Does a spinlock prevent preemptive time slicing?

[arludwig]

I was wondering, what measures I have to take to really ensure atomicity on an operation. On a single core without multithreading I would have to irq_lock() my critical section. With the possibility of multi core I (and for good practice anyways) I would use k_spin_lock() instead. If I have preemtive threads with timeslicing, k_sched_lock() can ensure my thread is not preempted while in the critical section. But how far do I have to go for SMP + preemtive? Do I need to combine k_spin_lock() and k_sched_lock(), or does k_spin_lock() implicitly prevent timeslicing?

My guess would be, that timeslicing is internally based on interrupts, and thus with interrupts disabled through k_spin_lock() no timeslicing would take place. Also, from a contract point of view, it would be a weakness of k_spin_lock() to not protect against timeslicing.

But I could not find any documentation adressing this question, neither in the threading / multithreading sections nor in the scheduling sections. So for my peace of mind, I would like to ask you to confirm my intuition - or teach me how it is done :)