Conflicting PCI ID in bdc driver

Conflicting PCI ID in bdc driver

[Patrik Jakobsson]

Hi Al,
The PCI device ID of 0x1570 in the bdc driver (bdc_pci.c) is
conflicting with the Apple FacetimeHD webcam [1]. Is this caused by an
incorrect ID in the bdc driver or are there actually two devices with
the same ID in the wild? If we have a real conflict, how would we go
about solving this?

Thanks
Patrik Jakobsson

[1] https://github.com/patjak/bcwc_pcie

Re: Conflicting PCI ID in bdc driver

[Greg Kroah-Hartman]

On Mon, Jan 11, 2021 at 08:34:01PM +0100, Patrik Jakobsson wrote:
> Hi Al,
> The PCI device ID of 0x1570 in the bdc driver (bdc_pci.c) is
> conflicting with the Apple FacetimeHD webcam [1]. Is this caused by an
> incorrect ID in the bdc driver or are there actually two devices with
> the same ID in the wild? If we have a real conflict, how would we go
> about solving this?

Looks like someone at broadcom messed up :(

Can you look for any other fields in the device other than just the
vendor/device ids to verify that this really is a webcam so you can not
bind to the same thing this driver wants to bind to?

thanks,

greg k-h

Re: Conflicting PCI ID in bdc driver

[Patrik Jakobsson]

On Mon, Jan 11, 2021 at 9:29 PM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Mon, Jan 11, 2021 at 08:34:01PM +0100, Patrik Jakobsson wrote:
> > Hi Al,
> > The PCI device ID of 0x1570 in the bdc driver (bdc_pci.c) is
> > conflicting with the Apple FacetimeHD webcam [1]. Is this caused by an
> > incorrect ID in the bdc driver or are there actually two devices with
> > the same ID in the wild? If we have a real conflict, how would we go
> > about solving this?
>
> Looks like someone at broadcom messed up :(
>
> Can you look for any other fields in the device other than just the
> vendor/device ids to verify that this really is a webcam so you can not
> bind to the same thing this driver wants to bind to?

Right, we could check the class. But I suppose it must be fixed in
both drivers? bdc must not bind to the webcam and facetimehd must not
bind to the UDC. So which class is bdc reporting?
PCI_CLASS_SERIAL_USB_DEVICE?

>
> thanks,
>
> greg k-h

Re: Conflicting PCI ID in bdc driver

[Alan Cooper]

This driver was written for a PCI FPGA development board used to
verify the controller logic and to help with driver development before
adding the logic to our SoC's. I'm not sure why the driver was pushed
upstream but I'd like to remove it. I'm checking with a few other
groups to make sure I'm not missing anything.

Thanks
Al


On Mon, Jan 11, 2021 at 5:02 PM Patrik Jakobsson
<patrik.r.jakobsson@gmail.com> wrote:
>
> On Mon, Jan 11, 2021 at 9:29 PM Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> >
> > On Mon, Jan 11, 2021 at 08:34:01PM +0100, Patrik Jakobsson wrote:
> > > Hi Al,
> > > The PCI device ID of 0x1570 in the bdc driver (bdc_pci.c) is
> > > conflicting with the Apple FacetimeHD webcam [1]. Is this caused by an
> > > incorrect ID in the bdc driver or are there actually two devices with
> > > the same ID in the wild? If we have a real conflict, how would we go
> > > about solving this?
> >
> > Looks like someone at broadcom messed up :(
> >
> > Can you look for any other fields in the device other than just the
> > vendor/device ids to verify that this really is a webcam so you can not
> > bind to the same thing this driver wants to bind to?
>
> Right, we could check the class. But I suppose it must be fixed in
> both drivers? bdc must not bind to the webcam and facetimehd must not
> bind to the UDC. So which class is bdc reporting?
> PCI_CLASS_SERIAL_USB_DEVICE?
>
> >
> > thanks,
> >
> > greg k-h

Re: Conflicting PCI ID in bdc driver

[Patrik Jakobsson]

On Tue, Jan 12, 2021 at 2:06 AM Alan Cooper <alcooperx@gmail.com> wrote:
>
> This driver was written for a PCI FPGA development board used to
> verify the controller logic and to help with driver development before
> adding the logic to our SoC's. I'm not sure why the driver was pushed
> upstream but I'd like to remove it. I'm checking with a few other
> groups to make sure I'm not missing anything.

That would solve my problem. Is removing a driver acceptable for
stable submission? If not, it would be helpful to have a patch
suitable for stable that disables the driver before removal. I'm
somewhat tired of explaining to people how to blacklist bdc.

