[9fans] usbuhci bug

[9fans] usbuhci bug

[erik quanstrom]

speaking of usb ...

using a serial-to-usb converter and usb/serial, i've been able to
put my terminal into an unusable state on a fairly regular basis.

here's the step-by-step.
1.  run usb/serial
2.  wait for a usb disconnect due to a failed read
3.  notice that your machine is having trouble scheduling.
4.  note that usbuhci is blasting 100,000 interrupts/sec!
5.  try to slay and restart usb/serial ... no joy there.
6.  give up and reboot terminal.

unfortunately, i don't have any time to dig into this one atm.

- erik

Re: [9fans] usbuhci bug

[Gorka Guardiola]

Can you send me off-list usb/serial -dd output. At least so I know what device we are talking about :-).

On Sep 1, 2011, at 5:44 PM, erik quanstrom <quanstro@labs.coraid.com> wrote:

> speaking of usb ...
> 
> using a serial-to-usb converter and usb/serial, i've been able to
> put my terminal into an unusable state on a fairly regular basis.
> 
> here's the step-by-step.
> 1.  run usb/serial
> 2.  wait for a usb disconnect due to a failed read
> 3.  notice that your machine is having trouble scheduling.
> 4.  note that usbuhci is blasting 100,000 interrupts/sec!
> 5.  try to slay and restart usb/serial ... no joy there.
> 6.  give up and reboot terminal.
> 
> unfortunately, i don't have any time to dig into this one atm.
> 
> - erik
> 

Re: [9fans] usbuhci bug

[erik quanstrom]

> > 5.  try to slay and restart usb/serial ... no joy there.

i should have been more clear.  restarting usb/serial worked
fine.  the interrupt problem didn't go away.

- erik

Re: [9fans] usbuhci bug

[Gorka Guardiola]

On Thu, Sep 1, 2011 at 6:35 PM, erik quanstrom <quanstro@labs.coraid.com> wrote:
>> > 5.  try to slay and restart usb/serial ... no joy there.
>
> i should have been more clear.  restarting usb/serial worked
> fine.  the interrupt problem didn't go away.
>

Hmm, but the old usb/serial dies/is killed?. Because it is a very
different problem (in one case it
is most probably a serial problem in the other it is an ehci problem
or hardware).

Another question is, the fail read is because the serial is unplugged
or a hardware
issued error?.

Thanks,

G.

Re: [9fans] usbuhci bug

[erik quanstrom]

On Thu Sep  1 16:08:21 EDT 2011, paurea@gmail.com wrote:
> On Thu, Sep 1, 2011 at 6:35 PM, erik quanstrom <quanstro@labs.coraid.com> wrote:
> >> > 5.  try to slay and restart usb/serial ... no joy there.
> >
> > i should have been more clear.  restarting usb/serial worked
> > fine.  the interrupt problem didn't go away.
> >
>
> Hmm, but the old usb/serial dies/is killed?. Because it is a very
> different problem (in one case it
> is most probably a serial problem in the other it is an ehci problem
> or hardware).
>
> Another question is, the fail read is because the serial is unplugged
> or a hardware
> issued error?.

it is a hardware issued error.  the serial was in the same state overnight.
it was also interesting that unplugging the serial dongle didn't fix the
interrupt problem either.

- erik