Re: [PATCH resend] Add sc16is7x2 driver

[Greg KH <greg@kroah.com>]

On Mon, Oct 18, 2010 at 11:46:38AM +0200, Manuel Stahl wrote:
> On 15.10.2010 18:49, Greg KH wrote:
> >On Fri, Oct 15, 2010 at 05:09:03PM +0200, Manuel Stahl wrote:
> >>On 15.10.2010 16:27, Greg KH wrote:
> >>>On Fri, Oct 15, 2010 at 09:24:34AM +0200, Manuel Stahl wrote:
> >>>>On 14.10.2010 21:01, Greg KH wrote:
> >>>>>On Thu, Oct 07, 2010 at 10:15:07AM +0200, Manuel Stahl wrote:
> >>>>>>--- /dev/null
> >>>>>>+++ b/include/linux/spi/sc16is7x2.h
> >>>>>>@@ -0,0 +1,17 @@
> >>>>>>+#ifndef LINUX_SPI_SC16IS752_H
> >>>>>>+#define LINUX_SPI_SC16IS752_H
> >>>>>>+
> >>>>>>+#define SC16IS7X2_NR_GPIOS 8
> >>>>>>+
> >>>>>>+struct sc16is7x2_platform_data {
> >>>>>>+	unsigned int	uartclk;
> >>>>>>+	/* uart line number of the first channel */
> >>>>>>+	unsigned	uart_base;
> >>>>>>+	/* number assigned to the first GPIO */
> >>>>>>+	unsigned	gpio_base;
> >>>>>>+	char		*label;
> >>>>>>+	/* list of GPIO names (array length = SC16IS7X2_NR_GPIOS) */
> >>>>>>+	const char	*const *names;
> >>>>>>+};
> >>>>>>+
> >>>>>>+#endif
> >>>>>
> >>>>>Why is this needed in a .h file?  What other .c file uses this define
> >>>>>and structure?
> >>>>>
> >>>>>thanks,
> >>>>>
> >>>>>greg k-h
> >>>>
> >>>>The board setup code can pass information about the uart and gpio
> >>>>enumeration.
> >>>
> >>>What "board setup code"?  I don't see that here in the patch anywhere.
> >>
> >>SPI devices are generally set up for a specific hardware board. I.e. for
> >>some ARM boards there is setup code in
> >>linux/arch/arm/mach-at91/board-<xyz>.c
> >>
> >>I use the sc16is7x2 for a board that is not yet in vanilla, so the board
> >>setup code is not included in the patch.
> >
> >Then why not add this header (or move it to the header file) when your
> >board setup code is merged?  No need to have it there now, right?
> 
> Cause everyone who want's to use this driver for his board needs the
> header. The board I use is a prototype, nearly the same as some
> Atmel Demo board, but with the sc16is752 attached to the SPI port.
> Therefore I don't see any benefit in publishing my board setup code.

How nice, you are saying that we need a .h file in the kernel for no
other reason than to support out-of-tree code?  Well, feel free to
resend this driver, and watch as the .h file is removed as it's not
needed by any in-tree code...

good luck,

greg k-h