>
> Thanks
> Al
>
>
> On Mon, Jan 11, 2021 at 5:02 PM Patrik Jakobsson
> <patrik.r.jakobsson@gmail.com> wrote:
> >
> > On Mon, Jan 11, 2021 at 9:29 PM Greg Kroah-Hartman
> > <gregkh@linuxfoundation.org> wrote:
> > >
> > > On Mon, Jan 11, 2021 at 08:34:01PM +0100, Patrik Jakobsson wrote:
> > > > Hi Al,
> > > > The PCI device ID of 0x1570 in the bdc driver (bdc_pci.c) is
> > > > conflicting with the Apple FacetimeHD webcam [1]. Is this caused by an
> > > > incorrect ID in the bdc driver or are there actually two devices with
> > > > the same ID in the wild? If we have a real conflict, how would we go
> > > > about solving this?
> > >
> > > Looks like someone at broadcom messed up :(
> > >
> > > Can you look for any other fields in the device other than just the
> > > vendor/device ids to verify that this really is a webcam so you can not
> > > bind to the same thing this driver wants to bind to?
> >
> > Right, we could check the class. But I suppose it must be fixed in
> > both drivers? bdc must not bind to the webcam and facetimehd must not
> > bind to the UDC. So which class is bdc reporting?
> > PCI_CLASS_SERIAL_USB_DEVICE?
> >
> > >
> > > thanks,
> > >
> > > greg k-h

Re: Conflicting PCI ID in bdc driver

[Greg Kroah-Hartman]

On Tue, Jan 12, 2021 at 09:42:58AM +0100, Patrik Jakobsson wrote:
> On Tue, Jan 12, 2021 at 2:06 AM Alan Cooper <alcooperx@gmail.com> wrote:
> >
> > This driver was written for a PCI FPGA development board used to
> > verify the controller logic and to help with driver development before
> > adding the logic to our SoC's. I'm not sure why the driver was pushed
> > upstream but I'd like to remove it. I'm checking with a few other
> > groups to make sure I'm not missing anything.
> 
> That would solve my problem. Is removing a driver acceptable for
> stable submission?

Not really, if there are users of that driver.  Why not just mark the
config to depend on BROKEN instead?

But first we need to ensure that the users of it are really all gone.

> If not, it would be helpful to have a patch
> suitable for stable that disables the driver before removal. I'm
> somewhat tired of explaining to people how to blacklist bdc.

Why not submit your driver for inclusion in the kernel tree?  We can
make sure it doesn't conflict when that happens.

thanks,

greg k-h

Re: Conflicting PCI ID in bdc driver

[Alan Cooper]

I just checked with the hardware engineer that designed the BDC core.
The BDC PCI driver was used with an FPGA/PCIe board for design
verification and no one should be using this driver. I'll send a patch
that will remove this driver.

Thanks
Al

On Tue, Jan 12, 2021 at 3:57 AM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Tue, Jan 12, 2021 at 09:42:58AM +0100, Patrik Jakobsson wrote:
> > On Tue, Jan 12, 2021 at 2:06 AM Alan Cooper <alcooperx@gmail.com> wrote:
> > >
> > > This driver was written for a PCI FPGA development board used to
> > > verify the controller logic and to help with driver development before
> > > adding the logic to our SoC's. I'm not sure why the driver was pushed
> > > upstream but I'd like to remove it. I'm checking with a few other
> > > groups to make sure I'm not missing anything.
> >
> > That would solve my problem. Is removing a driver acceptable for
> > stable submission?
>
> Not really, if there are users of that driver.  Why not just mark the
> config to depend on BROKEN instead?
>
> But first we need to ensure that the users of it are really all gone.
>
> > If not, it would be helpful to have a patch
> > suitable for stable that disables the driver before removal. I'm
> > somewhat tired of explaining to people how to blacklist bdc.
>
> Why not submit your driver for inclusion in the kernel tree?  We can
> make sure it doesn't conflict when that happens.
>
> thanks,
>
> greg k-h

Re: Conflicting PCI ID in bdc driver

[Greg Kroah-Hartman]

On Tue, Jan 12, 2021 at 02:51:05PM -0500, Alan Cooper wrote:
> I just checked with the hardware engineer that designed the BDC core.
> The BDC PCI driver was used with an FPGA/PCIe board for design
> verification and no one should be using this driver. I'll send a patch
> that will remove this driver.

Based on the recent patches for this driver on the mailing list, are you
sure that no one is using this?  We can't delete a driver that people
are using.

thanks,

greg k-h

Re: Conflicting PCI ID in bdc driver

[Alan Cooper]

Just to be clear, I'm only interested in removing the bdc_pci.c file
and not the rest of the BDC driver.
I talked to the hardware engineer that designed the BDC core and he
said that there were only a few of these PCIe/FPGA boards. They were
used for development until the BDC core was added to various ARM SoC's
as a memory mapped platform bus controller. The original PCIe boards
are no longer in existence.

Al

On Wed, Jan 13, 2021 at 5:22 AM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Tue, Jan 12, 2021 at 02:51:05PM -0500, Alan Cooper wrote:
> > I just checked with the hardware engineer that designed the BDC core.
> > The BDC PCI driver was used with an FPGA/PCIe board for design
> > verification and no one should be using this driver. I'll send a patch
> > that will remove this driver.
>
> Based on the recent patches for this driver on the mailing list, are you
> sure that no one is using this?  We can't delete a driver that people
> are using.
>
> thanks,
>
> greg k-h

Re: Conflicting PCI ID in bdc driver

[Greg Kroah-Hartman]

On Wed, Jan 13, 2021 at 03:07:03PM -0500, Alan Cooper wrote:
> Just to be clear, I'm only interested in removing the bdc_pci.c file
> and not the rest of the BDC driver.

Ok, that's better, please send a patch for this.

thanks,

greg k-h