On Tue, Sep 24, 2002, Doug Alcorn <[EMAIL PROTECTED]> wrote: > Johannes Erdfelt <[EMAIL PROTECTED]> writes: > > > It's probably just a race condition with checking is_suspended and > > harmless. > > > > Can you try this patch? This should fix that race as well as make sure > > we only suspend the HC when we have nothing else waiting. (Another bug) > > I think I hit the bug again, but can't be sure. I don't have a serial > console. I only capture kernel messages using syslogd/klogd. That > said, I had something bad happen last night. My test setup is two > notebooks. The test notebook is usually closed, and I ssh into it > from my devel notebook. I needed to reload my driver and the only way > to rmmod it is if I disconnect my device (so my use count goes down). > I did the unplug, and my network connection between my devel box and > my test box hung. After the small lag it took me to figure out my > network connection had hung, I opened up the lid on my test notebook > and saw what looked like a very large stack trace scrolling by. It > would dump a half dozen numbers in the [<number>] format, pause for a > second or two, then dump some more. I finally had to hard cycle the > box. Unfortunately, nothing was in the kern.log. > > Obviously I don't know if this is a controller issue or my driver > issue. It could be some bug in the disconnect routine. I have an > interrupt urb that I don't think I'm unlinking at disconnect time. I > didn't study your patch and don't pretend to understand the HC code. > The problem could be in there. I don't know. I just thought I would > give you a heads up that something bad happened that might be related > to this patch. I'm going to try and scare up a null-modem cable to > setup a serial console.
While I can't guarantee that the patch I gave you wouldn't cause this, I don't think it would. It sounds like this is a different bug. BTW - Your driver probably shouldn't be incrementing the usage count at connect time, only when an application opens the device. JE ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
