| Message ID | 20190612164726.26768-3-pierre-louis.bossart@linux.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | ASoC: SOF: suspend/resume debug tools | expand |
On Wed, Jun 12, 2019 at 11:47:25AM -0500, Pierre-Louis Bossart wrote: > Add debug option to disable pm_runtime. This is not intended for > production devices but is very useful for platform bringup. I can't immediately find it right now but isn't there some generic way of doing this in the runtime PM framework? If not it seems like it'd be a good thing to add, these can't be the only devices where it'd be useful.
On Thu, 13 Jun 2019 20:48:01 +0200, Mark Brown wrote: > > On Wed, Jun 12, 2019 at 11:47:25AM -0500, Pierre-Louis Bossart wrote: > > Add debug option to disable pm_runtime. This is not intended for > > production devices but is very useful for platform bringup. > > I can't immediately find it right now but isn't there some generic way > of doing this in the runtime PM framework? If not it seems like it'd be > a good thing to add, these can't be the only devices where it'd be > useful. Well, runtime PM can be fully controlled via sysfs, but the problem is that the driver declares itself being runtime-enabled. So, either we leave it default and let user-space enabling it (via udev or other way), or introduce some condition in the driver side. Takashi
On Thu, Jun 13, 2019 at 09:04:28PM +0200, Takashi Iwai wrote: > Mark Brown wrote: > > I can't immediately find it right now but isn't there some generic way > > of doing this in the runtime PM framework? If not it seems like it'd be > > a good thing to add, these can't be the only devices where it'd be > > useful. > Well, runtime PM can be fully controlled via sysfs, but the problem is > that the driver declares itself being runtime-enabled. > So, either we leave it default and let user-space enabling it (via > udev or other way), or introduce some condition in the driver side. I thought someone had added a command line parameter to do it based on dev_name(), perhaps they were just talking about it or it was in some BSP somewhere though.
>>> I can't immediately find it right now but isn't there some generic way >>> of doing this in the runtime PM framework? If not it seems like it'd be >>> a good thing to add, these can't be the only devices where it'd be >>> useful. > >> Well, runtime PM can be fully controlled via sysfs, but the problem is >> that the driver declares itself being runtime-enabled. >> So, either we leave it default and let user-space enabling it (via >> udev or other way), or introduce some condition in the driver side. > > I thought someone had added a command line parameter to do it based on > dev_name(), perhaps they were just talking about it or it was in some > BSP somewhere though. If there is a better way I am all ears. It's indeed not very elegant to duplicate the same parameter for two different modules and it's not an SOF-specific need. The only way I am aware of is to play with /sys/bus/pci/devices/xyz/power/ files but it's not very useful if you want to disable the initial runtime pm transition which is often the more problematic one. Completely removing runtime_pm support from all drivers at compile time is also not very good either.
diff --git a/sound/soc/sof/sof-acpi-dev.c b/sound/soc/sof/sof-acpi-dev.c index c8dafb1ac54e..93a8e15bbd2c 100644 --- a/sound/soc/sof/sof-acpi-dev.c +++ b/sound/soc/sof/sof-acpi-dev.c @@ -29,6 +29,12 @@ static char *tplg_path; module_param(tplg_path, charp, 0444); MODULE_PARM_DESC(tplg_path, "alternate path for SOF topology."); +static int sof_acpi_debug; +module_param_named(sof_debug, sof_acpi_debug, int, 0444); +MODULE_PARM_DESC(sof_debug, "SOF ACPI debug options (0x0 all off)"); + +#define SOF_ACPI_DISABLE_PM_RUNTIME BIT(0) + #if IS_ENABLED(CONFIG_SND_SOC_SOF_HASWELL) static const struct sof_dev_desc sof_acpi_haswell_desc = { .machines = snd_soc_acpi_intel_haswell_machines, @@ -121,6 +127,9 @@ static const struct dev_pm_ops sof_acpi_pm = { static void sof_acpi_probe_complete(struct device *dev) { + if (sof_acpi_debug & SOF_ACPI_DISABLE_PM_RUNTIME) + return; + /* allow runtime_pm */ pm_runtime_set_autosuspend_delay(dev, SND_SOF_SUSPEND_DELAY_MS); pm_runtime_use_autosuspend(dev); @@ -221,7 +230,8 @@ static int sof_acpi_probe(struct platform_device *pdev) static int sof_acpi_remove(struct platform_device *pdev) { - pm_runtime_disable(&pdev->dev); + if (!(sof_acpi_debug & SOF_ACPI_DISABLE_PM_RUNTIME)) + pm_runtime_disable(&pdev->dev); /* call sof helper for DSP hardware remove */ snd_sof_device_remove(&pdev->dev);
Add debug option to disable pm_runtime. This is not intended for production devices but is very useful for platform bringup. Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> --- sound/soc/sof/sof-acpi-dev.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)