Linux Kernel Architecture

Chapter 2: Process Management and Scheduling

0 99 100 139

Higher Priority

Realtime Normal

− 20 Nice + 19

Figure 2-14: Kernel priority scale.

The following macros are used to convert between the different forms of representation (MAX_RT_PRIO

specifies the maximum priority of real-time processes, andMAX_PRIOis the maximal priority value for

regular processes):

<sched.h>

#define MAX_USER_RT_PRIO 100

#define MAX_RT_PRIO MAX_USER_RT_PRIO

#define MAX_PRIO (MAX_RT_PRIO + 40)

#define DEFAULT_PRIO (MAX_RT_PRIO + 20)

kernel/sched.c

#define NICE_TO_PRIO(nice) (MAX_RT_PRIO + (nice) + 20)

#define PRIO_TO_NICE(prio) ((prio) - MAX_RT_PRIO - 20)

#define TASK_NICE(p) PRIO_TO_NICE((p)->static_prio)

Computing Priorities

Recall that it is not sufficient to consider just the static priority of a process, but that three priorities

must be taken into account: dynamic priority (task_struct->prio), normal priority

(task_struct->normal_prio), and static priority (task_struct->static_prio). These priorities

are related to each other in interesting ways, and in the following I discuss how.

static_priois the starting point of the calculations. Assume that it has been already set and that the

kernel now wants to compute the other priorities. This is done by a one-liner:

p->prio = effective_prio(p);

The auxiliary functioneffective_prioperforms the following operations:

kernel/sched.c

static int effective_prio(struct task_struct *p)

{

p->normal_prio = normal_prio(p);

/*

* If we are RT tasks or we were boosted to RT priority,

* keep the priority unchanged. Otherwise, update priority

* to the normal priority:

*/

if (!rt_prio(p->prio))

return p->normal_prio;

return p->prio;

}

First of all, the normal priority is computed and stored innormal_priority. This side effect allows for

setting bothprioandnormal_priowith a single function invocation. Another helper function,rt_prio,