Description of problem: do_timer_interrupt does not acquire or release the xtime_lock (which is a sequential lock) when servicing the timer interrupt. The rest of the i386 timing code does use the lock and we've imported a lot of code into RHEL4 which uses the xtime_lock. Version-Release number of selected component (if applicable): 2.6.9-80.el4 I can almost guarantee that some of the timing instability is due to this missing lock. This issue, however, maybe moot. RHEL4 is almost at the end of its life, and I'm not sure we should change this behavior now. OTOH, as mentioned, using the lock in do_timer_interrupt() will certainly increase stability in gettimeofday calls, system time, etc. P.