David,

> It's extremely unlikely that changes to ISO code had _any_ effect
> on INTR transfers.
You are right. I've changed now my driver from IRQ to ISO and the problem is exactly the same.


> So that bug you reported earlier now seems to be gone?
Yes. That seems to be fixed now.

> Was that after changes to your device driver? Or was it maybe the
> periodic schedule scanning patch that changed this behavior?
I haven't changed anything in my driver for a while. Under 2.6.1 the behaviour was ok. I could shut down without any problem. Seems to be caused by the patch. More specific: it seems to be the ehci code. Now I'm running USB 1.1 and 2.0 through exactly the same code (ISO transfer, before is was ISO under 1.1 and IRQ under 2.0). With USB 1.1 everyhing is fine. With USB 2.0 everything is fine as long as I don't submit periodic URBs.


> Don't forget to send ksymoops output when reporting such things;
> there are thousands of potential reasons for such things, and the
> ksymoops output often rules out most of them.
In fact there are no oopses. The last thing I see is my driver reporing the end of its own disconnect function. After that the kernel completely crashes. No output. Can I get more output somehow?


/Bernd


David Brownell wrote:
Bernd Porr wrote:

There seems to be a side effect on the IRQ transfer mode in high speed:

- submit IRQ URB (high speed)
- resubmit it a couple of times in the completion handler
- stop resubmitting

The above can be repeated without any problem.


So that bug you reported earlier now seems to be gone?

Was that after changes to your device driver?  Or was it maybe the
periodic schedule scanning patch that changed this behavior?

It's extremely unlikely that changes to ISO code had _any_ effect
on INTR transfers.


However, when I shut down the kernel it competely crashes after my disconnect has been called.


Don't forget to send ksymoops output when reporting such things;
there are thousands of potential reasons for such things, and the
ksymoops output often rules out most of them.

- Dave




------------------------------------------------------- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to