ndiswrapper

Hi Jan,

I checked out a new version from CVS and applied your patch. I re-built 
ndiswrapper with DEBUG=3 and gave it another try. The kernel was hung after 
doing a modprobe ndiswrapper.

I am using a system with the Intel 865G chipset. USB 2.0

Kernel messages attached. I would be happy to do more tests.

Thanks,
Mark

On Friday 17 September 2004 06:42, Jan Kiszka wrote:
> Hugo Visser wrote:
> > Jan,
> >
> > For what its worth, I'm experiencing the exact same problems on a EPIA
> > M10k. When I load ehci-hcd ndiswrapper crashes. The cvs of ndiswrapper
> > doesn't fix this, it seems even more unstable. I've tried this on my
> > current kernel 2.6.7-epia1 and a 2.6.8.1 kernel with epia patches.
> >
> > If you'd like me to test things just let me know.
>
> Thanks for offering help, I may contact you soon again.
>
> Yesterday, after many unsuccessful experiments, I made at least a small
> progress: by increasing the timeout when waiting on the reset of the
> device, the driver now loads every second time or so. It fails sometime
> because the usb subsystem reports a handshake timeout. If you already
> like to do an experiment, apply the patch below. This whole problem is
> obviously a timing issue during driver startup. And such issues are hard
> to debug (e.g. direct your output to a serial console and you will have
> different results...).
>
> More news soon.
>
> Jan

Mark Cannon wrote:

>Hi Jan,
>
>I checked out a new version from CVS and applied your patch. I re-built 
>ndiswrapper with DEBUG=3 and gave it another try. The kernel was hung after 
>doing a modprobe ndiswrapper.
>  
>
Yes, there must be still some other bug with different effects on 
different systems.

>I am using a system with the Intel 865G chipset. USB 2.0
>
>Kernel messages attached. I would be happy to do more tests.
>
>  
>
Is the log complete, i.e. written to a serial console, or it is what 
made it into the log file before the crash?

Jan

Unfortunately, this is only what made it into the log file. 
How do I redirect the debug output to the serial console (sorry if you have 
answered this already)

Mark
On Friday 17 September 2004 12:04, Jan Kiszka wrote:
> Mark Cannon wrote:
> >Hi Jan,
> >
> >I checked out a new version from CVS and applied your patch. I re-built
> >ndiswrapper with DEBUG=3 and gave it another try. The kernel was hung
> > after doing a modprobe ndiswrapper.
>
> Yes, there must be still some other bug with different effects on
> different systems.
>
> >I am using a system with the Intel 865G chipset. USB 2.0
> >
> >Kernel messages attached. I would be happy to do more tests.
>
> Is the log complete, i.e. written to a serial console, or it is what
> made it into the log file before the crash?
>
> Jan

Mark Cannon wrote:

>Unfortunately, this is only what made it into the log file. 
>How do I redirect the debug output to the serial console (sorry if you have 
>answered this already)
>
>  
>
Take a look at 
http://lxr.linux.no/source/Documentation/serial-console.txt (also 
available with your local kernel source tree).

Jan

On Friday 17 September 2004 17:04, Jan Kiszka wrote:
> Mark Cannon wrote:
> >Hi Jan,
> >
> >I checked out a new version from CVS and applied your patch. I
> > re-built ndiswrapper with DEBUG=3 and gave it another try. The
> > kernel was hung after doing a modprobe ndiswrapper.
>
> Yes, there must be still some other bug with different effects on
> different systems.
>

I am using a SiS952 chipset with a Netgear WG111.

This locks up when using EHCI in recent cvs versions, but cvs CVS top of 
tree from 23rd August 2004 worked fine in EHCI mode, but failed in OHCI 
mode! Is this a useful data point?

I will try the patch if I get a chance.

Thanks,
Tim


******************Legal Disclaimer**************************
"This email may contain confidential and privileged material for the sole use of the intended recipient.  Any unauthorized review, use or distribution by others is strictly prohibited.  If you have received the message in error, please advise the sender by reply email and delete the message. Thank you."
************************************************************

Mark,

I actually had better results not modprobing but putting an alias in 
modules.conf. First I had ndiswrapper loaded (modprobed) at startup, but 
for some reason this seems to be more stable, though it looks quite the 
same to me. I guess its due to the timing of loading the usb modules, 
but anyway you might try that too...

