Bug in usbser.sys when using a high speed hub

I have a full speed USB device that I'm developing that uses usbser.sys.
I found that if I connect the device directly to my Dell laptop everything
works find. If I connect the device through a high speed USB 2.0 hub
everything is very flaky. I see hangs when opening the com port, Windows
lockups, and other wierd falures.

I found a workaround. If I disabled the enhanced USB controller in
device manager to force the hub to use full speed everything is rock solid.
So this makes me think the problem is either a bug in Windows or a bug with
the Intel chipset in my Dell Inspiron laptop.

1. Is this a known bug in usbser.sys or Windows?
2. What do I need to do to report the problem?

My operating system is Vista SP1.

chris
Thu May 01 21:13:27 CDT 2008

On May 1, 7:53 pm, Armageddon <Armaged...@discussions.microsoft.com>
wrote:

> I found a workaround. If I disabled the enhanced USB controller in
> device manager to force the hub to use full speed everything is rock solid.
> So this makes me think the problem is either a bug in Windows or a bug with
> the Intel chipset in my Dell Inspiron laptop.

I highly doubt this is an issue with usbser.sys as most class drivers
don't care, or even know, about whether the underlying device is USB
1.1, USB 2.0, or even a real physical device at all. I wouldn't blame
Windows or your laptop -- I'd say 90% chance your device is buggy.

Armageddon
Thu May 01 21:29:11 CDT 2008



"chris.aseltine@gmail.com" wrote:

> On May 1, 7:53 pm, Armageddon <Armaged...@discussions.microsoft.com>
> wrote:
>
> > I found a workaround. If I disabled the enhanced USB controller in
> > device manager to force the hub to use full speed everything is rock solid.
> > So this makes me think the problem is either a bug in Windows or a bug with
> > the Intel chipset in my Dell Inspiron laptop.
>
> I highly doubt this is an issue with usbser.sys as most class drivers
> don't care, or even know, about whether the underlying device is USB
> 1.1, USB 2.0, or even a real physical device at all. I wouldn't blame
> Windows or your laptop -- I'd say 90% chance your device is buggy.
>

And Windows is bug free?

The firmware on the device also doesn't care if a hub is used or not. The
device is a full speed device so I'm confused how having a high speed link
between the hub and the PC can make a difference as far as the device is
concerned. Doesn't the link from the hub to the device simply look like a
plain full speed link?

I'm also confused why Windows malfunctions. It's not just that the device
doesn't work. Isn't the driver stack different if a hub is used? Perhaps
it's some kind of wierd interaction between device drivers.

chris
Fri May 02 10:06:56 CDT 2008

On May 1, 9:29 pm, Armageddon <Armaged...@discussions.microsoft.com>
wrote:

> And Windows is bug free?

Did I say that?

> The firmware on the device also doesn't care if a hub is used or not. The
> device is a full speed device so I'm confused how having a high speed link
> between the hub and the PC can make a difference as far as the device is
> concerned. Doesn't the link from the hub to the device simply look like a
> plain full speed link?

Get a bus analyzer and compare traces between connecting it directly
and connecting it through a hub, and look for errors. If your device
is going wild on the wire, you could potentially bugcheck the machine,
but that's not really the host's fault (inasmuch as you're not going
to get a hotfix to work around a device's buggy behavior).

> I'm also confused why Windows malfunctions. It's not just that the device
> doesn't work. Isn't the driver stack different if a hub is used? Perhaps
> it's some kind of wierd interaction between device drivers.

Assuming it's a 2.0 hub, then you'll be hitting the EHCI stack on the
host side, yes.

Pavel
Fri May 02 09:19:35 CDT 2008

"Armageddon" <Armageddon@discussions.microsoft.com> wrote in message
news:D80B0827-3D9C-41E5-A470-036D4D6BA4EE@microsoft.com...

>Doesn't the link from the hub to the device simply look like a
> plain full speed link?

No, the physical signaling is different in high speed mode (besides of the
obvious timing differences)

--PA

Gianluca
Fri May 02 10:46:03 CDT 2008

Can it be a buggy USB hub?

Just my two cents
GV


--
Gianluca Varenni, Windows DDK MVP

CACE Technologies
http://www.cacetech.com



"Armageddon" <Armageddon@discussions.microsoft.com> wrote in message
news:F800F5FC-0F95-47FF-AC76-6375216EF142@microsoft.com...
> I have a full speed USB device that I'm developing that uses usbser.sys.
> I found that if I connect the device directly to my Dell laptop everything
> works find. If I connect the device through a high speed USB 2.0 hub
> everything is very flaky. I see hangs when opening the com port, Windows
> lockups, and other wierd falures.
>
> I found a workaround. If I disabled the enhanced USB controller in
> device manager to force the hub to use full speed everything is rock
> solid.
> So this makes me think the problem is either a bug in Windows or a bug
> with
> the Intel chipset in my Dell Inspiron laptop.
>
> 1. Is this a known bug in usbser.sys or Windows?
> 2. What do I need to do to report the problem?
>
> My operating system is Vista SP1.

