Hi Alan, > Good work; I figured it out. Although it's easy to forget, there are > _two_ routines which can add a QH to the schedule, and they both need > to initialize qh->xacterrs. Heh, I suspected as much. > I decided to simplify the transaction error counting at the same time. I was wondering about that as well, this looks a lot easier to understand :-) > The resulting patch below should fix the problem. It is independent of > the other patch, although it touches some adjacent lines. Yup, it works. Now the XactErrs have proper retry counts, starting at 1. The errors are still there, though. However, it seems that the errors appear when the hub is plugged in, and more (slightly different) errors appear when a device is plugged into the "inner" hub. I guess this makes it likely that there is something faulty in the connection between the two hubs. I'll have a go with an older kernel version and another OS to see what happens there. Thanks for the fixes, I hope to seem them in mainline soon :-) Ironically though, I think I upgraded to the latest RC kernel because of the XactErrs occuring when plugging in my Wacom tablet, which we didn't solve yet :-p Gr. Matthijs
Attachment:
signature.asc
Description: Digital signature
