, and use ndiswrapper plus windows driver. then can 
use "ping" but can't use ssh and firefox(http). so i think it's not 
friewall issue, because not change anything.

The different of TCP and ICMP is a good thread, thanks again!

Bob

bobl 写道:
> Hin-Tak Leung 写道:
>   
>> --- On Fri, 23/10/09, bobl <bobl@...> wrote:
>>
>>   
>>     
>>> Thanks! We just test Broadcom BCM5751 under ndiswrapper,
>>> "ping" is ok,
>>> but can't use ssh and firefox...
>>>     
>>>       
>> ping is icmp echo. ssh is tcp (and so is http - I assume that's what you meant by firefox). but the icmp vs tcp difference isn't done in hardware/driver, is it? (I think it is in the rest of the kernel, and regardless of driver/hardware). If ping works and ssh doesn't it is firewall issues, not hardware?
>>
>>
>>       
>>
>>   
>>     
> Thanks for your reply frist.
>
> In fact, i only want to confirm ndiswrapper can support normal ethernal 
> card... so i just do a test...
>
> In my linux system, the BCM5751 work well use tg3 kernel driver. then i 
> rmmod the tg3 driver, and use ndiswrapper plus windows driver. then can 
> use "ping" but can't use ssh and firefox(http). so i think it's not 
> friewall issue, because not change anything.
>
> The different of TCP and ICMP is a good thread, thanks again!
>   
Oh, i just use ssh -v, it report the tcp session already established, 
and i write a simple C/S program use tcp, it's also OK. so TCP is OK.

Bob
> Bob
>
> ------------------------------------------------------------------------------
> Come build with us! The BlackBerry(R) Developer Conference in SF, CA
> is the only developer event you need to attend this year. Jumpstart your
> developing skills, take BlackBerry mobile applications to market and stay 
> ahead of the curve. Join us from November 9 - 12, 2009. Register now!
> http://p.sf.net/sfu/devconference
> _______________________________________________
> Ndiswrapper-general mailing list
> Ndiswrapper-general@...
> https://lists.sourceforge.net/lists/listinfo/ndiswrapper-general
>

"Hin-Tak Leung" <hintak_leung@...> wrote

> --- On Tue, 29/9/09, Malte Gell <malte.gell@...> wrote:
> > I prefer to wait ;-) This way I can stick with kernel
> > 2.6.27 and just install
> > this package, otherwise I would have to upgrade my kernel,
> > because ar9170
> > needs kernel >= 2.6.29 as far as I know. And i always
> > fear, a kernel update
> > could break other things,
 
> Where did you learn this from? It says bleeding-edge compat-wireless
> (http://linuxwireless.org/en/users/Download#Compat-wireless_release_types)
> requires Kernel >= 2.6.27 . and it is not a *full* kernel update -

Yes, compat-wirless-2.6 needs kernel 2.6.27, but, the ar9170 driver requires 
kernel 2.6.29, below you can see the "make" output when building compat-
wireless-2.6 from today on kernel 2.6.27. It is this tracepoint issue that 
requires kernel 2.6.29

Regards
Malte


./scripts/driver-select ar9170

make

./scripts/gen-compat-autoconf.sh config.mk > include/linux/compat_autoconf.h
make -C /lib/modules/2.6.27.29-0.1-k7/build 
M=/home/malte_gell/download/src/wifi/compat-wireless-2009-09-29 modules
make[1]: Entering directory `/usr/src/linux-2.6.27.29-0.1'
  LD      /home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/drivers/net/wireless/ath/built-in.o
  CC [M]  /home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/drivers/net/wireless/ath/main.o
In file included from /home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat.h:19,
                 from <command-line>:0:
/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat-2.6.28.h:152: error: redefinition of 
‘struct tracepoint’
In file included from /home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat.h:19,
                 from <command-line>:0:
/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat-2.6.28.h:182:1: warning: "DEFINE_TRACE" 
redefined
In file included from include/linux/module.h:19,
                 from include/linux/textsearch.h:7,
                 from include/linux/skbuff.h:26,
                 from /home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat-2.6.28.h:10,
                 from /home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat.h:19,
                 from <command-line>:0:
include/linux/tracepoint.h:86:1: warning: this is the location of the previous 
definition
/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/include/net/compat-2.6.28.h:186: error: redefinition of 
‘tracepoint_update_probe_range’
include/linux/tracepoint.h:100: error: previous definition of 
‘tracepoint_update_probe_range’ was here
make[4]: *** [/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/drivers/net/wireless/ath/main.o] Fehler 1
make[3]: *** [/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/drivers/net/wireless/ath] Fehler 2
make[2]: *** [/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29/drivers/net/wireless] Fehler 2
make[1]: *** [_module_/home/malte_gell/download/src/wifi/compat-
wireless-2009-09-29] Fehler 2
make[1]: Leaving directory `/usr/src/linux-2.6.27.29-0.1'
make: *** [modules] Fehler 2

