Re: [PATCH 1/2] dt-bindings: watchdog: Add arm,smc-wdt watchdog arm,smc-wdt compatible

[Guenter Roeck]

On Thu, Feb 20, 2020 at 05:41:09PM +1100, Evan Benn wrote:
> Dear Xingyu,
> 
> Could this driver also cover your usecase? I am not familiar with
> meson, but it seems like the meson calls could
> be replaced with arm_smccc calls. Then this driver will cover both
> chips. I am not sure if your firmware is upstream
> somewhere, but this might be adapted;
> https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/3405
> 
FWIW, the Meson driver has more functionality.

Guenter

> Thanks
> 
> 
> On Thu, Feb 20, 2020 at 10:20 AM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> >
> > On Wed, Feb 19, 2020 at 03:04:54PM -0800, Julius Werner wrote:
> > > > You are not the first 'watchdog in firmware accessed via an SMC call'.
> > > > Is there some more detail about what implementation this is? Part of
> > > > TF-A? Defined by some spec (I can dream)?
> > >
> > > This is just some random implementation written by me because we
> > > needed one. I would like it to be the new generic implementation, but
> > > it sounds like people here prefer the naming to be MediaTek specific
> > > (at least for now). The other SMC watchdog we're aware of is
> > > imx_sc_wdt but unfortunately that seems to hardcode platform-specific
> >
> > There is one more pending, for Meson SMC.
> >
> > https://patchwork.kernel.org/project/linux-watchdog/list/?series=227733
> >
> > Unfortunately it uses Meson firmware API functions, though it has pretty
> > much the same functionality since those ultimately end up calling
> > arm_smccc_smc().
> >
> > Guenter
> >
> > > details in the interface (at least in the pretimeout SMC) so we can't
> > > just expand that. With this driver I tried to directly wrap the kernel
> > > watchdog interface so it should be platform-agnostic and possible to
> > > expand this driver to other platforms later if desired. The SMC
> > > function ID would still always have to be platform-specific,
> > > unfortunately (but we could pass it in through the device tree), since
> > > the Arm SMC spec doesn't really leave any room for OS-generic SMCs
> > > like this.