21CSC202J-Operating Systems

UNIT - 4
Course • Identify and tell the various embedded
Learning operating systems and computer security
Rationale concepts

Course •Experiment the performance of different

Learning algorithms used in management of memory,
Outcomes file and I/O and select the appropriate one.

Learning Resources
• Abraham Silberschatz, Peter Baer Galvin, Greg Gagne, Operating systems, 10th ed.,
John Wiley & Sons, 2018

1
Memory Management
 Background
 Swapping
 Contiguous Memory Allocation
 Segmentation
 Paging
 Structure of the Page Table
 Example: The Intel 32 and 64-bit Architectures
 Example: ARM Architecture

2
Objectives

 To provide a detailed description of various ways of

organizing memory hardware
 To discuss various memory-management techniques,
including paging and segmentation
 To provide a detailed description of the Intel Pentium,
which supports both pure segmentation and segmentation
with paging

3
 Background

 Program must be brought (from disk) into memory and

placed within a process for it to be run
 Main memory and registers are only storage CPU can
access directly
 Memory unit only sees a stream of addresses + read
requests, or address + data and write requests
 Register access in one CPU clock (or less)
 Main memory can take many cycles, causing a stall
 Cache sits between main memory and CPU registers
 Protection of memory required to ensure correct
operation

4
 Base and Limit Registers
 A pair of base and limit registers define the logical address space
 CPU must check every memory access generated in user mode to
be sure it is between base and limit for that user

5
Hardware Address Protection

6
Address Binding
 Programs on disk, ready to be brought into memory to execute form an
input queue
 Without support, must be loaded into address 0000
 Inconvenient to have first user process physical address always at 0000
 How can it not be?
 Further, addresses represented in different ways at different stages of a
program’s life
 Source code addresses usually symbolic
 Compiled code addresses bind to relocatable addresses
 i.e. “14 bytes from beginning of this module”
 Linker or loader will bind relocatable addresses to absolute addresses
 i.e. 74014
 Each binding maps one address space to another

7
Binding of Instructions and Data to Memory

 Address binding of instructions and data to memory addresses

can happen at three different stages
 Compile time: If memory location known a priori, absolute code
can be generated; must recompile code if starting location changes
 Load time: Must generate relocatable code if memory location is
not known at compile time
 Execution time: Binding delayed until run time if the process can
be moved during its execution from one memory segment to
another
 Need hardware support for address maps (e.g., base and limit registers)

8
Multistep Processing of a User Program

9
 Logical vs. Physical Address Space

 The concept of a logical address space that is bound to a

separate physical address space is central to proper memory
management
 Logical address – generated by the CPU; also referred to as virtual
address
 Physical address – address seen by the memory unit
 Logical and physical addresses are the same in compile-time
and load-time address-binding schemes; logical (virtual) and
physical addresses differ in execution-time address-binding
scheme
 Logical address space is the set of all logical addresses
generated by a program
 Physical address space is the set of all physical addresses
generated by a program

10
Memory-Management Unit (MMU)
 Hardware device that at run time maps virtual to physical
address
 Many methods possible, covered in the rest of this chapter
 To start, consider simple scheme where the value in the
relocation register is added to every address generated by a
user process at the time it is sent to memory
 Base register now called relocation register
 MS-DOS on Intel 80x86 used 4 relocation registers
 The user program deals with logical addresses; it never sees
the real physical addresses
 Execution-time binding occurs when reference is made to location
in memory
 Logical address bound to physical addresses

11
Dynamic relocation using a relocation register

 Routine is not loaded until it is

called
 Better memory-space utilization;
unused routine is never loaded
 All routines kept on disk in
relocatable load format
 Useful when large amounts of
code are needed to handle
infrequently occurring cases
 No special support from the
operating system is required
 Implemented through program
design
 OS can help by providing libraries
to implement dynamic loading

12
Dynamic Linking
 Static linking – system libraries and program code combined by
the loader into the binary program image
 Dynamic linking –linking postponed until execution time
 Small piece of code, stub, used to locate the appropriate
memory-resident library routine
 Stub replaces itself with the address of the routine, and
executes the routine
 Operating system checks if routine is in processes’ memory
address
 If not in address space, add to address space
 Dynamic linking is particularly useful for libraries
 System also known as shared libraries
 Consider applicability to patching system libraries
 Versioning may be needed

13
Swapping
 A process can be swapped temporarily out of memory to a
backing store, and then brought back into memory for
continued execution
 Total physical memory space of processes can exceed physical
memory
 Backing store – fast disk large enough to accommodate copies
of all memory images for all users; must provide direct access
to these memory images
 Roll out, roll in – swapping variant used for priority-based
scheduling algorithms; lower-priority process is swapped out so
higher-priority process can be loaded and executed
 Major part of swap time is transfer time; total transfer time is
directly proportional to the amount of memory swapped
 System maintains a ready queue of ready-to-run processes
which have memory images on disk

