Re: [PATCH] watchdog: Remove iop_wdt

[Arnd Bergmann]

On Tue, Nov 19, 2019 at 3:08 AM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
>
> On 11/18/19 2:04 PM, Laura Abbott wrote:
> >
> > Commit 59d3ae9a5bf6 ("ARM: remove Intel iop33x and iop13xx support")
> > removed support for some old platforms. Given this driver depends on
> > a now removed platform, just remove the driver.
> >
> > Signed-off-by: Laura Abbott <labbott@xxxxxxxxxx>
> > ---
> > Found this while reviewing config options. Not sure if this was kept
> > around for other reasons or just missed.
> > ---
> >   drivers/watchdog/Kconfig   |  16 ---
> >   drivers/watchdog/Makefile  |   1 -
> >   drivers/watchdog/iop_wdt.c | 249 -------------------------------------
> >   3 files changed, 266 deletions(-)
> >   delete mode 100644 drivers/watchdog/iop_wdt.c
> >
> > diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
> > index 58e7c100b6ad..fef9078a44b6 100644
> > --- a/drivers/watchdog/Kconfig
> > +++ b/drivers/watchdog/Kconfig
> > @@ -554,22 +554,6 @@ config PNX4008_WATCHDOG
> >
> >         Say N if you are unsure.
> >
> > -config IOP_WATCHDOG
> > -     tristate "IOP Watchdog"
> > -     depends on ARCH_IOP13XX
> > -     select WATCHDOG_NOWAYOUT if (ARCH_IOP32X || ARCH_IOP33X)
>
> This is a bit confusing, but it suggests that the watchdog may also work
> with ARCH_IOP32X, which is still supported. I don't know anything about
> those architectures, but I hesitate to have the driver removed unless
> we have confirmation that it won't work with ARCH_IOP32X.
> Maybe the dependency needs to be updated instead ?

See commit ec2e32ca661e ("watchdog: iop_wdt only builds for
mach-iop13xx") from 2014: the watdog hardware exists on iop32x
but the driver only successfully built on iop13xx, which is now gone.

Adding Russell (who said he still uses iop32x hardware) and Lennert
(who is still listed in the MAINTAINERS file, but previously said he
does not use it any more) to Cc. If neither of them see a reason to
keep the driver, I'd say we can remove it.

It seems unlikely that anyone wants to revive the driver if they have
not done it yet, and if they want to do it later, it is barely harder to revert
the removal than to fix the compile-time problem.

        Arnd