Comment 5 for bug 1070755

The fact that display gets rotated but not input makes me think it's some flaw in the tegra3 driver's RANDR handling on resume. If this were on an open source driver, I'd suggest switching on DRI debug flags so we could detect if RANDR events are getting triggered on suspend, however I don't know if tegra provides that debug data in any form. It does not look like there are xorg.conf debugging options that would help.

gnome-power-manager on the other hand does have log data available. If someone can turn that on, reproduce this issue and then collect and attach that log here, it might give some clues.

It would also be interesting to see whether this can be triggered just by ordinary suspend/resume cycles or if it depends on the battery level and/or cord plugging. The g-p-m log might answer that. Or perhaps manually suspend/resume (via pm-suspend I guess?) and experiment with the power cord until you find a reliable reproducer.