14
Swapping (Cont.)
 Does the swapped out process need to swap back in to same
physical addresses?
 Depends on address binding method
 Plus consider pending I/O to / from process memory space
 Modified versions of swapping are found on many systems (i.e.,
UNIX, Linux, and Windows)
 Swapping normally disabled
 Started if more than threshold amount of memory allocated
 Disabled again once memory demand reduced below threshold

15
Schematic View of Swapping

16
 Context Switch Time including Swapping

 If next processes to be put on CPU is not in memory, need to

swap out a process and swap in target process
 Context switch time can then be very high
 100MB process swapping to hard disk with transfer rate of
50MB/sec
 Swap out time of 2000 ms
 Plus swap in of same sized process
 Total context switch swapping component time of 4000ms (4
seconds)
 Can reduce if reduce size of memory swapped – by knowing
how much memory really being used
 System calls to inform OS of memory use via
request_memory() and release_memory()

17
 Context Switch Time and Swapping (Cont.)

 Other constraints as well on swapping

 Pending I/O – can’t swap out as I/O would occur to wrong process
 Or always transfer I/O to kernel space, then to I/O device
 Known as double buffering, adds overhead

 Standard swapping not used in modern operating systems

 But modified version common
 Swap only when free memory extremely low

18
Swapping on Mobile Systems
 Not typically supported
 Flash memory based
 Small amount of space
 Limited number of write cycles
 Poor throughput between flash memory and CPU on mobile platform

 Instead use other methods to free memory if low

 iOS asks apps to voluntarily relinquish allocated memory
 Read-only data thrown out and reloaded from flash if needed
 Failure to free can result in termination

 Android terminates apps if low free memory, but first writes application
state to flash for fast restart
 Both OSes support paging as discussed below

19
Contiguous Allocation
 Main memory must support both OS and user processes
 Limited resource, must allocate efficiently
 Contiguous allocation is one early method
 Main memory usually into two partitions:
 Resident operating system, usually held in low memory with
interrupt vector
 User processes then held in high memory
 Each process contained in single contiguous section of memory

20
Contiguous Allocation (Cont.)
 Relocation registers used to protect user processes from each
other, and from changing operating-system code and data
 Base register contains value of smallest physical address
 Limit register contains range of logical addresses – each logical
address must be less than the limit register
 MMU maps logical address dynamically
 Can then allow actions such as kernel code being transient and
kernel changing size

21
Hardware Support for Relocation and Limit Registers

22
Multiple-partition allocation
 Multiple-partition allocation
 Degree of multiprogramming limited by number of partitions
 Variable-partition sizes for efficiency (sized to a given process’ needs)
 Hole – block of available memory; holes of various size are scattered
throughout memory
 When a process arrives, it is allocated memory from a hole large enough
to accommodate it
 Process exiting frees its partition, adjacent free partitions combined
 Operating system maintains information about:
a) allocated partitions b) free partitions (hole)

23
 Dynamic Storage-Allocation Problem
How to satisfy a request of size n from a list of free holes?

 First-fit: Allocate the first hole that is big enough

 Best-fit: Allocate the smallest hole that is big enough; must

search entire list, unless ordered by size
 Produces the smallest leftover hole

 Worst-fit: Allocate the largest hole; must also search entire

list
 Produces the largest leftover hole
First-fit and best-fit better than worst-fit in terms of speed and storage
utilization

24
Fragmentation
 External Fragmentation – total memory space exists to
satisfy a request, but it is not contiguous
 Internal Fragmentation – allocated memory may be slightly
larger than requested memory; this size difference is
memory internal to a partition, but not being used
 First fit analysis reveals that given N blocks allocated, 0.5 N
blocks lost to fragmentation
 1/3 may be unusable -> 50-percent rule

25
Fragmentation (Cont.)

 Reduce external fragmentation by compaction

 Shuffle memory contents to place all free memory together in
one large block
 Compaction is possible only if relocation is dynamic, and is done
at execution time
 I/O problem
 Latch job in memory while it is involved in I/O
 Do I/O only into OS buffers

 Now consider that backing store has same fragmentation

problems

26
Segmentation
 Memory-management scheme that supports user view of memory
 A program is a collection of segments
 A segment is a logical unit such as:
main program
procedure
function
method
object
local variables, global variables
common block
stack
symbol table
arrays

27
User’s View of a Program

28
Logical View of Segmentation

4
1

3 2
4

user space physical memory space

29
Segmentation Architecture
 Logical address consists of a two tuple:
<segment-number, offset>,

 Segment table – maps two-dimensional physical addresses; each

table entry has:
 base – contains the starting physical address where the segments
reside in memory
 limit – specifies the length of the segment

