Re: Low speed BULK endpoints on UHCI

On Sat, 15 Sep 2007, Brad Campbell wrote:

> Manuel Stahl wrote:
> > I would also be very glad to see linux support that 'non standard' mode of 
> > bulk transfers for low speed devices. Maybe this could just be a parameter of 
> > the uhci module to switch it on or off.
> > 
> > Also I don't see how a device could starve the bus with bulk transfers, 
> > because they should only be handled if there is enough bandwidth free.
> > 
> > The only limitation the driver would have to check is the maximum size of 8 
> > bytes.
> 
> What about if I submitted a patch that had a module parameter to usbcore along the lines of 
> "deliberately_contravene_specifications" ? and we exported it such that it could be used to enable 
> such gross hacks as bulk transfers on low speed devices?
> 
> At the moment all I could see it doing is removing the endpoint remapping from bulk to interrupt and 
> removing the check for bulk submission in uhci, but I guess there are possible future uses where you 
> may want to prevent the stack from exactly following the specs to allow bug-for-bug compatibility 
> with the other desktop OS's.
> 
> If such a patch might be acceptable then I'll see about knocking something up. If not then my next 
> thought was to see if I could make the cdc-acm driver detect interrupt endpoints and use those in 
> lieu of bulk. But then that only works for a single driver. As we saw before with the MIDI 
> interface, there *are* other devices out there that do contravene the specs and do work on the other 
> OS's.

Your idea is okay with me.

Alan Stern


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel