11:09 <arnd> Esmil: ok, so an 8-byte access on this devices puts the lower 32 bits at the low address, which means lo-hi is the correct one.

11:22 <geertu> Esmil: and when adding support for the audio clock controller, the *q operations will be replaced by *l operations again?

11:25 <geertu> Perhaps that can be done now?

11:43 <Esmil> geertu: Yes, for the audio clocks the status register isn't 64bit aligned so I couldn't see a good alternative but to go back to the originl readl/writel implementation, but then we're back to the solution Andy didn't like, so I don't think we should sneak that in without proper review. So it's probably better to just add this this fix now and then I'll send in the audio clock/reset patches later

11:45 <Esmil> arnd: I see, thanks

19:02 <marex> Another next, another kernel BUG in stm32 crypto :-)

20:01 <marex> oh well ... the driver has been broken since 4.12 ... patch is out

20:02 <geertu> marex: you should have stayed on v4.9 ;-)

20:04 <marex> geertu: this wouldn't be a problem if Linux had stable driver ABI

20:12 <geertu> marex: run the old kernel in a guest?