 Segment-table base register (STBR) points to the segment

table’s location in memory

 Segment-table length register (STLR) indicates number of

segments used by a program;
segment number s is legal if s < STLR

30
Segmentation Architecture (Cont.)
 Protection
 With each entry in segment table associate:
 validation bit = 0  illegal segment
 read/write/execute privileges

 Protection bits associated with segments; code sharing

occurs at segment level
 Since segments vary in length, memory allocation is a
dynamic storage-allocation problem
 A segmentation example is shown in the following diagram

31
Segmentation Hardware

32
Paging
 Physical address space of a process can be noncontiguous;
process is allocated physical memory whenever the latter is
available
 Avoids external fragmentation
 Avoids problem of varying sized memory chunks
 Divide physical memory into fixed-sized blocks called frames
 Size is power of 2, between 512 bytes and 16 Mbytes
 Divide logical memory into blocks of same size called pages
 Keep track of all free frames
 To run a program of size N pages, need to find N free frames
and load program
 Set up a page table to translate logical to physical addresses
 Backing store likewise split into pages
 Still have Internal fragmentation

33
Address Translation Scheme
 Address generated by CPU is divided into:
 Page number (p) – used as an index into a page table which contains
base address of each page in physical memory
 Page offset (d) – combined with base address to define the physical
memory address that is sent to the memory unit

page number page offset

p d
m -n n

 For given logical address space 2m and page size 2n

34
Paging Hardware

35
Paging Model of Logical and Physical Memory

36
Paging Example

n=2 and m=4 32-byte memory and 4-byte pages

37
Paging (Cont.)

 Calculating internal fragmentation

 Page size = 2,048 bytes
 Process size = 72,766 bytes
 35 pages + 1,086 bytes
 Internal fragmentation of 2,048 - 1,086 = 962 bytes
 Worst case fragmentation = 1 frame – 1 byte
 On average fragmentation = 1 / 2 frame size
 So small frame sizes desirable?
 But each page table entry takes memory to track
 Page sizes growing over time
 Solaris supports two page sizes – 8 KB and 4 MB

 Process view and physical memory now very different

 By implementation process can only access its own memory

38
Free Frames

Before allocation After allocation

39
Implementation of Page Table
 Page table is kept in main memory
 Page-table base register (PTBR) points to the page table
 Page-table length register (PTLR) indicates size of the page
table
 In this scheme every data/instruction access requires two
memory accesses
 One for the page table and one for the data / instruction
 The two memory access problem can be solved by the use of
a special fast-lookup hardware cache called associative
memory or translation look-aside buffers (TLBs)

40
 Implementation of Page Table (Cont.)
 Some TLBs store address-space identifiers (ASIDs) in each
TLB entry – uniquely identifies each process to provide
address-space protection for that process
 Otherwise need to flush at every context switch
 TLBs typically small (64 to 1,024 entries)
 On a TLB miss, value is loaded into the TLB for faster access
next time
 Replacement policies must be considered
 Some entries can be wired down for permanent fast access

41
Associative Memory

 Associative memory – parallel search

Page # Frame #

 Address translation (p, d)

 If p is in associative register, get frame # out
 Otherwise get frame # from page table in memory

42
Paging Hardware With TLB

43
Effective Access Time
 Associative Lookup =  time unit
 Can be < 10% of memory access time
 Hit ratio = 
 Hit ratio – percentage of times that a page number is found in the
associative registers; ratio related to number of associative registers
 Consider  = 80%,  = 20ns for TLB search, 100ns for memory access
 Effective Access Time (EAT)
EAT = (1 + )  + (2 + )(1 – )
=2+–
 Consider  = 80%,  = 20ns for TLB search, 100ns for memory access
 EAT = 0.80 x 100 + 0.20 x 200 = 120ns
 Consider more realistic hit ratio ->  = 99%,  = 20ns for TLB search,
100ns for memory access
 EAT = 0.99 x 100 + 0.01 x 200 = 101ns

44
Memory Protection
 Memory protection implemented by associating protection bit
with each frame to indicate if read-only or read-write access
is allowed
 Can also add more bits to indicate page execute-only, and so on
 Valid-invalid bit attached to each entry in the page table:
 “valid” indicates that the associated page is in the process’
logical address space, and is thus a legal page
 “invalid” indicates that the page is not in the process’ logical
address space
 Or use page-table length register (PTLR)
 Any violations result in a trap to the kernel

45
Valid (v) or Invalid (i) Bit In A Page Table

46
Shared Pages
 Shared code
 One copy of read-only (reentrant) code shared among processes
(i.e., text editors, compilers, window systems)
 Similar to multiple threads sharing the same process space
 Also useful for interprocess communication if sharing of read-
write pages is allowed
 Private code and data
 Each process keeps a separate copy of the code and data
 The pages for the private code and data can appear anywhere in
the logical address space

