Chapter 9: Memory Questions? What is main memory? CSCI [4|6]730 - PDF document

Chapter 9: Memory Questions? � � What is main memory? CSCI [4|6]730 � � How does multiple processes share memory Operating Systems space? � � What is static and dynamic allocation? � � What is segmentation ? Main Memory 2 Maria Hybinette, UGA Maria Hybinette, UGA Review: Motivation for Multiprogramming Multiprogramming Goals Uniprocessing: One process runs at a time Low Address 2 n -1 � � Sharing (0x00000000) Physical » � Several processes coexist in main memory Memory OS Stack » � Cooperating processes can share portions of address space � � Transparency » � Processes are not aware that memory is shared » � Works regardless of number and/or location of processes User Address � � Protection Space Heap Process » � Cannot corrupt OS or other processes » � Privacy: Cannot read data of other processes Code 0 � � Efficiency � � Disadvantages: High Address » � Do not waste CPU or memory resources (0x7fffffff) » � Only one process runs at a time » � Keep fragmentation low (later) » � Process can destroy OS 3 4 Maria Hybinette, UGA Maria Hybinette, UGA Static Relocation Static Reallocation System Memory System Memory � � Goal: Allow transparent sharing - � � Disadvantages: Each address space may be » � No protection Initial P 0 Initial P 0 placed anywhere in memory – � Process can destroy OS or other processes » � OS finds free space for new process Process P 3 Process P 3 – � No privacy » � Modify addresses statically (similar » � Address space must be allocated to linker) when loading process contiguously Process P 2 Process P 2 � � Advantages: – � Allocate space for worst-case stack and heap » � Allows multiple processes to run – � Processes may not grow » � Requires no hardware support » � Cannot move process after they are placed Process P 1 Process P 1 » � Fragmentation (later) 5 6 Maria Hybinette, UGA Maria Hybinette, UGA

Hardware Support for Dynamic Relocation Dynamic Relocation � � Goal: Protect processes from one another � � Two operating modes � � Requires hardware support » � Privileged (protected, kernel) mode: OS runs » � Memory Management Unit (MMU) – � When enter OS (trap, system calls, interrupts, exceptions) � � MMU dynamically changes process address at every – � Allows certain instructions to be executed memory reference (compute address on-the-fly) � � Can manipulate contents of MMU – � Allows OS to access all of physical memory » � Process generates logical or virtual addresses » � User mode: User processes run » � Memory hardware uses physical or real addresses – � Perform translation of logical address to physical address Process runs here � � MMU contains base and bounds registers OS can control MMU » � base: start location for address space CPU MMU Memory » � bounds: size limit of address space 7 8 Logical address Physical address Maria Hybinette, UGA Maria Hybinette, UGA Implementation of Dynamic Relocation Example of Dynamic Relocation � � Translation on every memory access of user process � � What are the physical addresses for the following 16- » � MMU compares logical address to bounds register bit logical addresses (Hex: highest F:1111)? – � if logical address is greater, then generate error � � Process 1: base: 0x4320 , bounds: 0x2220 � » � MMU adds base register to logical address to form physical address » � 0x0000: � » � 0x1110: � » � 0x3000: � logical mode � no � physical � � Process 2: base: 0x8540 , bounds: 0x3330 � = � address user? � address » � 0x0000: � Registers: � » � 0x1110: � mode � 1 bit � yes � » � 0x3000: � � � Operating System bounds � no � + � 32 < � error bounds? � » � 0x0000: � bits � base � base � 9 10 Maria Hybinette, UGA Maria Hybinette, UGA Managing Processes with Base and Bounds Base and Bounds Discussion 0 � � � Advantages Operating � � Context-switch » � Provides protection (both read and System write) across address spaces » � Add base and bounds registers to PCB » � Supports dynamic relocation » � Steps 30004 � 30004 � – � Can move address spaces 1. Change to privileged mode Process base � – � Why might you want to do this? 2. Save base and bounds registers of old process 42094 � 12090 � » � Simple, inexpensive: Few registers, 3. Load base and bounds registers of new process bound � little logic in MMU Process 4. Change to user mode and jump to new process » � Fast: Add and compare can be done in � � What if don’t change base and bounds registers when parallel switch? � � Disadvantages � � Protection requirement » � Each process must be allocated » � User process cannot change base and bounds registers contiguously in physical memory Process » � User process cannot change to privileged mode – � Must allocate memory that may not be used by process » � No partial sharing: Cannot share limited parts of address space 11 12 Maria Hybinette, UGA Maria Hybinette, UGA

Segmentation Segmented Addressing heap � � � Divide address space into � � How does process designate a particular logical segments stacj � segment? » � Each segment symbol � corresponds to logical subroutine � » � Use part of logical address table � entity in address space – � Top bits of logical address select segment – � code, stack, heap – � Low bits of logical address select offset within � � Each segment can symbol � subroutine � segment table � main � independently: table � program � » � be placed separately in physical memory stack � » � grow and shrink main � program � » � be protected (separate heap � read/write/execute protection bits) Logical Address Physical Address 13 14 Maria Hybinette, UGA Space Space Maria Hybinette, UGA Segmentation Implementation Discussion of Segmentation � � MMU contains Segment Table (per process) � � Advantages » � Each segment has own base and bounds, protection bits » � Example: 14 bit logical address, 4 segments » � Enables sparse allocation of address space – � Stack and heap can grow independently Segment Base Bounds R W – � Heap: If no data on free list, dynamic memory allocator requests more from OS (e.g., UNIX: malloc calls sbrk()) 0 0x2000 0x06ff 1 0 – � Stack: OS recognizes reference outside legal segment, 1 0x0000 0x04ff 1 0 extends stack implicitly » � Different protection for different segments 2 0x3000 0x0fff 1 1 – � Read-only status for code 3 0x0000 0xffff 0 0 » � Enables sharing of selected segments » � Supports dynamic relocation of each segment � � Translate logical addresses � physical addresses: � � Disadvantages » � 0x0240: 0 th segment 240 internal address within segment � what address? � » � Each segment must be allocated contiguously » � 0x1108: � – � May not have sufficient physical memory for large segments » � 0x265c: � » � 0x3002: � 15 16 Maria Hybinette, UGA Maria Hybinette, UGA When to Bind Physical & Logical Addresses Motivation for Dynamic Memory Source � Program � � � Compile time: If memory location � � Why do processes need dynamic allocation of memory? known a priori, absolute code can be » � Do not know amount of memory needed at compile time Compiler or generated; must recompile code if Assembler » � Must be pessimistic when allocate memory statically starting location changes – � Allocate enough for worst possible case � � Load time: Must generate Object � – � Storage is used inefficiently Other � Module � Object � relocatable code if memory location � � Recursive procedures Modules � is not known at compile time » � Do not know how many times procedure will be nested Linkage � � Execution time: Binding delayed editor � � Complex data structures: lists and trees until run time if the process can be System � » � struct my_t Library � Load � moved during its execution from module � *p = (struct my_t *)malloc(sizeof(struct one memory segment to another. my_t)); Need hardware support for address � � Two types of dynamic allocation loader System � maps (e.g., base and limit registers) Library � » � Stack In-memory » � Heap Binary 17 18 Memory image Maria Hybinette, UGA Maria Hybinette, UGA