On Mon, Jan 10, 2022 at 09:24:38PM +0000, Robert Hancock wrote: > On Mon, 2022-01-10 at 20:31 +0000, Mark Brown wrote: > > If the device is using mclk-fs then either there's a fixed sample rate > > (in which case simple-card probably ought to force it without the driver > > worrying) or the sysclk will vary in which case simple-card should be > > setting the sysclk to 0 when the card goes idle to clear any > > constraints (which as you say later it does). > It sounds like to fix that, simple-card needs to keep track of whether the DAI > has a fixed clock rate or not. If it does, then it shouldn't be zeroing out the > sysclk when closing the device as that's wiping out information that needs to > be persistent. I guess if the frequency was read from a system-clock-frequency > property then we know it is fixed. There are other cases where it could be a > fixed rate, such as if the clock is connected to a fixed-clock. Maybe we need > an explicit DT flag "system-clock-fixed" or something for those cases? If the clock is fixed in the clock API we should arrange to be able to enumerate that from the clock API - we shouldn't be requiring redundant properties for something like that. > Then at least in the case where mclk-fs is set and one or more of the DAIs have > a fixed rate, simple-card can add a constraint to restrict the sample rate to > the fixed clock divided by mclk-fs? Yes, however we figure out that sysclk is fixed.
Attachment:
signature.asc
Description: PGP signature