47
Shared Pages Example

48
Structure of the Page Table
 Memory structures for paging can get huge using straight-
forward methods
 Consider a 32-bit logical address space as on modern computers
 Page size of 4 KB (212)
 Page table would have 1 million entries (232 / 212)
 If each entry is 4 bytes -> 4 MB of physical address space / memory
for page table alone
 That amount of memory used to cost a lot
 Don’t want to allocate that contiguously in main memory

 Hierarchical Paging
 Hashed Page Tables
 Inverted Page Tables

49
Hierarchical Page Tables

 Break up the logical address space into multiple

page tables
 A simple technique is a two-level page table
 We then page the page table

50
Two-Level Page-Table Scheme

51
Two-Level Paging Example
 A logical address (on 32-bit machine with 1K page size) is divided into:
 a page number consisting of 22 bits
 a page offset consisting of 10 bits

 Since the page table is paged, the page number is further divided
into:
 a 12-bit page number
 a 10-bit page offset

 Thus, a logical address is as follows:

 where p1 is an index into the outer page table, and p2 is the

displacement within the page of the inner page table
 Known as forward-mapped page table

52
Address-Translation Scheme

53
64-bit Logical Address Space

 Even two-level paging scheme not sufficient

 If page size is 4 KB (212)
 Then page table has 252 entries
 If two level scheme, inner page tables could be 210 4-byte entries
 Address would look like

 Outer page table has 242 entries or 244 bytes

 One solution is to add a 2nd outer page table
 But in the following example the 2nd outer page table is still 234 bytes in size
 And possibly 4 memory access to get to one physical memory location

54
Three-level Paging Scheme

55
Hashed Page Tables
 Common in address spaces > 32 bits
 The virtual page number is hashed into a page table
 This page table contains a chain of elements hashing to the same
location
 Each element contains (1) the virtual page number (2) the value of
the mapped page frame (3) a pointer to the next element
 Virtual page numbers are compared in this chain searching for a
match
 If a match is found, the corresponding physical frame is extracted
 Variation for 64-bit addresses is clustered page tables
 Similar to hashed but each entry refers to several pages (such as 16)
rather than 1
 Especially useful for sparse address spaces (where memory references
are non-contiguous and scattered)

56
Hashed Page Table

57
Inverted Page Table

 Rather than each process having a page table and keeping

track of all possible logical pages, track all physical pages
 One entry for each real page of memory
 Entry consists of the virtual address of the page stored in that
real memory location, with information about the process that
owns that page
 Decreases memory needed to store each page table, but
increases time needed to search the table when a page
reference occurs
 Use hash table to limit the search to one — or at most a few —
page-table entries
 TLB can accelerate access
 But how to implement shared memory?
 One mapping of a virtual address to the shared physical address

58
Inverted Page Table Architecture

59
Virtual Memory
Chapter 9: Virtual Memory
 Background
 Demand Paging
 Copy-on-Write
 Page Replacement
 Allocation of Frames
 Thrashing
 Memory-Mapped Files
 Allocating Kernel Memory
 Other Considerations
 Operating-System Examples

61
Objectives

 To describe the benefits of a virtual memory system

 To explain the concepts of demand paging, page-replacement
algorithms, and allocation of page frames
 To discuss the principle of the working-set model
 To examine the relationship between shared memory and
memory-mapped files
 To explore how kernel memory is managed

62
Background
 Code needs to be in memory to execute, but entire program
rarely used
 Error code, unusual routines, large data structures
 Entire program code not needed at same time
 Consider ability to execute partially-loaded program
 Program no longer constrained by limits of physical memory
 Each program takes less memory while running -> more programs
run at the same time
 Increased CPU utilization and throughput with no increase in response
time or turnaround time

 Less I/O needed to load or swap programs into memory -> each user
program runs faster

63
Background (Cont.)
 Virtual memory – separation of user logical memory from
physical memory
 Only part of the program needs to be in memory for execution
 Logical address space can therefore be much larger than physical
address space
 Allows address spaces to be shared by several processes
 Allows for more efficient process creation
 More programs running concurrently
 Less I/O needed to load or swap processes

64
Background (Cont.)
 Virtual address space – logical view of how process is stored
in memory
 Usually start at address 0, contiguous addresses until end of
space
 Meanwhile, physical memory organized in page frames
 MMU must map logical to physical
 Virtual memory can be implemented via:
 Demand paging
 Demand segmentation

