Mbed-os: Serious regression on NRF52 idle power consumption

[NeilMacMullen]

The current 5.9RC( and head of master) show very poor power consumption when executing

while (1) Thread::wait(5000);

The attached picture shows two traces on the same RF52 based hardware platform. Idle power has increased from ~180uA to 6.4mA suggesting that the processor is not idling at all.

[NeilMacMullen]

Adding @MarceloSalazar and @marcuschangarm as interested parties.

[ciarmcom]

ARM Internal Ref: MBOTRIAGE-596

[jrobeson]

related? https://github.com/ARMmbed/mbed-os/issues/7154

[kjbracey-arm]

related? #7154

Don't think so. The lock discussed there has been present since 5.7.

Perhaps this is relevant: https://github.com/ARMmbed/mbed-os/pull/7073

[0xc0170]

Perhaps this is relevant: #7073

👍 @maciejbocianski Can you confirm?

@NeilMacMullen Thanks for reporting

[maciejbocianski]

definitely, lack of #7073 cause this problem
SLEEP was disabled during HAL feature branches merging

[0xc0170]

@NeilMacMullen This should be fixed in 5.9.1 (3 patches are waiting for CI). You can cherry-pick 7073 (it has some dependencies on another 1-2 PR opened) to fix this and report back that is not causing any regression.

[NeilMacMullen]

I can confirm that https://github.com/ARMmbed/mbed-os/pull/7073/commits/f6d3058899636cfd22a8b6273f67e37fd55d24f9 does seem to fix the issue.