Sergei Ianovich [off-list ref] writes:

On Sun, 2015-12-20 at 00:12 +0100, Robert Jarzmik wrote:

quoted

Sergei Ianovich [off-list ref] writes:

quoted

On Sat, 2015-12-19 at 20:31 +0100, Robert Jarzmik wrote:
I understand that people are afraid of taking this patch. If it
starts
causing troubles at runtime, it will be difficult to diagnose. There
will be no console for most people. So it is probably good idea to
fail
at boot time.

If it's about something already written in a mailing
list, please point me to it so that it can help me think about it.

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-December/2167
73.html

I can explain why I think so. Greg acked the patch, but hasn't merged it
since then. He has good reasons for this most probably. Russell's
comment pointed by the link seems to be the reason.

I think the problem raised by Russell could be addressed. My best guess
is compile time error, despite your comment above.

I re-read carefully Russell's answer in [1].

What Russell asked is that for a period of time, the old pxa serial code and the
new will be in the kernel, so that maintainers have the option to switch over to
the old drivers/tty/serial/pxa.c if the new 8250 based version breaks their
userspace getty.

Then, once the transition is done, and that for a period (let's say 1 year) no
maintainer had complained while its defconfig was switched over to the new 8520
version, then and only then you will remove drivers/tty/serial/pxa.c.

I have one more plan. For transition period, we can introduce a
temporary Kconfig option SERIAL_8250_PXA_OFF, and fail at build time if
neither SERIAL_8250_PXA nor SERIAL_8250_PXA_OFF is set. This way all
interested parties will be notified of this driver update.

No, I'd like to stick with Russell's original plan :
 - phase 1: both SERIAL_8250_PXA and SERIAL_PXA exist in KConfig
            both are selectable
            This lasts one year or something like that
 - phase 2: remove SERIAL_PXA from KConfig and drivers/tty/serial/pxa.c

This means a different patch from the one acked by Greg, and a new serie of
acks. The diffstat will be way worse (as you won't have the -970 for pxa.c), but
in the end it will end up on that -970.

That sounds like a good transition plan to me.

Cheers.

-- 
Robert

[1]
http://lists.infradead.org/pipermail/linux-arm-kernel/2013-December/216773.html