65
Virtual Memory That is Larger Than Physical Memory

66
 Virtual-address Space
 Usually design logical address space for
stack to start at Max logical address and
grow “down” while heap grows “up”
 Maximizes address space use
 Unused address space between
the two is hole
 No physical memory needed
until heap or stack grows to a
given new page
 Enables sparse address spaces with
holes left for growth, dynamically linked
libraries, etc
 System libraries shared via mapping into
virtual address space
 Shared memory by mapping pages read-
write into virtual address space
 Pages can be shared during fork(),
speeding process creation

67
Shared Library Using Virtual Memory

68
Demand Paging
 Could bring entire process into memory
at load time
 Or bring a page into memory only when
it is needed
 Less I/O needed, no unnecessary
I/O
 Less memory needed
 Faster response
 More users
 Similar to paging system with swapping
(diagram on right)
 Page is needed  reference to it
 invalid reference  abort
 not-in-memory  bring to memory
 Lazy swapper – never swaps a page
into memory unless page will be
needed
 Swapper that deals with pages is a
pager

69
Basic Concepts
 With swapping, pager guesses which pages will be used before
swapping out again
 Instead, pager brings in only those pages into memory
 How to determine that set of pages?
 Need new MMU functionality to implement demand paging
 If pages needed are already memory resident
 No difference from non demand-paging
 If page needed and not memory resident
 Need to detect and load the page into memory from storage
 Without changing program behavior
 Without programmer needing to change code

70
Valid-Invalid Bit
 With each page table entry a valid–invalid bit is associated
(v  in-memory – memory resident, i  not-in-memory)
 Initially valid–invalid bit is set to i on all entries
 Example of a page table snapshot:

 During MMU address translation, if valid–invalid bit in page table

entry is i  page fault
71
Page Table When Some Pages Are Not in Main Memory

72
Page Fault

 If there is a reference to a page, first reference to that page

will trap to operating system:
page fault
1. Operating system looks at another table to decide:
 Invalid reference  abort
 Just not in memory
2. Find free frame
3. Swap page into frame via scheduled disk operation
4. Reset tables to indicate page now in memory
Set validation bit = v
5. Restart the instruction that caused the page fault

73
Steps in Handling a Page Fault

74
Aspects of Demand Paging
 Extreme case – start process with no pages in memory
 OS sets instruction pointer to first instruction of process, non-memory-
resident -> page fault
 And for every other process pages on first access
 Pure demand paging
 Actually, a given instruction could access multiple pages -> multiple
page faults
 Consider fetch and decode of instruction which adds 2 numbers from
memory and stores result back to memory
 Pain decreased because of locality of reference
 Hardware support needed for demand paging
 Page table with valid / invalid bit
 Secondary memory (swap device with swap space)
 Instruction restart

75
Instruction Restart
 Consider an instruction that could access several different locations
 block move

 auto increment/decrement location

 Restart the whole operation?
 What if source and destination overlap?

76
Performance of Demand Paging
 Stages in Demand Paging (worse case)
1. Trap to the operating system
2. Save the user registers and process state
3. Determine that the interrupt was a page fault
4. Check that the page reference was legal and determine the location of the page on the disk
5. Issue a read from the disk to a free frame:
1. Wait in a queue for this device until the read request is serviced
2. Wait for the device seek and/or latency time
3. Begin the transfer of the page to a free frame
6. While waiting, allocate the CPU to some other user
7. Receive an interrupt from the disk I/O subsystem (I/O completed)
8. Save the registers and process state for the other user
9. Determine that the interrupt was from the disk
10. Correct the page table and other tables to show page is now in memory
11. Wait for the CPU to be allocated to this process again
12. Restore the user registers, process state, and new page table, and then resume the interrupted
instruction

77
 Performance of Demand Paging (Cont.)
 Three major activities
 Service the interrupt – careful coding means just several hundred instructions needed
 Read the page – lots of time
 Restart the process – again just a small amount of time
 Page Fault Rate 0  p  1
 if p = 0 no page faults
 if p = 1, every reference is a fault
 Effective Access Time (EAT)
EAT = (1 – p) x memory access
+ p (page fault overhead
+ swap page out
+ swap page in )

78
Demand Paging Example
 Memory access time = 200 nanoseconds
 Average page-fault service time = 8 milliseconds
 EAT = (1 – p) x 200 + p (8 milliseconds)
= (1 – p x 200 + p x 8,000,000
= 200 + p x 7,999,800
 If one access out of 1,000 causes a page fault, then
EAT = 8.2 microseconds.
This is a slowdown by a factor of 40!!
 If want performance degradation < 10 percent
 220 > 200 + 7,999,800 x p
20 > 7,999,800 x p
 p < .0000025
 < one page fault in every 400,000 memory accesses

