suspend_stress_test is a shell script that asks powerd to suspend the system for a given number of seconds for configurable iterations. powerd then notifies every interested daemon about imminent suspend. Once all daemons are ready for suspend, powerd sets an RTC that fires after asked seconds and suspends either to idle (S0iX) or ram (S3) based on the configuration of the device. Thus this test helps us catch bugs in suspend/resume path of drivers and devices. This test (with memory_check flag) also helps in verifying that memory does not get corrupted during the suspend/resume process.
To run the test first flash a test image onto the device. Then execute the following command on the device.
# suspend_stress_test
Note that this test suspends to ram (S3) or idle (S0iX) based on existing configuration. Existing configuration can be verified using check_powerd_config.
# check_powerd_config --suspend_to_idle; echo $?
If the above command returns 0, the device will suspend to idle (S0iX) otherwise it will suspend to ram (S3). To configure the suspend state temporarily for test, execute the following command. Writing 1 to the below file makes sure device suspends to idle (S0iX). Writing 0 will make sure device suspends to ram (S3).
# echo 0/1 > /var/lib/power_manager/suspend_to_idle # restart powerd
This test by default runs 10000 cycles of suspend/resume. Number of iterations can be configured using count option.
# suspend_stress_test --count=2500
Duration of suspend can be configured using suspend_max and suspend_min options. Device spends random seconds between suspend_min and suspend_max in S3/S0iX. If configured to same value, device spends the exact specified time in suspend on every iteration.
# suspend_stress_test --count=2500 --suspend_min=5 --suspend_max=10
Note that this test generates several GB of logs. Most of the log includes suspend/resume times of individual drivers on every suspend. If you are not interested in suspend/resume times of individual drivers, you can turn that off using the nopm_print_times option. Note that this is not recommended as it will hide the exact driver that caused the suspend failure.
# suspend_stress_test --count=2500 --suspend_min=5 --suspend_max=10 --nopm_print_times
After running the test, check for any crashes in /var/spool/crash. Also check if all functionality such as display, touchpad, Keyboard, mouse, media playback etc. are still working as intended.
Simply pass memory_check option to suspend_stress_test script to verify that memory does not get corrupted during suspend/resume process.
# suspend_stress_test --count=2500 --suspend_min=5 --suspend_max=10 --memory_check
