i.MX21 USB OTG

[Derald D. Woods]

On Sat, Mar 24, 2012 at 03:10:25PM +0100, Martin Fuzzey wrote:
> Hi Derald,
> 
> 
> On Fri, Mar 23, 2012 at 2:09 PM, Derald Woods <woods.rt at gmail.com> wrote:
> > Hello Martin,
> >
> > I have an existing design that uses the i.MX21 ARM9 processor. The
> > design currently uses an external USB OTG chip. The chip is now
> > end-of-life. I know that this is an older ARM processor. It still meets
> > most of the design needs. Is the i.MX21 OTG functionality considered
> > reliable? I have not seen any real success stories with the i.MX21 USB
> > OTG implementation. Our board currently utilizes the USB OTG chip at
> > the bootloader and Linux kernel level.
> >
> 
> It depends what you mean by OTG.
> 
> The i.MX21 silicon has 3 USB ports, two of which can only be used as
> USB hosts and one which is configurable as host only, function
> (device) only or OTG (dynamic switching).
> Somewhat confusingly the Freescale documentation refers to the whole
> module as "USB-OTG".
> 
> However the mainline code only supports host mode (there is no
> function driver nor OTG support)
> 
> I mainlined the i.MX21 HCD driver for 2.6.34 and fixed a few bugs for
> 2.6.37 (in particular problems with non aligned buffers causing usbnet
> to fail).
> 
> Since then I haven't had any bug reports, the driver also passes the
> USB test suite (which I updated to check the buffer alignment
> behaviour).
> 
> 
> > I would appreciate being pointed in the right direction or warned of
> > impending danger. Basically I want to know if the mx21 USB OTG has
> > performed well for other embedded Linux designs.
> >
> Something very close to the original code on which I based the driver
> was shipped with the Chumby (which uses  a heavily patched 2.6.16
> kernel). However that code had quite a few bugs and didn't support
> isochronous transfers at all.
> 
> I don't have any direct feedback myself of real world use of the
> driver however since we ended up not shipping an iMX21 based Linux
> product for non technical reasons.
> 
> > I had originally posted to the Barebox mailing list.
> >
> Ah I'm not subscribed to that list - adding it as a CC hoping it's not
> subscriber only
> 
> Cheers,
> 
> Martin Fuzzey

Hello Martin,


Thanks for the detailed feedback. This information is consistent with
what I have observed and followed with regard to the i.MX21. It will
help us chart a meaningful development course. 


Best regards,

Derald D. Woods