79
Demand Paging Optimizations
 Swap space I/O faster than file system I/O even if on the same device
 Swap allocated in larger chunks, less management needed than file system
 Copy entire process image to swap space at process load time
 Then page in and out of swap space
 Used in older BSD Unix
 Demand page in from program binary on disk, but discard rather than paging out
when freeing frame
 Used in Solaris and current BSD
 Still need to write to swap space
 Pages not associated with a file (like stack and heap) – anonymous
memory
 Pages modified in memory but not yet written back to the file system
 Mobile systems
 Typically don’t support swapping
 Instead, demand page from file system and reclaim read-only pages (such as
code)

80
Copy-on-Write
 Copy-on-Write (COW) allows both parent and child processes to initially
share the same pages in memory
 If either process modifies a shared page, only then is the page copied
 COW allows more efficient process creation as only modified pages are
copied
 In general, free pages are allocated from a pool of zero-fill-on-demand
pages
 Pool should always have free frames for fast demand page execution
 Don’t want to have to free a frame as well as other processing
on page fault
 Why zero-out a page before allocating it?
 vfork() variation on fork() system call has parent suspend and child
using copy-on-write address space of parent
 Designed to have child call exec()
 Very efficient

81
Before Process 1 Modifies Page C

82
After Process 1 Modifies Page C

83
 What Happens if There is no Free Frame?

 Used up by process pages

 Also in demand from the kernel, I/O buffers, etc
 How much to allocate to each?
 Page replacement – find some page in memory, but not really in
use, page it out
 Algorithm – terminate? swap out? replace the page?
 Performance – want an algorithm which will result in minimum
number of page faults
 Same page may be brought into memory several times

84
Page Replacement

 Prevent over-allocation of memory by modifying page-

fault service routine to include page replacement
 Use modify (dirty) bit to reduce overhead of page
transfers – only modified pages are written to disk
 Page replacement completes separation between logical
memory and physical memory – large virtual memory can
be provided on a smaller physical memory

85
Need For Page Replacement

86
 Basic Page Replacement
1. Find the location of the desired page on disk

2. Find a free frame:

- If there is a free frame, use it
- If there is no free frame, use a page replacement algorithm to
select a victim frame
- Write victim frame to disk if dirty

3. Bring the desired page into the (newly) free frame; update the
page and frame tables

4. Continue the process by restarting the instruction that caused the

trap

Note now potentially 2 page transfers for page fault – increasing EAT

87
Page Replacement

88
 Page and Frame Replacement Algorithms

 Frame-allocation algorithm determines

 How many frames to give each process
 Which frames to replace
 Page-replacement algorithm
 Want lowest page-fault rate on both first access and re-access
 Evaluate algorithm by running it on a particular string of memory
references (reference string) and computing the number of page
faults on that string
 String is just page numbers, not full addresses
 Repeated access to the same page does not cause a page fault
 Results depend on number of frames available
 In all our examples, the reference string of referenced page
numbers is
7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1

89
Graph of Page Faults Versus The Number of Frames

90
First-In-First-Out (FIFO) Algorithm
 Reference string: 7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1
 3 frames (3 pages can be in memory at a time per process)

15 page faults

 Can vary by reference string: consider 1,2,3,4,1,2,5,1,2,3,4,5

 Adding more frames can cause more page faults!
 Belady’s Anomaly

 How to track ages of pages?

 Just use a FIFO queue

91
FIFO Illustrating Belady’s Anomaly

92
Optimal Algorithm
 Replace page that will not be used for longest period of time
 9 is optimal for the example
 How do you know this?
 Can’t read the future
 Used for measuring how well your algorithm performs

93
 Least Recently Used (LRU) Algorithm
 Use past knowledge rather than future
 Replace page that has not been used in the most amount of time
 Associate time of last use with each page

 12 faults – better than FIFO but worse than OPT

 Generally good algorithm and frequently used
 But how to implement?

94
LRU Algorithm (Cont.)
 Counter implementation
 Every page entry has a counter; every time page is referenced through
this entry, copy the clock into the counter
 When a page needs to be changed, look at the counters to find smallest
value
 Search through table needed

 Stack implementation
 Keep a stack of page numbers in a double link form:
 Page referenced:
 move it to the top
 requires 6 pointers to be changed

 But each update more expensive

 No search for replacement
 LRU and OPT are cases of stack algorithms that don’t have Belady’s
Anomaly

95
Use Of A Stack to Record Most Recent Page References

96
LRU Approximation Algorithms
 LRU needs special hardware and still slow
 Reference bit
 With each page associate a bit, initially = 0
 When page is referenced bit set to 1
 Replace any with reference bit = 0 (if one exists)
 We do not know the order, however