--- On Tue, 29/9/09, Malte Gell <malte.gell@...> wrote:

> "Hin-Tak Leung" <hintak_leung@...>
> wrote
> 
> > --- On Tue, 29/9/09, Malte Gell <malte.gell@...>
> wrote:
> > > I prefer to wait ;-) This way I can stick with
> kernel
> > > 2.6.27 and just install
> > > this package, otherwise I would have to upgrade
> my kernel,
> > > because ar9170
> > > needs kernel >= 2.6.29 as far as I know. And i
> always
> > > fear, a kernel update
> > > could break other things,
>  
> > Where did you learn this from? It says bleeding-edge
> compat-wireless
> > (http://linuxwireless.org/en/users/Download#Compat-wireless_release_types)
> > requires Kernel >= 2.6.27 . and it is not a *full*
> kernel update -
> 
> Yes, compat-wirless-2.6 needs kernel 2.6.27, but, the
> ar9170 driver requires 
> kernel 2.6.29, below you can see the "make" output when
> building compat-
> wireless-2.6 from today on kernel 2.6.27. It is this
> tracepoint issue that 
> requires kernel 2.6.29
> 
> Regards
> Malte

Hmm, I see you have already posted to linux-wireless and already got your answers. Hauke has already given you the answer - your kernel tree is *not* 2.6.27(.0), and as it is obvious that you have 2.6.27.29-0.1, which is 2.6.27.29 + suse-patches. Even 2.6.27.29 is not 2.6.27(.0), but somewhat closer to 2.6.28. And you really have already had your answer - your tree is somewhat between 2.6.27 and 2.6.28, and *neither*. 

My suggestion would be simply to delete all the 'error: redefinition' parts from net/compat-2.6.28.h , because you already have them in the suse tree.
You may also need to delete something from net/compat-2.6.28.c, but that will happen during the linking stage (much later).

> 
> 
> ./scripts/driver-select ar9170
> 
> make
> 
> ./scripts/gen-compat-autoconf.sh config.mk >
> include/linux/compat_autoconf.h
> make -C /lib/modules/2.6.27.29-0.1-k7/build 
> M=/home/malte_gell/download/src/wifi/compat-wireless-2009-09-29
> modules
> make[1]: Entering directory `/usr/src/linux-2.6.27.29-0.1'
>   LD     
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/drivers/net/wireless/ath/built-in.o
>   CC [M] 
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/drivers/net/wireless/ath/main.o
> In file included from
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat.h:19,
>              
>    from <command-line>:0:
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat-2.6.28.h:152: error:
> redefinition of 
> ‘struct tracepoint’
> In file included from
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat.h:19,
>              
>    from <command-line>:0:
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat-2.6.28.h:182:1:
> warning: "DEFINE_TRACE" 
> redefined
> In file included from include/linux/module.h:19,
>              
>    from include/linux/textsearch.h:7,
>              
>    from include/linux/skbuff.h:26,
>              
>    from
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat-2.6.28.h:10,
>              
>    from
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat.h:19,
>              
>    from <command-line>:0:
> include/linux/tracepoint.h:86:1: warning: this is the
> location of the previous 
> definition
> /home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/include/net/compat-2.6.28.h:186: error:
> redefinition of 
> ‘tracepoint_update_probe_range’
> include/linux/tracepoint.h:100: error: previous definition
> of 
> ‘tracepoint_update_probe_range’ was here
> make[4]: *** [/home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/drivers/net/wireless/ath/main.o] Fehler
> 1
> make[3]: *** [/home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/drivers/net/wireless/ath] Fehler 2
> make[2]: *** [/home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29/drivers/net/wireless] Fehler 2
> make[1]: ***
> [_module_/home/malte_gell/download/src/wifi/compat-
> wireless-2009-09-29] Fehler 2
> make[1]: Leaving directory `/usr/src/linux-2.6.27.29-0.1'
> make: *** [modules] Fehler 2
> 
>

