Chapter 12 Virtual Memory Operating Systems by Dhananjay

Chapter 12 Virtual Memory Operating Systems, by Dhananjay Dhamdhere Copyright © 2008 12.

Introduction • • • Virtual Memory Basics Demand Paging The Virtual Memory Manager Page Replacement Policies Controlling Memory Allocation to a Process Shared Pages Memory-Mapped Files Case Studies of Virtual Memory Using Paging Virtual Memory Using Segmentation Operating Systems, by Dhananjay Dhamdhere 2

Virtual Memory Basics • MMU translates logical address into physical one • Virtual memory manager is a software component – Uses demand loading – Exploits locality of reference to improve performance Operating Systems, by Dhananjay Dhamdhere 3

Virtual Memory Basics (continued) Operating Systems, by Dhananjay Dhamdhere 4

Virtual Memory Using Paging • MMU performs address translation using page table Effective memory address of logical address (pi, bi) = start address of the page frame containing page pi + bi Operating Systems, by Dhananjay Dhamdhere 5

Demand Paging Preliminaries Operating Systems, by Dhananjay Dhamdhere 6

Demand Paging Preliminaries (continued) • Memory Management Unit (MMU) raises a page fault interrupt if page containing logical address not in memory Operating Systems, by Dhananjay Dhamdhere 7

Demand Paging Preliminaries (continued) A page fault interrupt is raised because Valid bit of page 3 is 0 Operating Systems, by Dhananjay Dhamdhere 8

Demand Paging Preliminaries (continued) • At a page fault, the required page is loaded in a free page frame • If no page frame is free, virtual memory manager performs a page replacement operation – Page replacement algorithm – Page-out initiated if page is dirty (modified bit is set) • Page-in and page-out: page I/O or page traffic • Effective memory access time in demand paging: Operating Systems, by Dhananjay Dhamdhere 9

Page Replacement • (Empirical) law of locality of reference: logical addresses used by process in a short interval tend to be grouped in certain portions of its logical address space Operating Systems, by Dhananjay Dhamdhere 10

Memory Allocation to a Process • How much memory to allocate to a process Operating Systems, by Dhananjay Dhamdhere 11

Optimal Page Size • Size of a page is defined by computer hardware • Page size determines: – No of bits required to represent byte number in a page – Memory wastage due to internal fragmentation – Size of the page table for a process – Page fault rates when a fixed amount of memory is allocated to a process • Use of larger page sizes than optimal value implies somewhat higher page fault rates for a process – Tradeoff between HW cost and efficient operation Operating Systems, by Dhananjay Dhamdhere 12

Paging Hardware • Page-table-address-register (PTAR) points to the start of a page table Operating Systems, by Dhananjay Dhamdhere 13

Paging Hardware (continued) Operating Systems, by Dhananjay Dhamdhere 14

Memory Protection • Memory protection violation raised if: – Process tries to access a nonexistent page – Process exceeds its (page) access privileges • It is implemented through: – Page table size register (PTSR) of MMU • Kernel records number of pages contained in a process in its PCB – Loads number from PCB in PTSR when process is scheduled – Prot info field of the page’s entry in the page table Operating Systems, by Dhananjay Dhamdhere 15

Address Translation and Page Fault Generation • Translation look-aside buffer (TLB): small and fast associative memory used to speed up address translation Operating Systems, by Dhananjay Dhamdhere 16

Address Translation and Page Fault Generation (continued) • TLBs can be HW or SW managed Operating Systems, by Dhananjay Dhamdhere 17

Address Translation and Page Fault Generation (continued) TLB hit ratio • Some mechanisms used to improve performance: – Wired TLB entries for kernel pages: never replaced – Superpages Operating Systems, by Dhananjay Dhamdhere 18

Superpages • TLB reach is stagnant even though memory sizes increase rapidly as technology advances – TLB reach = page size x no of entries in TLB – It affects performance of virtual memory • Superpages are used to increase the TLB reach – A superpage is a power of 2 multiple of page size – Its start address (both logical and physical) is aligned on a multiple of its own size – Max TLB reach = max superpage size x no of entries in TLB – Size of a superpage is adapted to execution behavior of a process through promotions and demotions Operating Systems, by Dhananjay Dhamdhere Copyright © 2008 12. 19 19

Support for Page Replacement • Virtual memory manager needs following information for minimizing page faults and number of page-in and page-out operations: – The time when a page was last used • Expensive to provide enough bits for this purpose • Solution: use a single reference bit – Whether a page is dirty • A page is clean if it is not dirty • Solution: modified bit in page table entry Operating Systems, by Dhananjay Dhamdhere 20

Practical Page Table Organizations • A process with a large address space requires a large page table, which occupies too much memory • Solutions: – Inverted page table • Describes contents of each page frame – Size governed by size of memory – Independent of number and sizes of processes – Contains pairs of the form (program id, page #) • Con: information about a page must be searched – Multilevel page table • Page table of process is paged Operating Systems, by Dhananjay Dhamdhere 21

Inverted Page Tables Use of hash table Speeds up search Operating Systems, by Dhananjay Dhamdhere 22

