>> aren't there known issues with the tc you mentioned and SMP systems? > Don't know - but one possibility that would fit the symptoms would be > if there are multiple usage counters (one per cpu) and somehow the > values are just coming from one of them, and it isn't the right one.
Hmm...have you tried a non-MP kernel? Have you tried nailing your test down on each of the available CPUs? Have you tried running various numbers of infinite-loop CPU-burners at the same time? Each of those might hint at the actual problem. Maybe ktrace could use a "migrated to a different CPU" tracepoint? /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mo...@rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B