Regards,

Hugo

Mark Cannon wrote:

> Hi Jan,
> 
> I checked out a new version from CVS and applied your patch. I re-built 
> ndiswrapper with DEBUG=3 and gave it another try. The kernel was hung after 
> doing a modprobe ndiswrapper.

Hugo Visser wrote:
> Mark,
> 
> I actually had better results not modprobing but putting an alias in 
> modules.conf. First I had ndiswrapper loaded (modprobed) at startup, but 
> for some reason this seems to be more stable, though it looks quite the 
> same to me. I guess its due to the timing of loading the usb modules, 
> but anyway you might try that too...
> 

Mmh, will try this. However, it remains mysterious. Yesterday I gave it 
a try on my notebook with USB 2.0 (Thinkpad R40): no chance to crash it 
here, it simply works in every order to plug/load the device and driver. 
But on the VIA system, there is a very high chance to get either a USB 
timeout or even a crash (always at the same position inside the windows 
driver) when you load the driver the second time, and a medium change 
when you load at system startup. As soon as it runs, it seems to be 
rock-stable (the EPIA system runs permanently as a router/server). The 
best chance for a successful startup is to power-cycle the USB device 
(unplug once) before booting.

I will try to have another look at it again next week. In the meantime, 
some crash logs from other users could be helpful, maybe they point a 
bit closer to the core of the issue. My system is not that useful in 
this regard :(

Jan

Jan Kiszka wrote:


> But on the VIA system, there is a very high chance to get either a USB 
> timeout or even a crash (always at the same position inside the windows 

Maybe it is related, maybe not, but on my Acer 1357LMi (VIA KM400) all 
the times I've tried to connect an external USB2 drive I had my logs 
full of timeouts. To be able to use it, I had to stop hotplug, remove 
the USB2 module and force the disk to use USB1.1, at which point it 
would work fine, albeit very slowly.

Luckily for me, my RTL8180 is onboard and not USB2 8-)

HTH,
Andrea.

-- 
Homepage: http://andrea.borgia.bo.it     /    Amateur radio: IZ4FHT
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

Jan,

If a crash occurs, most of the time I have to poweroff the machine, pull 
out the powercord, power the machine on so that it discharges and then 
the linksys will work again :) Else it won't find my ap. Anyway, I 
digged through my pile of cables but haven't found a null-modem cable 
yet, so I can't produce any serial logs (yet).

What I did notice on the epia's is that it tends to share irq with the 
pci slot. I tried windows on it but the machine would just freeze if the 
linksys was connected and the pvr 250 in the pci slot used. So it can 
also be a irq sharing issue. I found no way however to move either the 
pci or the usb to a separate irq.

The patch you sent me yesterday didn't crash on me at all while 
transfering about 2gb from my epia using ftp. One thing I noticed is 
that the transfer stops once in a while. This was also the case with the 
unpatched/release version. I don't know for sure if this is ndiswrapper 
related or something in oftpd so I need to investigate that.

I'll go shopping for a null modem cable now :)



Jan Kiszka wrote:

> Hugo Visser wrote:
> 
>> Mark,
>>
>> I actually had better results not modprobing but putting an alias in 
>> modules.conf. First I had ndiswrapper loaded (modprobed) at startup, 
>> but for some reason this seems to be more stable, though it looks 
>> quite the same to me. I guess its due to the timing of loading the usb 
>> modules, but anyway you might try that too...
>>
> 
> Mmh, will try this. However, it remains mysterious. Yesterday I gave it 
> a try on my notebook with USB 2.0 (Thinkpad R40): no chance to crash it 
> here, it simply works in every order to plug/load the device and driver. 
> But on the VIA system, there is a very high chance to get either a USB 
> timeout or even a crash (always at the same position inside the windows 
> driver) when you load the driver the second time, and a medium change 
> when you load at system startup. As soon as it runs, it seems to be 
> rock-stable (the EPIA system runs permanently as a router/server). The 
> best chance for a successful startup is to power-cycle the USB device 
> (unplug once) before booting.
> 
> I will try to have another look at it again next week. In the meantime, 
> some crash logs from other users could be helpful, maybe they point a 
> bit closer to the core of the issue. My system is not that useful in 
> this regard :(
> 
> Jan

Jan,

The unit I was testing on ran into some bad memory problems. It took me a 
while to figure out that it was memory (errors accessing hard drive or 
cdrom). Of course this happened while I was rebuilding the kernel to include 
serial console output :-| 

I am swamped at work for the next day or so, but I should be able to get you a 
crash log by Saturday, I hope.

