The scheduler uses the load tracking logic to compute the load of
threads to be enqueued into the run queue. The time delta between the
last enqueue and the next enqueue may grow very large for threads
that mostly wait on conditions. In such cases the int "n" period count
variable would become too small and wrap around, leading to an
assertion failure.
For this to happen, the thread in question would have to have slept for
at least ~25 days and then wake up. Threads often affected would be ones
waiting for some other process to end, for example shell threads waiting
for a long running process to exit.
Fixes#13558.
Previous implementation based on the actual load of each core and share
each thread has in that load turned up to be very problematic when
balancing load on very heavily loaded systems (i.e. more threads
consuming all available CPU time than there is logical CPUs).
The new approach is to estimate how much load would a thread produce
if it had all CPU time only for itself. Summing such load estimations
of each thread assigned to a given core we get a rank that contains
much more information than just simple actual core load.