 Second-chance algorithm
 Generally FIFO, plus hardware-provided reference bit
 Clock replacement
 If page to be replaced has
 Reference bit = 0 -> replace it
 reference bit = 1 then:
 set reference bit 0, leave page in memory

 replace next page, subject to same rules

97
Second-Chance (clock) Page-Replacement Algorithm

98
 Enhanced Second-Chance Algorithm

 Improve algorithm by using reference bit and modify bit (if

available) in concert
 Take ordered pair (reference, modify)
1. (0, 0) neither recently used not modified – best page to replace
2. (0, 1) not recently used but modified – not quite as good, must
write out before replacement
3. (1, 0) recently used but clean – probably will be used again soon
4. (1, 1) recently used and modified – probably will be used again
soon and need to write out before replacement
 When page replacement called for, use the clock scheme but
use the four classes replace page in lowest non-empty class
 Might need to search circular queue several times

99
Counting Algorithms

 Keep a counter of the number of references that have been made

to each page
 Not common

 Lease Frequently Used (LFU) Algorithm: replaces page with

smallest count

 Most Frequently Used (MFU) Algorithm: based on the argument

that the page with the smallest count was probably just brought
in and has yet to be used

100
Page-Buffering Algorithms
 Keep a pool of free frames, always
 Then frame available when needed, not found at fault time
 Read page into free frame and select victim to evict and add to free
pool
 When convenient, evict victim
 Possibly, keep list of modified pages
 When backing store otherwise idle, write pages there and set to
non-dirty
 Possibly, keep free frame contents intact and note what is in
them
 If referenced again before reused, no need to load contents again
from disk
 Generally useful to reduce penalty if wrong victim frame selected

101
Applications and Page Replacement

 All of these algorithms have OS guessing about future page

access
 Some applications have better knowledge – i.e. databases
 Memory intensive applications can cause double buffering
 OS keeps copy of page in memory as I/O buffer
 Application keeps page in memory for its own work
 Operating system can given direct access to the disk, getting
out of the way of the applications
 Raw disk mode
 Bypasses buffering, locking, etc

102
Allocation of Frames
 Each process needs minimum number of frames
 Example: IBM 370 – 6 pages to handle SS MOVE instruction:
 instruction is 6 bytes, might span 2 pages
 2 pages to handle from
 2 pages to handle to
 Maximum of course is total frames in the system
 Two major allocation schemes
 fixed allocation
 priority allocation
 Many variations

103
Fixed Allocation
 Equal allocation – For example, if there are 100 frames (after
allocating frames for the OS) and 5 processes, give each process
20 frames
 Keep some as free frame buffer pool

 Proportional allocation – Allocate according to the size of process

 Dynamic as degree of multiprogramming, process sizes change
m = 64
si  size of process pi s1 = 10
S   si s2 = 127
m  total number of frames a1 =
10
´ 62 » 4
137
s
ai  allocation for pi  i  m 127
S a2 = ´ 62 » 57
137

104
Priority Allocation

 Use a proportional allocation scheme using priorities rather

than size

 If process Pi generates a page fault,

 select for replacement one of its frames
 select for replacement a frame from a process with lower
priority number

105
Global vs. Local Allocation
 Global replacement – process selects a replacement frame
from the set of all frames; one process can take a frame from
another
 But then process execution time can vary greatly
 But greater throughput so more common

 Local replacement – each process selects from only its own

set of allocated frames
 More consistent per-process performance
 But possibly underutilized memory

106
Non-Uniform Memory Access
 So far all memory accessed equally
 Many systems are NUMA – speed of access to memory varies
 Consider system boards containing CPUs and memory,
interconnected over a system bus
 Optimal performance comes from allocating memory “close to”
the CPU on which the thread is scheduled
 And modifying the scheduler to schedule the thread on the same
system board when possible
 Solved by Solaris by creating lgroups
 Structure to track CPU / Memory low latency groups
 Used my schedule and pager
 When possible schedule all threads of a process and allocate all memory
for that process within the lgroup

107
Thrashing
 If a process does not have “enough” pages, the page-fault rate is
very high
 Page fault to get page
 Replace existing frame
 But quickly need replaced frame back
 This leads to:
 Low CPU utilization
 Operating system thinking that it needs to increase the degree of
multiprogramming
 Another process added to the system

 Thrashing  a process is busy swapping pages in and out

108
Thrashing (Cont.)

109
 Demand Paging and Thrashing
 Why does demand paging work?
Locality model
 Process migrates from one locality to another
 Localities may overlap

 Why does thrashing occur?

 size of locality > total memory size
 Limit effects by using local or priority page replacement

110
Locality In A Memory-Reference Pattern

111
Working-Set Model
   working-set window  a fixed number of page references
Example: 10,000 instructions
 WSSi (working set of Process Pi) =
total number of pages referenced in the most recent  (varies in time)
 if  too small will not encompass entire locality
 if  too large will encompass several localities
 if  =   will encompass entire program
 D =  WSSi  total demand frames
 Approximation of locality
 if D > m  Thrashing
 Policy if D > m, then suspend or swap out one of the processes