Armageddon
Fri May 02 11:56:03 CDT 2008



"chris.aseltine@gmail.com" wrote:

> On May 1, 9:29 pm, Armageddon <Armaged...@discussions.microsoft.com>
> wrote:
>
> > And Windows is bug free?
>
> Did I say that?
>
> > The firmware on the device also doesn't care if a hub is used or not. The
> > device is a full speed device so I'm confused how having a high speed link
> > between the hub and the PC can make a difference as far as the device is
> > concerned. Doesn't the link from the hub to the device simply look like a
> > plain full speed link?
>
> Get a bus analyzer and compare traces between connecting it directly
> and connecting it through a hub, and look for errors. If your device
> is going wild on the wire, you could potentially bugcheck the machine,
> but that's not really the host's fault (inasmuch as you're not going
> to get a hotfix to work around a device's buggy behavior).
>
> > I'm also confused why Windows malfunctions. It's not just that the device
> > doesn't work. Isn't the driver stack different if a hub is used? Perhaps
> > it's some kind of wierd interaction between device drivers.
>
> Assuming it's a 2.0 hub, then you'll be hitting the EHCI stack on the
> host side, yes.
>

I already did use a hardware bus analyzer. I put the analyzer between the
hub and the device and I didn't see anything suspicous other then the host
freezes communication midway for no reason. I put the analyzer between the
host and hub but couldn't measure anything since my analyzer is full speed
only. That's when I noticed that disabling the enhanced controller to force
the hub into full speed solves the problem.

I don't think the problem is the hub since I see the same problem on two
different hubs from two difference manufactures.

And like I said, the firmware on the device is independent of hubs beings
used or not. Since my microcontroller is a PIC 18F2550 it's possible it's a
silicon bug in the SIE but I kind of doubt it.

Is the problem here that usbser.sys just isn't very supported by Microsoft
because it isn't high priority. Has anybody else tried connecting a full
speed usbser.sys device to a host via a high speed hub? Unforntinatly, their
arn't very many usbser devices and I don't own any to try.

Armageddon
Fri May 02 12:31:05 CDT 2008

Is it possible for me to upload the captures and the portman trace somewhere?

chris
Fri May 02 13:28:59 CDT 2008

On May 2, 11:56 am, Armageddon <Armaged...@discussions.microsoft.com>
wrote:

> Is the problem here that usbser.sys just isn't very supported by Microsoft
> because it isn't high priority.

Look, no. Why are you blaming usbser.sys without any evidence? Why
do you think it's "not very supported" or "not a high priority"?

> Has anybody else tried connecting a full speed usbser.sys device to
> a host via a high speed hub?

Yes. I've personally done this hundreds, if not thousands of times
with no issues.

chris
Sun May 04 22:28:13 CDT 2008

On May 2, 12:31 pm, Armageddon <Armaged...@discussions.microsoft.com>
wrote:

> Is it possible for me to upload the captures and the portman trace somewhere?

By the way if you're talking about a CATC trace then yes go post it
somewhere, I'd be interested in seeing it.

Ben
Thu May 08 11:42:56 CDT 2008

Pavel A. wrote:
> "Armageddon" <Armageddon@discussions.microsoft.com> wrote in message
> news:D80B0827-3D9C-41E5-A470-036D4D6BA4EE@microsoft.com...
>
>> Doesn't the link from the hub to the device simply look like a
>> plain full speed link?
>
> No, the physical signaling is different in high speed mode (besides
> of the obvious timing differences)

The link from the full-speed device to the hub is the same as from
full-speed device to root port. The hub has to manage translating
everything to high-speed.

>
> --PA

Ben
Thu May 08 11:47:22 CDT 2008

> Is the problem here that usbser.sys just isn't very supported by
> Microsoft because it isn't high priority. Has anybody else tried
> connecting a full speed usbser.sys device to a host via a high speed
> hub? Unforntinatly, their arn't very many usbser devices and I don't
> own any to try.

We've seen various problems with different devices when connected through a
high-speed hub.

We're have some FTDI serial chips right now that don't use the usbser.sys
but will be evaling the Microchip PIC-based USB chips because the FTDI
drivers are plenty buggy.

You might additionally try putting a full-speed hub upstream of the
high-speed hub, and see what happens there. Probably the same as if you
forced the root port into full-speed by disabling EHCI.

Ben
Thu May 08 11:48:21 CDT 2008

chris.aseltine@gmail.com wrote:
> On May 2, 11:56 am, Armageddon <Armaged...@discussions.microsoft.com>
> wrote:
>
>> Is the problem here that usbser.sys just isn't very supported by
>> Microsoft because it isn't high priority.
>
> Look, no. Why are you blaming usbser.sys without any evidence? Why
> do you think it's "not very supported" or "not a high priority"?
>
>> Has anybody else tried connecting a full speed usbser.sys device to
>> a host via a high speed hub?
>
> Yes. I've personally done this hundreds, if not thousands of times
> with no issues.

Which chip is used for the device-side USB interface?