On 25/06/25 12:24 pm, Maxime Ripard wrote: > On Wed, Jun 25, 2025 at 10:26:10AM +0530, Dharma Balasubiramani wrote: >> Replace legacy .enable and .disable callbacks with their atomic >> counterparts .atomic_enable and .atomic_disable. >> >> Also, add turn off the serialiser inside atomic_disable(). >> >> Signed-off-by: Dharma Balasubiramani <[email protected]> > > As a rule of thumb, if you have "Also, do X" in your commit log, you > need a separate patch. > > And you need to explain why turning off the serialiser inside > atomic_disable() is needed. It might make sense to you, it's not really > obvious to me from that patch, and it will definitely not be to someone > trying to identify fixes and doing backports.
I initially introduced the turning off the serialiser to avoid having an empty disable() hook. Now that you've clarified it's perfectly fine to sleep in these contexts, I no longer see the need for the split. I'll drop both atomic_pre_enable(), atomic_post_disable() and turning off the serialiser as well and just keep just atomic_enable() and atomic_disable(). Thanks. > > Maxime -- With Best Regards, Dharma B.
