Comment # 9 on bug 996543 from Bernhard Wiedemann

so since it did not crash every time, I made a loop around the test

lenovo:~ # cat /sys/power/pm_test
none core processors platform devices [freezer]
lenovo:~ # echo mem > /sys/power/state 
lenovo:~ # for i in 1 2 3 4 5 ; do echo mem > /sys/power/state ; sleep 5 ; done
lenovo:~ # echo devices > /sys/power/pm_test
lenovo:~ # for i in 1 2 3 4 5 6 7 8 ; do echo mem > /sys/power/state ; sleep 5
; done
lenovo:~ # echo platform > /sys/power/pm_test
lenovo:~ # for i in 1 2 3 4 5 6 7 8 ; do echo mem > /sys/power/state ; sleep 5
; done
lenovo:~ # echo processors > /sys/power/pm_test
lenovo:~ # for i in 1 2 3 4 5 6 7 8 ; do echo mem > /sys/power/state ; sleep 5
; done
lenovo:~ # echo core > /sys/power/pm_test
lenovo:~ # for i in 1 2 3 4 5 6 7 8 ; do echo mem > /sys/power/state ; sleep 5
; done

but now it passed all of them.
As a side-note: each suspend caused 8 or 9 pings to be lost.

only with none it crashed (on first try even)

Could still be an esoteric HW-problem... e.g. touching the power button in a
certain way on resume causes a voltage to be introduced somewhere it should not
be... 
or the time of the sleep state matters with DRAM capacitors discharging over
time.



Then I also tried alternating
echo freeze > /sys/power/state
and
echo disk > /sys/power/state

and found that freeze did something on some tries
but returned device-or-resource-busy on other tries
and suspend-to-disk worked fine three times in a row.
None of these crashed the laptop.