Hi Geert,
> > static inline bool i2c_in_atomic_xfer_mode(void)
> > {
> > return system_state > SYSTEM_RUNNING && irqs_disabled();
>
> So i2c atomic transfers are really meant to be used during late system suspend
> only, and not in atomic context before, when irqs_disabled() is true?
Yes. It is all some time ago, I recall we agreed that there shouldn't be
any other I2C communication at irqs_disabled() stage.
> Perhaps we need a checker config option, to make sure the atomic transfer
> operation is exercised at least once during boot?
Testing I2C controllers (in various ways) is a well-desired feature :/
> Do we need a way to let i2c slaves indicate they plan to use atomic
> transfers later, so the i2c core can keep the i2c controller resumed?
I wanted to have this originally, but in the end I gave up on it. IIRC,
you don't want to whitelist a client in general, but only the very late
messages it sends. However, if a message needs to be flagged may be
board specific. It all looked messy and hard to configure, so we ended
up with what we have now.
Take all this with a grain of salt, it's been a while since the
discussions happened.
All the best,
Wolfram
Attachment:
signature.asc
Description: PGP signature
