Re: [PATCH] watchdog: indydog: Add dependency on SGI_HAS_INDYDOG

[Guenter Roeck <linux@roeck-us.net>]

On Tue, Nov 14, 2017 at 02:48:12PM +0100, Ralf Baechle wrote:
> On Tue, Nov 14, 2017 at 11:17:38AM +0000, James Hogan wrote:
> 
> > On Tue, Nov 14, 2017 at 10:52:54AM +0000, Matt Redfearn wrote:
> > > Commit da2a68b3eb47 ("watchdog: Enable COMPILE_TEST where possible")

Yes, that patch was a tremenuous failure and definitely not worth the trouble
it caused.

> > > enabled building the Indy watchdog driver when COMPILE_TEST is enabled.
> > > However, the driver makes reference to symbols that are only defined for
> > > certain platforms are selected in the config. These platforms select
> > > SGI_HAS_INDYDOG. Without this, link time errors result, for example
> > > when building a MIPS allyesconfig.
> > > 
> > > drivers/watchdog/indydog.o: In function `indydog_write':
> > > indydog.c:(.text+0x18): undefined reference to `sgimc'
> > > indydog.c:(.text+0x1c): undefined reference to `sgimc'
> > > drivers/watchdog/indydog.o: In function `indydog_start':
> > > indydog.c:(.text+0x54): undefined reference to `sgimc'
> > > indydog.c:(.text+0x58): undefined reference to `sgimc'
> > > drivers/watchdog/indydog.o: In function `indydog_stop':
> > > indydog.c:(.text+0xa4): undefined reference to `sgimc'
> > > drivers/watchdog/indydog.o:indydog.c:(.text+0xa8): more undefined
> > > references to `sgimc' follow
> > > make: *** [Makefile:1005: vmlinux] Error 1
> > > 
> > > Fix this by ensuring that CONFIG_INDIDOG can only be selected when the
> > > necessary dependent platform symbols are built in.
> > > 
> > > Fixes: da2a68b3eb47 ("watchdog: Enable COMPILE_TEST where possible")
> > > Signed-off-by: Matt Redfearn <matt.redfearn@mips.com>
> > > Cc: <stable@vger.kernel.org> # 4.11 +
> > > ---
> > > 
> > >  drivers/watchdog/Kconfig | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
> > > index ca200d1f310a..d96e2e7544fc 100644
> > > --- a/drivers/watchdog/Kconfig
> > > +++ b/drivers/watchdog/Kconfig
> > > @@ -1451,7 +1451,7 @@ config RC32434_WDT
> > >  
> > >  config INDYDOG
> > >  	tristate "Indy/I2 Hardware Watchdog"
> > > -	depends on SGI_HAS_INDYDOG || (MIPS && COMPILE_TEST)
> > > +	depends on SGI_HAS_INDYDOG || (MIPS && COMPILE_TEST && SGI_HAS_INDYDOG)
> > 
> > (MIPS && COMPILE_TEST && SGI_HAS_INDYDOG) implies SGI_HAS_INDYDOG
> > 
> > So I think you can just do:
> > -	depends on SGI_HAS_INDYDOG || (MIPS && COMPILE_TEST)
> > +	depends on SGI_HAS_INDYDOG
> > 
> > I.e. COMPILE_TEST isn't of any value in this case.
> 
> I agree, due to the references to sgimc this driver will only compile for
> the platforms which set SGI_HAS_INDYDOG; MIPS as the dependency is too
> generic.
> 
> Updated patch for the watchdog maintainers' ease below.
> 
>   Ralf
> 
> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
> Reported-by: Matt Redfearn <matt.redfearn@mips.com>
> Suggested-by: James Hogan <james.hogan@mips.com>
> 
Not sure if/how that helps. Now we have two patches with different
authors and sign-off statements, both of which need maintainer
manipulation. I'll pick the first patch into my tree and fix it up.
Let's see what Wim does with it.

Guenter

>  drivers/watchdog/Kconfig | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
> index c722cbfdc7e6..3ece1335ba84 100644
> --- a/drivers/watchdog/Kconfig
> +++ b/drivers/watchdog/Kconfig
> @@ -1451,7 +1451,7 @@ config RC32434_WDT
>  
>  config INDYDOG
>  	tristate "Indy/I2 Hardware Watchdog"
> -	depends on SGI_HAS_INDYDOG || (MIPS && COMPILE_TEST)
> +	depends on SGI_HAS_INDYDOG
>  	help
>  	  Hardware driver for the Indy's/I2's watchdog. This is a
>  	  watchdog timer that will reboot the machine after a 60 second