On Tue, 22 Jan 2019 08:33:15 +0100, Sameer Pujar wrote:
Background
The device power management operations can be controlled with the help of runtime power management (runtime PM) framework. In this case hda driver can register runtime PM callbacks with the power management core (PM core). Idea is to put the device in low power mode, when it is not getting used. The clocks and power domains can be turned off when no use case is running on the device. Current patch series adds necessary support to manage hda device at runtime.
Change log
v1:
Patch-1: "ALSA: hda/tegra: runtime power management support" * runtime PM enable/disable added in device probe/remove * runtime PM calls are forbidden if AZX_DCAPS_PM_RUNTIME is not enabled in driver flags * worker thread uses pm_runtime_get_sync to invoke ->runtime_resume and pm_runtime_put to invoke ->runtime_suspend. The callbacks are added in subsequent patches
Patch-2: "ALSA: hda/tegra: get clock handles early in probe" * Clock handles required for hda are acquired early in the probe. This is done to make use of runtime PM framework to enable/disable clocks in the callbacks.
Patch-3: "ALSA: hda/tegra: add runtime PM callbacks" * adds runtime PM callbacks for ->runtime_suspend(), ->runtime_resume() * only skeleton for the callbacks is added, implementation is done in the subsequent patches
Patch-4: "ALSA: hda/tegra: remove redundant clock enable API" * As worked thread is calling pm_runtime_get_sync() now, no need to enable the clock explicitly. This can be managed from callback. * Idea is to have clock enable and disable functionalities in runtime PM callbacks and the driver can make use of the above. Any device access should be preceded by ->runtime_resume() call.
Patch-5: "ALSA: hda/tegra: implement runtime suspend/resume" * clock enable and controller initialization is moved to ->runtime_reesume() and the reverse is done during ->runtime_suspend() * Also during system wide power transitions, pm_runtime_force_resume() is invoked in system resume() and similarly for suspend case.
Patch-6: "ALSA: hda/tegra: fix kernel panic" * kernel panic is happening befoe controller init is done * added a flag "probed" to indicate the completion of device init * this flag is checked before doing any device access.
Patch-7: "ALSA: hda/tegra: add driver flag for runtime PM" * driver flag AZX_DCAPS_PM_RUNTIME is added if CONFIG_PM is enabled. * otherwise runtime PM calls will be forbidden and act as dummy calls
v1-->v2
Patch-1: "ALSA: hda/tegra: runtime power management support" * no change Patch-2: "ALSA: hda/tegra: get clock handles early in probe" * no change Patch-3: "ALSA: hda/tegra: add runtime PM callbacks" * no change Patch-4: "ALSA: hda/tegra: remove redundant clock enable API" * no change Patch-5: "ALSA: hda/tegra: implement runtime suspend/resume" * Patch-6(v1) commit is melded into this commit * instead of new flag "probed", chip->running status is used to know the completion of probe Patch-6: "ALSA: hda/tegra: add driver flag for runtime PM" * This commit is similar to Patch-7(v1) * AZX_DCAPS_PM_RUNTIME is unconditionally added and CONFIG_PM under which above flag was added is removed
=========
Sameer Pujar (6): ALSA: hda/tegra: runtime power management support ALSA: hda/tegra: get clock handles early in probe ALSA: hda/tegra: add runtime PM callbacks ALSA: hda/tegra: remove redundant clock enable API ALSA: hda/tegra: implement runtime suspend/resume ALSA: hda/tegra: add driver flag for runtime PM
Applied all patches now. Thanks.
Takashi