On 12/12/2013 05:25 AM, Ingo Molnar wrote: > * John Stultz <john.stultz@xxxxxxxxxx> wrote: >> A think-o in the calculation of the monotonic -> tai time offset >> results in CLOCK_TAI timers and nanosleeps to expire late (the >> latency is ~2x the tai offset). >> >> Fix this by adding the tai offset from the realtime offset instead >> of subtracting. > Hm, it looks like the whole CLOCK_TAI feature was rushed in, with not > enough testing done. I wouldn't say rushed (I sat on the patches for awhile), but there was a hole in my testing and the order that I ran my automated tests had made it seem that all was well. To avoid this in the future, I've already committed improvements to my test set, and will be adding additional timer latency checks soon. > If the bugs extend to more than this two-liner then for -stable it > might be better to just disable CLOCK_TAI (userspace can deal with it > just fine), and queue up the right fixes for the next merge window or > so. I don't foresee further issues (famous last words, eh), but since I was planning on keeping patch #4 and #5 for 3.14 anyway, we can wait till those land upstream to decide if the two-liner is sufficient or if disabling CLOCK_TAI in older -stable kernels is the right approach. That sound ok? thanks -john -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html