[RFC PATCH] tty: serial: Add 8250-core based omap driver

[sre@kernel.org (Sebastian Reichel)]

Hi,

On Thu, Jul 03, 2014 at 09:07:36AM -0500, Felipe Balbi wrote:
> On Thu, Jul 03, 2014 at 08:34:44AM -0500, Robert Nelson wrote:
> > On Thu, Jul 3, 2014 at 8:25 AM, Felipe Balbi <balbi@ti.com> wrote:
> > > On Thu, Jul 03, 2014 at 12:34:11AM -0700, Tony Lindgren wrote:
> > >> * Robert Nelson <robertcnelson@gmail.com> [140702 12:27]:
> > >> > On Wed, Jul 2, 2014 at 2:09 PM, Aaro Koskinen <aaro.koskinen@iki.fi> wrote:
> > >> > > Hi,
> > >> > >
> > >> > > On Wed, Jul 02, 2014 at 11:09:32AM -0500, Felipe Balbi wrote:
> > >> > >> > It has been only tested as console UART.
> > >> > >> > The tty name is ttyS based instead of ttyO. How big is the pain here,
> > >> > >> > what could be the easiest way to provide compatibility?
> > >> > >>
> > >> > >> have been considering that myself for months. You could pass an optional
> > >> > >> argument to serial8250_register_8250_port() but that only solves part of
> > >> > >> the problem :-(
> > >>
> > >> Some kind of compability layer sure would be nice.
> > >>
> > >> > > When ttyS -> ttyO change was done on OMAP, compatibility was not an issue.
> > >> > > Why should we care about it now?
> > >> >
> > >> > It would be a good opportunity to force everyone to update their bootloader. ;)
> > >> >
> > >> > Besides the BeagleBoard forum is quiet now, no one is complaining
> > >> > about that old (ttyS -> ttyO) transition anymore..
> > >>
> > >> How about a Kconfig option to provide ttyO by default? The not even
> > >> do that if kernel has cmdline option nottyomap.
> > >
> > > what about single zImage ? I don't want to use ttyO on my
> > > Allwinner/Exynos/Snapdragon/whatever SoC just because OMAP is in the
> > > same image ;-)
> > 
> > What if we just kept it simple, leave the ttyO driver enabled and add
> > a warning (pr_info) that it's deprecated.  It's not like it's broken,
> > it just won't get later features or devices support added.
> 
> Fine by me, I'd switch to 8250 as soon as it's merged though :-) would
> be nice to get an example DTS change just so I can start testing on the
> boards I have around.

DT is supposed to contain information about the hardware, so it
should stay the same? I think there is no non-hackish way to decide
at runtime which driver should be loaded.

One possible solution is:
 * Keep both drivers for a couple of kernel releases
 * Add the deprecation warning to the older one
 * Add a conflict between both drivers in Kconfig

Thus its decided at build-time, which driver should be used. This
would keep existing .config files working for a couple of releases.

-- Sebastian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20140703/b3a4290e/attachment.sig>