[PATCH v3] i2c: stm32f7: Add atomic_xfer method to driver

[Andi Shyti]

Hi Sean,

On Mon, Sep 04, 2023 at 07:29:59AM +0200, Sean Nyekjaer wrote:
> Hi Andy,
> 
> > On 3 Sep 2023, at 14.46, Andi Shyti <andi.shyti at kernel.org> wrote:
> > 
> > Hi Pierre-Yves, Alain,
> > 
> > mind taking a look here?
> > 
> > [...]
> > 
> >> @@ -357,6 +357,7 @@ struct stm32f7_i2c_dev {
> >> u32 dnf_dt;
> >> u32 dnf;
> >> struct stm32f7_i2c_alert *alert;
> >> + bool atomic;
> > 
> > this smells a bit racy here, this works only if the xfer's are
> > always sequential.
> > 
> > What happens when we receive at the same time two xfer's, one
> > atomic and one non atomic?
> 
> From the include/i2c.h:
>  * @master_xfer_atomic: same as @master_xfer. Yet, only using atomic context
>  *   so e.g. PMICs can be accessed very late before shutdown. Optional.
> 
> So it’s only used very late in the shutdown.
> 
> It’s implemented the same way as in:
> drivers/i2c/busses/i2c-imx.c
> drivers/i2c/busses/i2c-meson.c
> drivers/i2c/busses/i2c-mv64xxx.c
> drivers/i2c/busses/i2c-tegra.c
> … etc…
> 
> 
> In drivers/i2c/i2c-core.h it’s determined whether it’s atomic transfer or not:
> 
> /*
>  * We only allow atomic transfers for very late communication, e.g. to access a
>  * PMIC when powering down. Atomic transfers are a corner case and not for
>  * generic use!
>  */
> static inline bool i2c_in_atomic_xfer_mode(void)
> {
>         return system_state > SYSTEM_RUNNING && irqs_disabled();
> }
> 
> So you would not have an atomic transfer and later an non atomic.

What about the opposite? I.e. a non atomic and later an atomic,
for very late tardive communications :)

Andi