"Hin-Tak Leung" <hintak_leung@...> wrote
 
> Hmm, I see you have already posted to linux-wireless and already got your
>  answers. Hauke has already given you the answer - your kernel tree is
>  *not* 2.6.27(.0), and as it is obvious that you have 2.6.27.29-0.1, which
>  is 2.6.27.29 + suse-patches. Even 2.6.27.29 is not 2.6.27(.0), but
>  somewhat closer to 2.6.28. And you really have already had your answer -
>  your tree is somewhat between 2.6.27 and 2.6.28, and *neither*.
> 
> My suggestion would be simply to delete all the 'error: redefinition' parts
>  from net/compat-2.6.28.h , because you already have them in the suse tree.

Thanx for your efforts. By the way, do I need to set  CONFIG_TRACEPOINTS in the 
kernel? I just see it is _not_ set in openSUSE 11.1.

>  You may also need to delete something from net/compat-2.6.28.c, but that
>  will happen during the linking stage (much later).

So, no need to modify compat-2.6.28.c? Does it automatically happen?

Regards
Malte

"Hin-Tak Leung" <hintak.leung@...> wrote

> > So, no need to modify compat-2.6.28.c? Does it automatically happen?
> 
> You may or may not  see any errors at the linker stage (the LD part in
> stage 2). If you get to that point without further error, then, well,
> count yourself lucky and just give "make install ; make unload;
> modprobe -v ar9170" a go.

Ho ho ;-) I got a lot of warnings like this:

 include/linux/tracepoint.h:86:1: warning: this is the location of the 
previous definition

But, it DID compile! Thanx a lot.

To be sure I did a find . -name "*.ko" and got these modules:

./drivers/net/wireless/ath/ath.ko
./net/rfkill/rfkill_backport.ko
./net/wireless/cfg80211.ko
./net/mac80211/mac80211.ko

I just wonder, where is ar9170usb !? I did ./scripts/driver-select ar9170

Malte

On Tue, 2009-09-29 at 06:57 +0200, Malte Gell wrote: 
> Hello,
> 
> Pavel Roskin <proski@...> wrote
> 
> > On Tue, 2009-09-29 at 03:59 +0200, Malte Gell wrote:
> > > kernel: ndiswrapper (import:242): unknown symbol:
> > > ntoskrnl.exe:'NlsMbCodePageTag'
> > 
> > NlsMbCodePageTag is defined in the current ndiswrapper to 0, and the
> > code in driver/pe_linker.c would consider the symbol as missing.
> > 
> > get_export() doesn't distinguish the case when the symbol is missing and
> > when it's present and set to 0.  I believe it's a bug, not an intended
> > behavior.  In other words, NlsMbCodePageTag is supposed to resolve to 0.
> > Please try the attached patch.
> 
> The patch works well with ndiswrapper 1.55 the module loaded perfectly, no 
> crash, no trouble.

Good to know.  Thank you for testing!  I have committed a similar fix.

> But, now I see an issue I saw before with an AVM Fritz USB N stick: the stick 
> cannot connect to the access point. The ar9170 driver does not show this 
> issue, the error looks like this:
> 
> ndiswrapper (iw_get_scan:1243): getting BSSID list failed (C0010016)

The message comes from iw_get_scan() in driver/iw_ndis.c.  C0010016 is
NDIS_STATUS_BUFFER_TOO_SHORT.

It looks like the driver wants to write more than 8 entries and requests
a longer buffer, but even the longer buffer is too short.  I think it's
a fair game to ignore that error.  The patch is attached.

> The adapther can see the access point and can connect for a second, but loses 
> connection. My wpa_supplicant.conf works well with ar9170.

I see it too, and I also have different problem with USB.  For instance,
I cannot unload ndiswrapper without a kernel panic.

