* Andreas Kemnade andreas@kemnade.info [231006 19:30]:
On Fri, 6 Oct 2023 13:23:48 +0300 Tony Lindgren tony@atomide.com wrote:
Here's what I think the regression fix for omap4 clocks would be, the old main_clk is not the same as the module clock that we get by default. If this looks OK I'll do a similar fix also for omap5.
Or is something else also needed?
hmm, audio output works, the waring is away, but something new is here:
OK good to hear it works, I'll send out fixes for omap4 and 5, seems the runtime PM warning is something different.
omap-mcbsp 40124000.mcbsp: Runtime PM usage count underflow! # cat /sys/bus/platform/devices/40124000.mcbsp/power/runtime_status active
even with no sound.
I guess if the mcbsp instance is not used, runtime PM is enabled but pm_runtime_resume_and_get() is never called by ASoC?
If so then the following might be a fix, not familiar with runtime PM done by ASoC though and not sure if some kind of locking would be needed here.
Regards,
Tony
8< ------------------------ diff --git a/sound/soc/ti/omap-mcbsp.c b/sound/soc/ti/omap-mcbsp.c index fdabed5133e83..b399d86f22777 100644 --- a/sound/soc/ti/omap-mcbsp.c +++ b/sound/soc/ti/omap-mcbsp.c @@ -74,14 +74,16 @@ static int omap2_mcbsp_set_clks_src(struct omap_mcbsp *mcbsp, u8 fck_src_id) return 0; }
- pm_runtime_put_sync(mcbsp->dev); + if (mcbsp->active) + pm_runtime_put_sync(mcbsp->dev);
r = clk_set_parent(mcbsp->fclk, fck_src); if (r) dev_err(mcbsp->dev, "CLKS: could not clk_set_parent() to %s\n", src);
- pm_runtime_get_sync(mcbsp->dev); + if (mcbsp->active) + pm_runtime_get_sync(mcbsp->dev);
clk_put(fck_src);