Scheduler glossary

These notes give the background on the scheduling concepts that the contributor article on the Scheduler and the guide to Writing a scheduler build on, for a reader who has not written a scheduler before. Each page explains the concept on its own, then closes with how Cosmos applies it. The table groups them in three parts: how a thread gets the CPU and loses it, what code that runs against the tick has to respect, then the properties a policy is judged by. The rules for hook code also build on a note from the driver glossary, interrupt context.

Concept	Summary
Policy and mechanism	Deciding which thread runs, kept apart from the code that switches threads
Thread states and the run queue	Running, ready and blocked threads, the structure ready threads wait in, and the idle thread
Preemption	The kernel takes the CPU away on an interrupt; threads never have to volunteer
The tick and the quantum	The timer interrupt that gives the scheduler control, the time slice, and accounting by sample
Context switch	Saving one thread's registers and restoring another's, and when the switch actually happens
Interrupts at instruction boundaries	An interrupt lands after one instruction and before the next; nothing longer is atomic against it
Spinlocks and interrupt masking	Why a lock does not keep out an interrupt handler on the same CPU, and what does
Virtual-time fair-share	Weighted CPU shares, enforced by always running the thread whose virtual clock is furthest behind
Starvation and aging	Ready threads that never run, and the guards against it
Priority inversion and inheritance	A high-priority thread stuck behind a lock holder that a medium-priority thread preempts
Real-time scheduling	Deadlines, Rate Monotonic and EDF, utilization bounds and admission control
CPU affinity and load balancing	Per-CPU run queues, placement, migration and pinned threads