I looked at the USB (2.0) spec, section 10.3.1 is what
seems relevant here.  It does specify that different
operating systems may do things in different ways.

I wonder if you're not thinking an OS-dependency
into your device firmware... SET_CONFIG doesn't
seem appropriate to trigger such behavior, that'd
be more appropriate in some vendor control
messages or other interface-specific requests.

On Linux, the hub driver initially sets devices into
a default configuration (first one) and then tries to
find a driver for each interface.  Such drivers can
change configurations if they need to.  They're also
responsible for choosing alternate interface settings,
typically as they're probed.


Now, I was recently reminded that there seem to
be some problems with how that works.  If one
driver changes the device configuration, trouble
could ensure for other drivers using its interfaces.
(I don't think we've seen a device yet that would
have such problems.)

But in general I think usb_set_configuration() is
something that needs to be done with caution
today.  It doesn't disconnect() currently bound
interface drivers, and doesn't restart any sort of
"find interface drivers" process.

Somewhere in 2.5 I think that needs to be fixed.

My first thought is that usb_set_configuration()
should be (a) disconnecting drivers, and (b) later
trying to find drivers for the interfaces.  Which
means that (c) the hub driver shouldn't try to
find the drivers, since it'd happen when it sets
the default configuration, and (d) that "find a
driver" process should notice such changes to
the device configuration, and behave sanely...

- Dave


----- Original Message ----- 
From: <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Friday, November 30, 2001 1:00 AM
Subject: [linux-usb-devel] SET_CONFIG


> Hello, linux-usb-devel!
> 
> Our devices are "vendor specific" ISDN devices, *not* CDC devices
> and receive a SET_CONFIGURATION after plugin and long time before
> the driver has determined if and when to send the first SET_CONFIG
> which is of certain importance for one of the devices. (This one
> loads firmware and a sequence of SET_CONFIGs is used to start and
> finish loading...)
> 
> The question is: Where in the specification can I find more details
> on this phenomenon in order to handle this early SET_CONFIG appro-
> priately? (This can not be seen in Windows or MacOS, BTW.) What was
> the reason for having the USB system configure the device instead
> of the driver?
> 
> With best regards,
> Olaf Stoyke
> 
> 



_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to