Re: [PATCH] serial/of-serial: Add 16654 chip to compatible string list

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Mon, May 28, 2012 at 07:48:57PM +0200, Roland Stigge wrote:
> On 28/05/12 18:31, Russell King - ARM Linux wrote:
> >>>> So maybe 16750 is the better choice for me, anyway. Already supported in
> >>>> of-serial. Works for now, but need more testing. Another hint is that
> >>>> 16750 is advertised as "IP core for Soc" which matches the case of LPC32xx.
> >>>
> >>> 16750 also has automatic hardware flow control support, selectable through
> >>> bit 5 in the MCR register.  If your UART has that, then it's probably a
> >>> 16750 derivative rather than a 16550 or 16650 derivative.
> >>>
> >>> 16650s have an EFR register at offset 2, selectable by writing 0xBF into
> >>> the LCR register, which the 16750 doesn't have.  16650 also has automatic
> >>> hardware flow control, bit this is selected through a couple of bits in
> >>> the EFR.
> >>
> >> The 4 LPC32xx's "Standard" UARTs have neither of those.
> >>
> >> Is it ok to use "ns16650", i.e. PORT_16650, or do I need to introduce a
> >> FIFO depth configuration?
> > 
> > I think you need a new type, because as I said above, 16650s have that
> > additional EFR, and we will attempt to access that register which
> > isn't present in yours.
> 
> I actually meant 16750 instead of 16650, sorry, but this basically means
> the same - I would refer to extensions that are actually not there...
> 
> Now, introducing a new type, can I add to 8250.c's uart_config[] by
> introducing a new type (no. 22) after PORT_XR17D15X? Unfortunately,
> there are the "ARM specific type numbers" after current PORT_MAX_8250
> (21), but those are not listed in 8250.c's uart_config[]. Or how am I
> supposed to add a new type?

If it's 8250, stick it in with the group, otherwise the array will
become stupidly large.  That's why there's a separation of the two.