Re: [PATCH v2 3/8] watchdog: JZ4740: Register a restart handler

[Guenter Roeck]

On 01/20/2018 08:04 AM, PrasannaKumar Muralidharan wrote:
> Hi Guenter,
>
> On 20 January 2018 at 21:15, Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> > On 01/19/2018 11:31 PM, PrasannaKumar Muralidharan wrote:
> > > Hi Paul,
> > >
> > > On 30 December 2017 at 19:21, Paul Cercueil <paul@xxxxxxxxxxxxxxx> wrote:
> > > > The watchdog driver can restart the system by simply configuring the
> > > > hardware for a timeout of 0 seconds.
> > > >
> > > > Signed-off-by: Paul Cercueil <paul@xxxxxxxxxxxxxxx>
> > > > Reviewed-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> > > > ---
> > > >    drivers/watchdog/jz4740_wdt.c | 9 +++++++++
> > > >    1 file changed, 9 insertions(+)
> > > >
> > > >    v2: No change
> > > >
> > > > diff --git a/drivers/watchdog/jz4740_wdt.c
> > > > b/drivers/watchdog/jz4740_wdt.c
> > > > index 92d6ca8ceb49..fa7f49a3212c 100644
> > > > --- a/drivers/watchdog/jz4740_wdt.c
> > > > +++ b/drivers/watchdog/jz4740_wdt.c
> > > > @@ -130,6 +130,14 @@ static int jz4740_wdt_stop(struct watchdog_device
> > > > *wdt_dev)
> > > >           return 0;
> > > >    }
> > > >
> > > > +static int jz4740_wdt_restart(struct watchdog_device *wdt_dev,
> > > > +                             unsigned long action, void *data)
> > > > +{
> > > > +       wdt_dev->timeout = 0;
> > > > +       jz4740_wdt_start(wdt_dev);
> > > > +       return 0;
> > > > +}
> > > > +
> > > >    static const struct watchdog_info jz4740_wdt_info = {
> > > >           .options = WDIOF_SETTIMEOUT | WDIOF_KEEPALIVEPING |
> > > > WDIOF_MAGICCLOSE,
> > > >           .identity = "jz4740 Watchdog",
> > > > @@ -141,6 +149,7 @@ static const struct watchdog_ops jz4740_wdt_ops = {
> > > >           .stop = jz4740_wdt_stop,
> > > >           .ping = jz4740_wdt_ping,
> > > >           .set_timeout = jz4740_wdt_set_timeout,
> > > > +       .restart = jz4740_wdt_restart,
> > > >    };
> > > >
> > > >    #ifdef CONFIG_OF
> > > > --
> > > > 2.11.0
> > >
> > > Noticed that min_timeout of the watchdog device is set to 1 but this
> > > function calls start with timeout set to 0. Even though this works I
> > > feel it is better to set min_timeout to 0.
> >
> > No. That would be wrong. If you want to be pedantic, write a new function
> > __jz4740_wdt_set_timeout(u16 clock_div, u16 timeout_value) and call it
> > instead, but don't mess with min_timeout.
> >
> > Guenter
>
> What is the effect of changing min_timeout? I could see only
> validation checks with it.
Let me ask you - you want to let userspace set a timeout of 0, which would
effectively reboot the system through the backdoor. Why ?

Guenter