> -----Original Message----- > From: Vinod Koul <vkoul@xxxxxxxxxx> > Sent: Wednesday, November 25, 2020 1:05 PM > To: Bard Liao <yung-chuan.liao@xxxxxxxxxxxxxxx> > Cc: alsa-devel@xxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; > gregkh@xxxxxxxxxxxxxxxxxxx; jank@xxxxxxxxxxx; > srinivas.kandagatla@xxxxxxxxxx; rander.wang@xxxxxxxxxxxxxxx; > ranjani.sridharan@xxxxxxxxxxxxxxx; hui.wang@xxxxxxxxxxxxx; pierre- > louis.bossart@xxxxxxxxxxxxxxx; Kale, Sanyog R <sanyog.r.kale@xxxxxxxxx>; Lin, > Mengdong <mengdong.lin@xxxxxxxxx>; Liao, Bard <bard.liao@xxxxxxxxx> > Subject: Re: [PATCH] soundwire: master: use pm_runtime_set_active() on > add > > On 24-11-20, 21:07, Bard Liao wrote: > > From: Pierre-Louis Bossart <pierre-louis.bossart@xxxxxxxxxxxxxxx> > > > > The 'master' device acts as a glue layer used during bus > > initialization only, and it needs to be 'transparent' for pm_runtime > > management. Its behavior should be that it becomes active when one of > > its children becomes active, and suspends when all of its children are > > suspended. > > > > In our tests on Intel platforms, we routinely see these sort of > > warnings on the initial boot: > > > > [ 21.447345] rt715 sdw:3:25d:715:0: runtime PM trying to activate > > child device sdw:3:25d:715:0 but parent (sdw-master-3) is not active > > > > This is root-caused to a missing setup to make the device 'active' on > > probe. Since we don't want the device to remain active forever after > > the probe, the autosuspend configuration is also enabled at the end of > > the probe - the device will actually autosuspend only in the case > > where there are no devices physically attached. In practice, the > > master device will suspend when all its children are no longer active. > > > > Fixes: bd84256e86ecf ('soundwire: master: enable pm runtime') > > Signed-off-by: Pierre-Louis Bossart > > <pierre-louis.bossart@xxxxxxxxxxxxxxx> > > Reviewed-by: Rander Wang <rander.wang@xxxxxxxxxxxxxxx> > > Signed-off-by: Bard Liao <yung-chuan.liao@xxxxxxxxxxxxxxx> > > --- > > drivers/soundwire/master.c | 14 ++++++++++++++ > > 1 file changed, 14 insertions(+) > > > > diff --git a/drivers/soundwire/master.c b/drivers/soundwire/master.c > > index 3488bb824e84..9b05c9e25ebe 100644 > > --- a/drivers/soundwire/master.c > > +++ b/drivers/soundwire/master.c > > @@ -8,6 +8,15 @@ > > #include <linux/soundwire/sdw_type.h> #include "bus.h" > > > > +/* > > + * The 3s value for autosuspend will only be used if there are no > > + * devices physically attached on a bus segment. In practice enabling > > + * the bus operation will result in children devices become active > > +and > > + * the master device will only suspend when all its children are no > > + * longer active. > > + */ > > +#define SDW_MASTER_SUSPEND_DELAY_MS 3000 > > + > > /* > > * The sysfs for properties reflects the MIPI description as given > > * in the MIPI DisCo spec > > @@ -154,7 +163,12 @@ int sdw_master_device_add(struct sdw_bus *bus, > struct device *parent, > > bus->dev = &md->dev; > > bus->md = md; > > > > + pm_runtime_set_autosuspend_delay(&bus->md->dev, > SDW_MASTER_SUSPEND_DELAY_MS); > > + pm_runtime_use_autosuspend(&bus->md->dev); > > + pm_runtime_mark_last_busy(&bus->md->dev); > > + pm_runtime_set_active(&bus->md->dev); > > pm_runtime_enable(&bus->md->dev); > > + pm_runtime_idle(&bus->md->dev); > > I understand that this needs to be done somewhere but is the core the right > place. In intel case it maybe a dummy device but other controllers are real > devices and may not support pm. > > I think better idea would be to do this in respective driver.. that way it > would be an opt-in for device supporting pm. Should it be put in the same place as pm_runtime_enable? IMHO, pm_runtime_enable is in the core already and it seems to be harmless for devices which don't support pm. And pm can still be optional on md's parent device. > > Thanks > -- > ~Vinod