Multilevel Page Tables • If size of a table entry is 2 e bytes, number of page table entries in one PT page is 2 nb/2 e • Logical address (pi , bi) is regrouped into three fields: – PT page with the number pi 1 contains entry for pi – pi 2 is entry number for pi in PT page – bi Operating Systems, by Dhananjay Dhamdhere 23

I/O Operations in a Paged Environment • Process makes system call for I/O operations – Parameters include: number of bytes to transfer, logical address of the data area • Call activates I/O handler in kernel – I/O subsystem does not contain an MMU, so I/O handler replaces logical address of data area with physical address, using information from process page table – I/O fix (bit in misc info field) ensures pages of data are not paged out – Scatter/gather feature can deposit parts of I/O operation’s data in noncontiguous memory areas Operating Systems, by Dhananjay Dhamdhere – Alternatively, data area pages put in contiguous 24

Example: I/O Operations in Virtual Memory Operating Systems, by Dhananjay Dhamdhere 25

The Virtual Memory Manager Operating Systems, by Dhananjay Dhamdhere 26

Example: Page Replacement Operating Systems, by Dhananjay Dhamdhere 27

Overview of Operation of the Virtual Memory Manager • Virtual memory manager makes two important decisions during its operation: – Upon a page fault, decides which page to replace – Periodically decides how many page frames should be allocated to a process Operating Systems, by Dhananjay Dhamdhere 28

Page Replacement Policies • A page replacement policy should replace a page not likely to be referenced in the immediate future • Examples: – Optimal page replacement policy • Minimizes total number of page faults; infeasible in practice – First-in first-out (FIFO) page replacement policy – Least recently used (LRU) page replacement policy • Basis: locality of reference • Page reference strings – Trace of pages accessed by a process during its operation Operating Systems, by Dhananjaya. Dhamdhere – We associate reference time string with each 29

Example: Page Reference String • A computer supports instructions that are 4 bytes in length – Uses a page size of 1 KB – Symbols A and B are in pages 2 and 5 Operating Systems, by Dhananjay Dhamdhere 30

Operating Systems, by Dhananjay Dhamdhere 31

Page Replacement Policies (continued) • To achieve desirable page fault characteristics, faults shouldn’t increase when memory allocation is increased – Policy must have stack (or inclusion) property Operating Systems, by Dhananjay Dhamdhere 32

FIFO page replacement policy does not exhibit stack property. Operating Systems, by Dhananjay Dhamdhere 33

Page Replacement Policies (continued) • Virtual memory manager cannot use FIFO policy – Increasing allocation to a process may increase page fault frequency of process • Would make it impossible to control thrashing Operating Systems, by Dhananjay Dhamdhere 34

Practical Page Replacement Policies • Virtual memory manager has two threads – Free frames manager implements page replacement policy – Page I/O manager performs page-in/out operations Operating Systems, by Dhananjay Dhamdhere 35

Practical Page Replacement Policies (continued) • LRU replacement is not feasible – Computers do not provide sufficient bits in the ref info field to store the time of last reference • Most computers provide a single reference bit – Not recently used (NRU) policies use this bit • Simplest NRU policy: Replace an unreferenced page and reset all reference bits if all pages have been referenced • Clock algorithms provide better discrimination between pages by resetting reference bits periodically – One-handed clock algorithm – Two-handed clock algorithm Operating Systems, by Dhananjay Dhamdhere 36 » Resetting pointer (RP) and examining pointer (EP)

Example: Two-Handed Clock Algorithm • Both pointers are advanced simultaneously • Algorithm properties defined by pointer distance: – If pointers are close together, only recently used pages will survive in memory – If pointers are far apart, only pages that have not in a. Dhamdhere long time would be removed Operatingbeen Systems, used by Dhananjay 37

Controlling Memory Allocation to a Process • Process Pi is allocated alloci number of page frames • Fixed memory allocation – Fixes alloc statically; uses local page replacement • Variable memory allocation – Uses local and/or global page replacement – If local replacement is used, handler periodically determines correct alloc value for a process • May use working set model Operating Systems, by Dhananjay Dhamdhere 38

Implementation of a Working Set Memory Allocator • Swap out a process if alloc page frames cannot be allocated • Expensive to determine WSi(t, ∆) and alloci at every time instant t – Solution: Determine working sets periodically • Sets determined at end of an interval are used to decide values of alloc for use during the next interval Operating Systems, by Dhananjay Dhamdhere 39

Shared Pages • Static sharing results from static binding performed by a linker/loader before execution of program • Dynamic binding conserves memory by binding same copy of a program/data to several processes – Program or data shared retains its identity – Two conditions should be satisfied: • Shared program should be coded as reentrant – Can be invoked by many processes at the same time • Program should be bound to identical logical addresses in every process that shared it Operating Systems, by Dhananjay Dhamdhere 40

Shared pages should have same page numbers in all processes Operating Systems, by Dhananjay Dhamdhere 41

Copy-on-Write • Feature used to conserve memory when data in shared pages could be modified – Copy-on-write flag in page table entries Memory allocation decisions are performed statically A private copy of page k is made when A modifies it Operating Systems, by Dhananjay Dhamdhere 42

