Re: [PATCH] watchdog: s3c2410_wdt: Report when the watchdog reset the system

[Tomasz Figa <t.figa@samsung.com>]

On Thursday 05 of December 2013 08:00:27 Guenter Roeck wrote:
> On Thu, Dec 05, 2013 at 01:27:13PM +0530, Leela Krishna Amudala wrote:
> > Hi Guenter Roeck,
> > 
> > On Tue, Dec 3, 2013 at 3:06 AM, Guenter Roeck <linux@roeck-us.net> wrote:
> > >
> > > On Mon, Dec 02, 2013 at 12:47:53PM -0800, Olof Johansson wrote:
> > > > On Mon, Dec 2, 2013 at 12:21 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> > > > > On Mon, Dec 02, 2013 at 10:14:41AM -0800, Doug Anderson wrote:
> > > > >> A good watchdog driver is supposed to report when it was responsible
> > > > >> for resetting the system.  Implement this for the s3c2410, at least on
> > > > >> exynos5250 and exynos5420 where we already have a pointer to the PMU
> > > > >> registers to read the information.
> > > > >>
> > > > >> Signed-off-by: Doug Anderson <dianders@chromium.org>
> > > > >> ---
> > > > >> This patch is based atop Leela Krishna's recent series that ends with
> > > > >> (ARM: dts: update watchdog device nodes for Exynos5250 and Exynos5420)
> > > > >> AKA <https://patchwork.kernel.org/patch/3251861/>.
> > > > >>
> > > > >>  drivers/watchdog/s3c2410_wdt.c | 42 +++++++++++++++++++++++++++++++++++++++---
> > > > >>  1 file changed, 39 insertions(+), 3 deletions(-)
> > > > >>
> > > > >> diff --git a/drivers/watchdog/s3c2410_wdt.c b/drivers/watchdog/s3c2410_wdt.c
> > > > >> index 47f4dcf..2c87d37 100644
> > > > >> --- a/drivers/watchdog/s3c2410_wdt.c
> > > > >> +++ b/drivers/watchdog/s3c2410_wdt.c
> > > > >> @@ -62,9 +62,13 @@
> > > > >>  #define CONFIG_S3C2410_WATCHDOG_ATBOOT               (0)
> > > > >>  #define CONFIG_S3C2410_WATCHDOG_DEFAULT_TIME (15)
> > > > >>
> > > > >> +#define RST_STAT_REG_OFFSET          0x0404
> > > > >>  #define WDT_DISABLE_REG_OFFSET               0x0408
> > > > >>  #define WDT_MASK_RESET_REG_OFFSET    0x040c
> > > > >>  #define QUIRK_NEEDS_PMU_CONFIG               (1 << 0)
> > > > >> +#define QUIRK_HAS_RST_STAT           (1 << 1)
> > > > >> +#define QUIRKS_NEED_PMUREG           (QUIRK_NEEDS_PMU_CONFIG | \
> > > > >> +                                      QUIRK_HAS_RST_STAT)
> > > > >>
> > > > > I am not really happy about the NEED (both of them, really) here.
> > > > > How about HAS instead ?
> > > >
> > > > Ah, I just commented on these things on our internal review site too
> > > > on this patch. I don't even think a quirk is needed -- just use the
> > > > presence of a non-0 rst_stat_reg to tell if you need to use regmap.
> > > >
> > > Agreed; same is true for the QUIRK_NEEDS_PMU_CONFIG related registers
> > > as well.
> > >
> > 
> > As Tomasz Figa suggested I introduced quirks,
> > 
> 'quirk' implies a workaround for non-standard or broken hardware,
> not a flag indicating device specific functionality.

I wouldn't limit meaning of "quirk" to only this. The word has been widely
used around the kernel as a name for differences between hardware
variants.

As for the original issue itself, IMHO Doug's original solution is the
most practical, as 0 can be a valid register offset and RST_STAT register
could be used for other purposes as well in future, depending on other
quirk flag.

Best regards,
Tomasz