Mark

On Saturday 18 September 2004 04:40, Jan Kiszka wrote:
> Hugo Visser wrote:
> > Mark,
> >
> > I actually had better results not modprobing but putting an alias in
> > modules.conf. First I had ndiswrapper loaded (modprobed) at startup, but
> > for some reason this seems to be more stable, though it looks quite the
> > same to me. I guess its due to the timing of loading the usb modules,
> > but anyway you might try that too...
>
> Mmh, will try this. However, it remains mysterious. Yesterday I gave it
> a try on my notebook with USB 2.0 (Thinkpad R40): no chance to crash it
> here, it simply works in every order to plug/load the device and driver.
> But on the VIA system, there is a very high chance to get either a USB
> timeout or even a crash (always at the same position inside the windows
> driver) when you load the driver the second time, and a medium change
> when you load at system startup. As soon as it runs, it seems to be
> rock-stable (the EPIA system runs permanently as a router/server). The
> best chance for a successful startup is to power-cycle the USB device
> (unplug once) before booting.
>
> I will try to have another look at it again next week. In the meantime,
> some crash logs from other users could be helpful, maybe they point a
> bit closer to the core of the issue. My system is not that useful in
> this regard :(
>
> Jan

Mark Cannon <random.bits <at> rogers.com> writes:

>Mark
> 
> On Saturday 18 September 2004 04:40, Jan Kiszka wrote:
> > Hugo Visser wrote:
> > > Mark,
> > >
> > > I actually had better results not modprobing but putting an alias in
> > > modules.conf. First I had ndiswrapper loaded (modprobed) at startup, but
> > > for some reason this seems to be more stable, though it looks quite the
> > > same to me. I guess its due to the timing of loading the usb modules,
> > > but anyway you might try that too...
> >


I, too, have been trying to get the WUSB54G to work on my system.  I've made 
it to the "modprobe ndiswrapper" step, but SuSE 9.1 (Kernel 2.6.5) locks up.  
I can run this command when the USB adapter is not hooked up, but as soon as I 
hook it up, the system freezes.  Has anyone with the USB54G had this problem 
and been able to figure out a fix for it?

Thanks!

TCA

TCA wrote:

>I, too, have been trying to get the WUSB54G to work on my system.  I've made 
>it to the "modprobe ndiswrapper" step, but SuSE 9.1 (Kernel 2.6.5) locks up.  
>I can run this command when the USB adapter is not hooked up, but as soon as I 
>hook it up, the system freezes.  Has anyone with the USB54G had this problem 
>and been able to figure out a fix for it?
>
>  
>
Not really a fix but a workaround - if you can live with 802.11b only 
(11 Mbit/s) - is to unload the ehci_hcd driver. To my experience, 
running the WUSB54G only with a uhci/ohci_hcd USB 1.1 host controller 
driver causes no problems.

This issue seems to be hardware-dependent: On an IBM notebook with Intel 
Centrino chipset (also SuSE 9.1), plugging in the WUSB54G is even 
possible with ehci_hcd loaded. On the other hand, my VIA EPIA M6000 box 
with any 2.6 kernel I tried tends to crash or time-out upon loading 
ndiswrapper. The only solution so far is a cold start (also for the 
WUSB54G!) which brings up the system and lets it run without significant 
problems (current uptime is 3 days). Sometimes my notebook seems to 
confuse the WUSB54G which then drops its configuration. A watchdog 
script catches this and re-runs iwconfig. But this issue may already be 
a problem of the Windows driver - you can't look inside.

BTW, what chipset do you use? Already tried it on some other system? 
Just for the statistics. :-/

Well, I know I should dig deeper into this issue, especially as it shows 
up on my own hardware. But the problem is that this will cost *a lot* of 
time (I already spent many hours on this particular bug) - and it 
"unfortunately" already works for me with acceptable limitations. Maybe 
there will be some rainy autumn day soon...

Jan

Jan Kiszka <jan.kiszka <at> web.de> writes:

> Not really a fix but a workaround - if you can live with 802.11b only 
> (11 Mbit/s) - is to unload the ehci_hcd driver. To my experience, 
> running the WUSB54G only with a uhci/ohci_hcd USB 1.1 host controller 
> driver causes no problems.
> 
.......

