On 12/05/2011 02:01 AM, Mark Brown wrote:
Every device that implements runtime power management for DAIs is doing it in pretty much the same way: in the startup callback they take a runtime PM reference and then in the shutdown callback they release that reference, keeping the device active while the DAI is active. Given the frequency with which this is done and the obviousness of the need to keep the device active in this period factor the code out into the core, taking references on the device for each CPU DAI, CODEC DAI and DMA device in the core.
As runtime PM is reference counted this shouldn't interfere with any other reference holding by the drivers, and since (in common with the existing implementations) we don't check for errors on enabling it shouldn't matter if the device actually has runtime PM enabled or not.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
sound/soc/soc-pcm.c | 15 +++++++++++++++ 1 files changed, 15 insertions(+), 0 deletions(-)
diff --git a/sound/soc/soc-pcm.c b/sound/soc/soc-pcm.c index 49aa71e..8aa7cec 100644 --- a/sound/soc/soc-pcm.c +++ b/sound/soc/soc-pcm.c @@ -19,6 +19,7 @@ #include <linux/kernel.h> #include <linux/init.h> #include <linux/delay.h> +#include <linux/pm_runtime.h> #include <linux/slab.h> #include <linux/workqueue.h> #include <sound/core.h> @@ -77,6 +78,10 @@ static int soc_pcm_open(struct snd_pcm_substream *substream) struct snd_soc_dai_driver *codec_dai_drv = codec_dai->driver; int ret = 0;
- pm_runtime_get_sync(cpu_dai->dev);
- pm_runtime_get_sync(codec_dai->dev);
- pm_runtime_get_sync(platform->dev);
- mutex_lock_nested(&rtd->pcm_mutex, rtd->pcm_subclass);
I think it is better to move the pm_runtime_get_sync calls after the mutex_lock_nested() to be really safe (and to not change the way DAI drivers were handling the pm_runtime).