-- 
Regards,
Pavel Roskin

Hello,

Pavel Roskin <proski@...> wrote

 
> > But, now I see an issue I saw before with an AVM Fritz USB N stick: the
> > stick cannot connect to the access point. The ar9170 driver does not show
> > this issue, the error looks like this:
> >
> > ndiswrapper (iw_get_scan:1243): getting BSSID list failed (C0010016)
> 
> The message comes from iw_get_scan() in driver/iw_ndis.c.  C0010016 is
> NDIS_STATUS_BUFFER_TOO_SHORT.
> 
> It looks like the driver wants to write more than 8 entries and requests
> a longer buffer, but even the longer buffer is too short.  I think it's
> a fair game to ignore that error.  The patch is attached.
> 
> > The adapther can see the access point and can connect for a second, but
> > loses connection. My wpa_supplicant.conf works well with ar9170.
 
> I see it too, and I also have different problem with USB.  For instance,
> I cannot unload ndiswrapper without a kernel panic.

I can unload ndiswrapper without any problems... kernel 2.6.27-29. But, in the 
meantime I switched to kernel 2.6.31 and I installed compat-wireless and now 
the Fritz! USBN N works with a native driver, so I don't need ndiswrapper any 
longer :-)

Regards
Malte

--- On Thu, 1/10/09, Malte Gell <malte.gell@...> wrote:

> I can unload ndiswrapper without any problems... kernel
> 2.6.27-29. But, in the 
> meantime I switched to kernel 2.6.31 and I installed
> compat-wireless and now 
> the Fritz! USBN N works with a native driver, so I don't
> need ndiswrapper any 
> longer :-)

It would still be nice to know if the new patch works or not. While native drivers (when they work) are almost always preferable to ndiswrapper, windows is always going to be what hardware vendors release drivers for, and if only just as a step in studying in reverse-engineering windows drivers, there is some incentive in improving ndiswrapper... so please can you give it a go (following the ususal blacklist advices for switching drivers)?

On Fri, 2009-10-02 at 08:46 +0200, Malte Gell wrote:

> Applying the patch failed:
> 
> File to patch: iw_ndis.c
> patching file iw_ndis.c
> Hunk #1 FAILED at 1239.
> 1 out of 1 hunk FAILED -- saving rejects to file iw_ndis.c.rej

Maybe the whitespace was damaged.  You can use the "-l" option to the
patch command to ignore whitespace changes.

> I have used ndiswrapper 1.55, I have attached the *original* iw_ndis.c so you 
> can look at it and correct the patch.

Please don't do it again.  I can find the 1.55 release.  Anyway, I could
apply my patch to the file you have posted, even without "-l".

If nothing helps, you can apply the patch manually.  It's just one line,
immediately above "getting BSSID list failed".

-- 
Regards,
Pavel Roskin

--- On Fri, 2/10/09, Malte Gell <malte.gell@...> wrote:

> "Hin-Tak Leung" <hintak_leung@...>
> wrote
> 
> > --- On Thu, 1/10/09, Malte Gell <malte.gell@...>
> wrote:
> > > I can unload ndiswrapper without any problems...
> kernel
> > > 2.6.27-29. But, in the
> > > meantime I switched to kernel 2.6.31 and I
> installed
> > > compat-wireless and now
> > > the Fritz! USBN N works with a native driver, so
> I don't
> > > need ndiswrapper any
> > > longer :-)
>  
> > It would still be nice to know if the new patch works
> or not. 
> 
> Ok, I will try it. 
> 
> Applying the patch failed:
> 
> File to patch: iw_ndis.c
> patching file iw_ndis.c
> Hunk #1 FAILED at 1239.
> 1 out of 1 hunk FAILED -- saving rejects to file
> iw_ndis.c.rej
> 
> I have used ndiswrapper 1.55, I have attached the
> *original* iw_ndis.c so you 
> can look at it and correct the patch.

As Pavel wrote - there is no need to include verbose unmodified/public version of source code, or any part of - we can get it elsewhere.

It is probably just white space (adding -l option may help), or a possiblity is that Pavel may be working against svn head instead of 1.55? In any case, since it is just one hunk, it should be applicable by hand...