> BTW, what chipset do you use? Already tried it on some other system? 
> Just for the statistics. :-/
> 
> Well, I know I should dig deeper into this issue, especially as it shows 
> up on my own hardware. But the problem is that this will cost *a lot* of 
> time (I already spent many hours on this particular bug) - and it 
> "unfortunately" already works for me with acceptable limitations. Maybe 
> there will be some rainy autumn day soon...
> 
> Jan
> 
> 
> Attachment (smime.p7s): application/x-pkcs7-signature, 4760 bytes


I'm using an older system with chipset VIA 693 (Apollo PRO133).  I haven't 
tried this on any other system...I just picked linux back up after SP2 made my 
system run incredibly slow.

Since I'm relatively new to the whole process, could you point me in the right 
direction to unload the ehci_hcd driver, and to load the uhci/ohci_hcd USB 1.1 
host controller driver?  I'm not worried about "only" 802.11b since I only use 
my connection at home for fun.  I'd just be thrilled to get an internet 
connection with this WUSB54G!  And I sincerely appreciate all the help you've 
been!

TCA

TCA wrote:

>I'm using an older system with chipset VIA 693 (Apollo PRO133).  I haven't 
>tried this on any other system...I just picked linux back up after SP2 made my 
>system run incredibly slow.
>  
>
Mmh, this likely means that there is no USB 2.0 support, thus no ehci_hcd...

>Since I'm relatively new to the whole process, could you point me in the right 
>direction to unload the ehci_hcd driver, and to load the uhci/ohci_hcd USB 1.1 
>host controller driver?  I'm not worried about "only" 802.11b since I only use 
>my connection at home for fun.  I'd just be thrilled to get an internet 
>connection with this WUSB54G!  And I sincerely appreciate all the help you've 
>been!
>
>  
>
Try "rmmod ehci_hcd", the uhci or ohci driver should already be loaded 
in parallel and will take over. You may check your list of driver 
modules with "lsmod".

Jan

Jan Kiszka <jan.kiszka <at> web.de> writes:

> Mmh, this likely means that there is no USB 2.0 support, thus no ehci_hcd...


Yeah, you are correct about this.  I went ahead and checked to see what mods 
were loaded, and only uhci_hcd appeared.  No ehci_hcd was found.  I tried 
loading ndiswrapper once again for kicks, same problem.  Does this mean I'm 
stuck??  I've made it this far...I'd hate to have to try to run a wire across 
half my house ;)

TCA

TCA <thalocsta <at> yahoo.com> writes:
> 
> Yeah, you are correct about this.  I went ahead and checked to see what mods 
> were loaded, and only uhci_hcd appeared.  No ehci_hcd was found.  I tried 
> loading ndiswrapper once again for kicks, same problem.  Does this mean I'm 
> stuck??  I've made it this far...I'd hate to have to try to run a wire across 
> half my house ;)
> 
> TCA

I finally got things going with my WUSB54G.  I had to go with no smp, and that
solved the modprobe freezing problem.  After that, configuration was not
difficult, and I've had no problem whatsoever (not even a dropped signal)!

Thanks again for all the help Jan!

TCA

TCA wrote:

>TCA <thalocsta <at> yahoo.com> writes:
>  
>
>>Yeah, you are correct about this.  I went ahead and checked to see what mods 
>>were loaded, and only uhci_hcd appeared.  No ehci_hcd was found.  I tried 
>>loading ndiswrapper once again for kicks, same problem.  Does this mean I'm 
>>stuck??  I've made it this far...I'd hate to have to try to run a wire across 
>>half my house ;)
>>
>>TCA
>>    
>>
>
>I finally got things going with my WUSB54G.  I had to go with no smp, and that
>solved the modprobe freezing problem.  After that, configuration was not
>difficult, and I've had no problem whatsoever (not even a dropped signal)!
>
>  
>
Good to know. I'm not running ndiswrapper with SMP enabled anywhere, so 
this might be worth a try if it crashes on my system as well - and may 
point me to potentially remaining and well-hidden locking bugs...

BTW, after 4 days of continuous operation, I actually had a first crash 
of my box which was very likely caused by ndiswrapper. :-/

Jan