Hi Bob,
> Here is some more info in no particular order.
>
> - Roman's driver has never worked with the Lucent chips.
Paul Mackerras submitted a patch in mid-January to make it work on
at least some Powerbooks that used Lucent chips. That's when NPS
(no power switching) mode started to be forcibly enabled on all
OHCIs, among other tweaks. So, maybe not _all_ Lucent setups.
He did comment that he hadn't done exhaustive debugging; perhaps
that NPS tweak was just masking other failures.
> - Greg Smith's driver did work fine.
Likewise with the NEC chip.
> - Looking at the USB with a CATC bus analyzer, here is what
> happens when the hub is plugged.
> Reset
> Suspend 594.026 ms
> Reset 10.613 ms
Was this Reset/Suspend/Reset before plugging in the hub? That
is, was that what ohci-hcd did when it initialized? It looks
like what the driver init does (including an extra bus reset).
> SOF
> SOF
> .
> .
> .
> There is never a Setup packet on the bus so it dies very
> early.
>
> I will ask the Lucent applications engineers what may be causing
> the UnrecoverableError interrupt bit to set on their chip.
That seems like it's useful info. I'd not known that the HC failed
before even sending any packets onto the bus!
One more thing you might try: a patch I posted last week labeled
"experimental", that provides more extensive register dumping.
It won't apply against the very latest kernels, but if there's
anything funky about OHCI register settings it might show them.
(If you have other ways to look at them, you may not need this.)
With the NEC chip I didn't find register funkiness that mattered,
so I'm currently hypothesizing (a) register assignment ordering as
an issue [ checking usb-ohci against the OHCI spec ], or (b) some
problem in TD setup that many controllers ignore ("offets" maybe),
but the NEC and Lucent ones don't.
- Dave
> Bob
>
> David Brownell wrote:
> >
> > I knew it wouldn't just be NEC chips seeing UE failures with
> > this particular OHCI driver!
> >
> > For reference, here's what the OHCI spec says about UE:
> >
> > The Host Controller sets the UnrecoverableError bit
> > when it detects a system error not related to USB or
> > an error that cannot be reported in any other way.
> >
> > HC should not proceed with any processing nor signaling
> > before the system error has been corrected. HCD clears
> > this bit after HC has been reset.
> >
> > There's also a comment about "if the controller checks offsets"
> > it'll report UE in some cases ... which gives another hint in
> > the direction of some TD queue problems causing this.
> >
> > Resetting (once) could be a good experiment to try, but I'd not
> > expect much ... after all, the controller _was_ just reset.
> >
> > Anyone with hints to share on what might cause such a UE, please
> > do so ... they've evidently been happening for a long time.
> >
> > - Dave
> >
> > Bob Cutler wrote:
> > >
> > > Randy,
> > >
> > > That patch did nothing to fix the system hangs I see with the
> > > Lucent OHCI controllers. However, I just tried the pre3-6 patch
> > > and the most resent OHCI driver changes do address this problem.
> > > Instead of a hang, I now get the following when plugging a TI
> > > 7 port hub into a Lucent USS-312 eval card:
> > >
> > > hub.c: port 1 connection change
> > > hub.c: portstatus 101, change 1, High Speed
> > > hub.c: portstatus 103, change 10, High Speed
> > > usb.c: USB new device connect, assigned device number 2
> > > usb-ohci.c: OHCI Unrecoverable Error, controller disabled
> > > usb_control/bulk_msg: timeout
> > > usb-ohci.c: unlink URB timeout!
> > > usb.c: USB device not responding, giving up (error=-108)
> > > hub.c: hub: disabling port 1
> > >
> > > This is a definate improvement! Now to find out how to recover
> > > from the "OHCI Unrecoverable Error".
> > >
> > > Also, do you know why the Linux USB mail archive page has not
> > > received any new posts in the last day and a half?
> > >
> > > Thanks,
> > > Bob
> > >
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
