On Fri, May 24, 2019 at 03:52:37PM +0200, Wolfram Sang wrote:
> On Mon, Apr 15, 2019 at 12:52:01PM +0200, Wolfram Sang wrote:
> > Support an already running watchdog by checking its enable bit and set
> > up the status accordingly before registering the device.
> >
> > Signed-off-by: Wolfram Sang <wsa+renesas@xxxxxxxxxxxxxxxxxxxx>
>
> After second thought, I am getting confused a little. If the WDT is
> already running then
>
> a) before this patch: after successful probe, RPM will disable the
> clock until userspace opens the watchdog device
>
> b) after this patch: during probe, our default timeout will be
> programmed and because of WDOG_HW_RUNNING, the core will generate pings
> until userspace opens the watchdog device.
>
> So, b) will protect from a crashing kernel (no pings anymore) but not
> from something like missing rootfs, or?
>
> The usecase I had in mind ("give the kernel <x> seconds to boot into
> working userspace") seems to be achieved by loading the WDT driver as a
> module then, I guess?
>
Would
https://lore.kernel.org/linux-watchdog/20190605140628.618-1-rasmus.villemoes@xxxxxxxxx/
solve your use case ?
Guenter
