Kalle Jokiniemi <kalle.jokiniemi@xxxxxxxxx> writes: > The runtime PM does not offer enough granularity to control > individual clocks separately as needed. Current pm implementation > of omap_wdt blocks the CORE power domain from idling in > OMAP SoC, as it keeps the interface clock of the watchdog always > on. This causes severe power leakage in devices where > the omap watchdog is in active use. The iclk remains active because it's in hardware-controlled auto-idle mode. I believe if you set the WDT hwmod to software-supported idle (OCPIF_SWSUP_IDLE), the runtime PM idle (which calls omap_device idle, then omap_hwmod_idle) should also gate the interface clock. > This patch set reverts the omap_wdt move to runtime_pm and > introduces a tight control of the interface clock in the driver > to only enable it when registers are accessed. The patch is > based on earlier patch by Atal Shargorodsky. > > If anyone has good suggestions how to fix this with runtime_pm, > feel free to propose patches. Otherwise I suggest we revert > back to something that works. Can you try with software-supported idle and just making the runtime PM calls more granular? Kevin > Tested with RX-51 S4.0 Macroboard, linux-omap-pm/pm branch and > MeeGo armv7l release. > > Kalle Jokiniemi (2): > Revert "OMAP: WDT: Use PM runtime APIs instead of clk FW APIs" > Watchdog: omap_wdt: fix interface clock handling > > drivers/watchdog/omap_wdt.c | 64 +++++++++++++++++++++++++++++++++++++------ > 1 files changed, 55 insertions(+), 9 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
