hello michel,

michel Xhaard wrote:
> Le Vendredi 10 Mars 2006 01:27, thomas schorpp a écrit :
> 
>>thomas schorpp wrote:
>>
>>>thomas schorpp wrote:
>>>
>>>>thomas schorpp wrote:
>>>>
>>>>>thomas schorpp wrote:
>>>>>
>>>>>>thomas schorpp wrote:
>>>>>>
>>>>>>>hello,
>>>>>>>
>>>>>>>touching ekiga gui video sliders triggers spurious stalls of 2.6
>>>>>>>kernels.
>>>>>>>
>>>>>>>any clue? it occours more often on touching the contrast slider.
>>>>>>>color and hue seems not to be affected.
>>>>>>>usb seems not affected - no stall until touching sliders.
>>>>>>>
>>>>>>>this will be hard to debug since there will be no kernel oops and
>>>>>>>no logs on serial cable console too.
>>>>>>>the kernel seems to lock up in a endless loop somewhere.
>>>>>>>
>>>>>>>2.6.15.0 K7 kernel.org
>>>>>>>aiptek dv3500 webcam mode
>>>>>>>ekiga cvs + libs cvs
>>>>>>>
>>>>>>>y
>>>>>>>tom
>>>>>>
>>>>>>any grabber triggers.
>>>>>>
>>>>>>the stall seems to occur if light gets too dark (black picture).
>>>>>
>>>>>no also occurs on picture change.
>>>>>
>>>>>
>>>>>>here are debug logs until deadlock (serial console gets no oops, even
>>>>>>only once so far flashing keyboard lights):
>>>>
>>>>news:
>>>>
>>>>its not the spca driver.
>>>>
>>>>i connected to another usb plug and all is fine again.
>>>>
>>>>seems like a bug in the 2.6.15 kernel usb drivers:
>>>
>>>it was a bug in kernel 2.6.15.x.
>>>
>>>fixed in 2.6.15.5.
>>>
>>>all fine again.
>>
>>reopened. occurred again with 2.6.15.6 kernel and spca5xxx 3/1/2006
>>
>>on fast and big brightness change.
>>sometimes on black picture (low light).
>>this points to a bug im picture-processing, maybe spcadecoder.
>>i cant see anything unusual in the interrupt handler.
>>
>>full kernel stall.
>>
>>debugfs is useless here. no access under stalled kernel,
>>i dont know if mounting to nfs mounted fs is possible.
>>
>>trying again with serial console kernel debugging and step tracing.
>>
>>y
>>tom
>>
>>
> 
> Thomas,
> The problem here is very complex that should be:
> software:
>       The usb layer on the kernel API
>       The driver specially the spcadecoder

most likely.

> Hardware or firmware:
>       The Aiptek DV3500
>       The usb Host controller

impossible. unchanged since a year and working fine under winxp.
and worked under linux fine until at least kernel 2.6.13.

> As in all your report you speak about light or contrast we can raisonably 
> disable the kernel API and the spcadecoder these two parts don't care about 
> light :)

hmm, ok theoretically.

> In the same way we can disable the Usb Host Controller.
> Now only the Aiptek DV3500 is on the list 
> What about light and the video processor in the DV
>       If there are a lot of light the video processor on the DV decrease the 
> times 
> exposure, of course decreasing times exposure will increase the frame rate. 
> As a second effect with a lot of light the jpeg encoder inside the chips find 
> more details in the picture so the picture size increase. With this context 
> we expect a very Hight CPU load 

ill check this.

>       If there are not enought light the video processor increase the times 
> exposure that will of course decrease the frame rate. 
> All this process are know to be Autoexposure, set inside the chips firmware.
> All spca jpg webcam is set with this feature by default. We did not have 
> acces 
> to this feature, know only by Sunplus and maybe M$ developper.
> I suspect the chips to go out the limit of the Autoexposure and goes to sleep 
> when the interrupt Handler withing the kernel is waiting for data .

this could be traceable with verbose printk logs or debugfs.
if it is i will try to handle this state somehow if possible.
but this was surely *no* problem before 2.6.14. it worked just fine.

> I have no 
> solution ATM and cannot produce the same effect as your DV3500 with my old 
> DVII sorry :(

sorry, but ive never seen autoexposure under linux with this device and the 
spca driver.
i had to have always adjust brightness manually.
so this would be out, too. 
but ill check with windows. would a usbsnoop log help? i could send you 
per postage on data cd. pls tell me which snooper you prefer.

but many thanks for your comprehensive system analysis and kind explanation.
i must get behind this. i wont give the device up. it worked fine for many 
months.
since i have had many trouble with linux kernel usb since using linux, there 
has been most probably something faulty introduced after 2.6.13/14.

Alan Stern was talking on a "bug in the spca driver" on the usb list lately,
maybe he can comment on this, since i were surely too "limited" to get this 
fixed, cause i didnt went to havard(?) ;)

> Regards

kind regards,
thomas


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to