Memory management in vx works

Memory Management in

VxWorksDhan V Sagar

CB.EN.P2CSE13007

Overview

Hard Real Time System from Wind River Systems

Used in automobiles, consumer devices, network

switches

Based on a monolithic kernel

VxWorks5 Vs VxWorks6

Address Translation

Real Addressing Mode

Using a Relocation Register

Full Virtual Memory Functionality

Memory Management (VxW 5)

• No Swapping or Paging

• Assumes that there is enough physical memory available

• Does not have a directly supported virtual memory system

• Virtual memory support separately as add-on

• Kernel is loaded into the bottom part of the memory

• System Memory Pool

• Dynamic memory routines for manipulating the heap

• malloc(), free() are used

MMU

MMU page table designs

Memory Allocation

First Fit Allocation (VxWorks 5)

• Allocates the first available space

• Free blocks can be stored in a simple linked list

• Leads to fragmentation

• Dynamic partition sizes are allocated on system initialisation

• Partition the memory into two or more sections

Memory Allocation

Best Fit Allocation (VxWorks 6)

Smallest-sized block that is big enough to satisfy the request

More complex data structure to store the free blocks

Whenever a new size is created, a new node is inserted in the tree

Probability of a larger block having to be split is smaller

Free blocks of the same size can be grouped together

Memory Allocation

Best Fit Allocation (VxWorks 6)

Memory AllocationPerformance Comparison

Memory Management Enhancements (VxWorks 6)

Automatic resource reclamation

Improved memory allocation using a “best-fit” algorithm

User-mode heap and memory partition support

Developer-replaceable user-mode heap allocator

Heap instrumentation for heaps and memory partitions in both the kernel and RTPs

Tight integration with error management

Memory Protection

o MMU-based memory protection provides isolation of the kernel from user-mode applications and of applications from each other, increasing device reliability.

o VxWorks’ preemptive, priority-based global task scheduler

ensures real-time deterministic behavior.

o The ability to create private or public objects in the kernel and in RTPs offers

flexibility to use objects

o The extensible system call interface enables application developers to employ custom-developed kernel services from user-mode execution.

o Support for shared libraries among RTPs improves code efficiency and reusability