Memory-Mapped Files • Memory mapping of a file by a process binds file to a part of the logical address space of the process – Binding is performed when process makes a memory map system call – Analogous to dynamic binding of programs and data Operating Systems, by Dhananjay Dhamdhere 43

Memory-Mapped Files (continued) Operating Systems, by Dhananjay Dhamdhere 44

Case Studies of Virtual Memory Using Paging • • Unix Virtual Memory Linux Virtual Memory in Solaris Virtual Memory in Windows Operating Systems, by Dhananjay Dhamdhere 45

Unix Virtual Memory • • • Paging hardware differs in architectures Pages can be: resident, unaccessed, swapped-out Allocation of as little swap space as possible Copy-on-write fork Lack reference bit in some HW architectures; compensated using valid bit in interesting manner • Process can fix some pages in memory • Pageout daemon uses a clock algorithm – Swaps out a process if all required pages cannot be in memory – A swap-in priority is used to avoid starvation Operating Systems, by Dhananjay Dhamdhere 46

Linux Virtual Memory • Page size of 4 KB • On 64 -bit architectures, uses three-level page table • States for page frames: free, active, inactive dirty, inactive laundered, inactive clean • Page replacement based on a clock algorithm – Uses two lists called active list and inactive list • Buddy system allocator for allocating page frames • Several virtual memory regions for a process: – Zero-filled, file-backed, private memory Operating Systems, by Dhananjay Dhamdhere 47

Virtual Memory in Solaris • Supports normal pages and superpages – Superpages: • Automatically for processes with large address spaces • Can be requested using memcntl system call • Not used for memory-mapped files • Solaris 6 introduced priority paging to avoid interference between file processing and virtual memory • Page scanner tries to keep a sufficient number of page frames on cyclic page cache (since Solaris 8) – lotsfree parameter indicates how many page frames should be free Operating Systems, by Dhananjay Dhamdhere 48 – Uses two-handed clock algorithm on a global basis

Virtual Memory in Windows • • Supports both 32 -bit and 64 -bit logical addresses Page size is 4 KB Process address space is either 2 GB or 3 GB Two-, three- or four-level page tables and various page table entry formats – On the X-86 architecture: • Page frame can be in one of eight states, including: valid, free, zeroed, standby, modified, and bad • A process must first reserve virtual address space and then commit it for specific entities Operating Systems, by Dhananjay Dhamdhere 49

Virtual Memory in Windows (continued) • A section object represents a section of memory that can be shared – A process maps a view to access part of a section – Copy-on-write feature used for sharing pages – Prototype PTE is set-up for shared pages • TLBs are managed by HW (32 -bit) or SW (64 -bit) • Exploits reference locality: loads a few pages before and after a page-faulted page into memory • Uses notion of working sets (for memory allocation) – Clock algorithm • Page lists: free, zero-initiated, modified, standby Operating Systems, by Dhananjay Dhamdhere 50

Virtual Memory Using Segmentation • A segment is a logical entity in a program, such as a function, a data structure, or an object – Or, a module that consists of some or all of these – Convenient unit for sharing and protection Operating Systems, by Dhananjay Dhamdhere 51

Example: Effective Address Calculation in Segmentation • Logical address (si, bi) can be specified as ids – (alpha, beta) • alpha: name of a segment • beta: id associated with a byte contained in alpha Operating Systems, by Dhananjay Dhamdhere 52

Management of Memory • Similarities to paging: – Segment fault indicates segment is not in memory – Segment-in operation is performed to load segment • Segment-out operations may be needed first – Can use working set of segment for allocation • Segments can be replaced on NRU basis • Differences to paging: – Can lead to external fragmentation • Tackled through compaction or through memory reuse techniques (first fit, best fit, etc. ) – Segments can dynamically grow or shrink in size Operating Systems, by Dhananjay Dhamdhere 53

Sharing and Protection • Two important issues in protection and sharing of segments are: – Static and dynamic sharing of segments – Detecting use of invalid addresses • Protection exception if bi exceeds size of si Operating Systems, by Dhananjay Dhamdhere 54

Segmentation with Paging • External fragmentation exists in a virtual memory using segmentation – Solution: segmentation with paging Operating Systems, by Dhananjay Dhamdhere 55

Summary • Basic actions in virtual memory using paging: address translation and demand loading of pages – Implemented jointly by • Memory Management Unit (MMU): Hardware • Virtual memory manager: Software • Memory is divided into page frames • Virtual memory manager maintains a page table – Inverted and multilevel page tables use less memory but are less efficient – A fast TLB is used to speed up address translation Operating Systems, by Dhananjay Dhamdhere 56

Summary (continued) • Which page should VM manager remove from memory to make space for a new page? – Page replacement algorithms exploit locality of reference • LRU has stack property, but is expensive • NRU algorithms are used in practice – E. g. , clock algorithms • How much memory should manager allocate? – Use working set model to avoid thrashing • Copy-on-write can be used for shared pages • Memory mapping of files speeds up access to data Operating Systems, by Dhananjay Dhamdhere 57