On Sat, Sep 15, 2001 at 05:00:42PM +0200, Georg Acher wrote:
> On Fri, Sep 14, 2001 at 06:09:52PM -0700, Jean Tourrilhes wrote:
>
> > Questions :
> > Why the completion handler is not called on timeout in this case ?
>
> Timeoutchecking is done by an UHCI interrupt, which is activated on demand
> if there are URBs with a set timeout. If the host controller itself dies,
> this interrupt dies, too.
Ok. This explains it.
> > Why transfer always fail ?
>
> You've written "generic USB card", which is usually a VIA chip. This chip
> has at least one bug/flaw/funny-behavior, which kills all processing if a
> babble is encountered. No error interrupts, it's simply dead. Sometimes a
> module reload helps, sometimes a HW reset is necessary. This flaw is not
> reproducible in Windows, and I'm still wondering what they are doing to
> prevent it.
I'll try to find some different hardware. Maybe I should grab
some OHCI.
> I don't know if your behaviour is related, but you can check this: Does
> unplug/replug of an USB device still work after the IRDA-death,
Usually unplug/replug work erratically, so I try to avoid it.
> are the
> UHCI-interrupts incrementing? If not, you've hit the problem.
I'll check.
> > Why usb-uhci lock the box on removal ?
>
> That's definitely a bug, can you check with alt-sys-req-p if the box still
> lives (and where...)
I'll check.
> > What's going on ?
>
> Good question...
>
> PS: If this bug is lurking around for over 6 months when IRDA is used, why
> I'm heearing it now for the first time?
The bug about the timeout was mentionned in the second part of
this e-mail to the usb mailing list :
http://www.geocrawler.com/archives/3/2571/2001/2/0/5249167/
The bug about removal was mentionned in a small paragraph of
this e-mail to the usb mailing list :
http://www.geocrawler.com/archives/3/2571/2001/2/0/5250955/
In other words, I've got absolutely no idea why you never
heard about it...
On why I'm not pushing harder on this bug : it seems to happen
on my SMP box only when I put two dongles, with only one dongle this
box is stable. Other people seems to not come across it, and overall
there is few users. The other single board PC (UP) where it was
happening with one dongle has been put to production and I've switched
it to using a serial IrDA dongle, because I could not fix its problem
and USB was messing up the Ethernet card.
That's it... Thanks for you attention ;-)
Jean
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
