Hi, While testing the cpuidle on the OMAP3 platform, I observed that the usage statistics did not match the actual idle state transitions. This was happening due to two reasons: 1) The menu governor is selecting a "deep" idle state that cannot be attained. The system is able to attain one of the higher states; however, the statistics get updated for the "originally" selected state. 2) Some of the deep states can be defined as "invalid" at run-time via a sysfs entry "enable_off_mode". The since argument passed into the "enter_idle()" is pointer to the state selected by governor, it is possble to reach an invalid state. I am suggesting the following: 1) An additional flag for each idle state "valid". Any change in the "valid" flag is reflected back. 2) Governor takes this flag into account before selecting a target idle state. Looking for comments/suggestions. Here are link to this discussion on the linux-omap list: - http://marc.info/?l=linux-omap&m=125214460417989&w=2 - http://marc.info/?l=linux-omap&m=125260684907329&w=2 - http://marc.info/?l=linux-omap&m=125423861614898&w=2 Best regards, Sanjeev _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