112
Keeping Track of the Working Set
 Approximate with interval timer + a reference bit
 Example:  = 10,000
 Timer interrupts after every 5000 time units
 Keep in memory 2 bits for each page
 Whenever a timer interrupts copy and sets the values of all reference bits
to 0
 If one of the bits in memory = 1  page in working set
 Why is this not completely accurate?
 Improvement = 10 bits and interrupt every 1000 time units

113
Page-Fault Frequency
 More direct approach than WSS
 Establish “acceptable” page-fault frequency (PFF) rate
and use local replacement policy
 If actual rate too low, process loses frame
 If actual rate too high, process gains frame

114
 Working Sets and Page Fault Rates
n Direct relationship between working set of a process and its
page-fault rate
n Working set changes over time
n Peaks and valleys over time

115
Memory-Mapped Files
 Memory-mapped file I/O allows file I/O to be treated as routine
memory access by mapping a disk block to a page in memory
 A file is initially read using demand paging
 A page-sized portion of the file is read from the file system into a
physical page
 Subsequent reads/writes to/from the file are treated as ordinary
memory accesses
 Simplifies and speeds file access by driving file I/O through
memory rather than read() and write() system calls
 Also allows several processes to map the same file allowing the
pages in memory to be shared
 But when does written data make it to disk?
 Periodically and / or at file close() time
 For example, when the pager scans for dirty pages

116
 Memory-Mapped File Technique for all I/O

 Some OSes uses memory mapped files for standard I/O

 Process can explicitly request memory mapping a file via mmap()
system call
 Now file mapped into process address space
 For standard I/O (open(), read(), write(), close()), mmap
anyway
 But map file into kernel address space
 Process still does read() and write()
 Copies data to and from kernel space and user space
 Uses efficient memory management subsystem
 Avoids needing separate subsystem

 COW can be used for read/write non-shared pages

 Memory mapped files can be used for shared memory (although
again via separate system calls)

117
Memory Mapped Files

118
Shared Memory via Memory-Mapped I/O

119
Shared Memory in Windows API
 First create a file mapping for file to be mapped
 Then establish a view of the mapped file in process’s virtual address
space
 Consider producer / consumer
 Producer create shared-memory object using memory mapping
features
 Open file via CreateFile(), returning a HANDLE
 Create mapping via CreateFileMapping() creating a named
shared-memory object
 Create view via MapViewOfFile()

 Sample code in Textbook

120
Allocating Kernel Memory
 Treated differently from user memory
 Often allocated from a free-memory pool
 Kernel requests memory for structures of varying sizes
 Some kernel memory needs to be contiguous
 I.e. for device I/O

121
Buddy System
 Allocates memory from fixed-size segment consisting of physically-
contiguous pages
 Memory allocated using power-of-2 allocator
 Satisfies requests in units sized as power of 2
 Request rounded up to next highest power of 2
 When smaller allocation needed than is available, current chunk split into
two buddies of next-lower power of 2
 Continue until appropriate sized chunk available

 For example, assume 256KB chunk available, kernel requests 21KB

 Split into AL and AR of 128KB each
 One further divided into BL and BR of 64KB
 One further into CL and CR of 32KB each – one used to satisfy request

 Advantage – quickly coalesce unused chunks into larger chunk

 Disadvantage - fragmentation

122
Buddy System Allocator

123
 Slab Allocator
 Alternate strategy
 Slab is one or more physically contiguous pages
 Cache consists of one or more slabs
 Single cache for each unique kernel data structure
 Each cache filled with objects – instantiations of the data
structure
 When cache created, filled with objects marked as free
 When structures stored, objects marked as used
 If slab is full of used objects, next object allocated from empty
slab
 If no empty slabs, new slab allocated
 Benefits include no fragmentation, fast memory request
satisfaction

124
Slab Allocation

125
 Slab Allocator in Linux
 For example process descriptor is of type struct task_struct
 Approx 1.7KB of memory
 New task -> allocate new struct from cache
 Will use existing free struct task_struct

 Slab can be in three possible states

1. Full – all used
2. Empty – all free
3. Partial – mix of free and used
 Upon request, slab allocator
1. Uses free struct in partial slab
2. If none, takes one from empty slab
3. If no empty slab, create new empty

126
 Slab Allocator in Linux (Cont.)
 Slab started in Solaris, now wide-spread for both kernel mode
and user memory in various OSes
 Linux 2.2 had SLAB, now has both SLOB and SLUB allocators
 SLOB for systems with limited memory
 Simple List of Blocks – maintains 3 list objects for small, medium, large
objects

 SLUB is performance-optimized SLAB removes per-CPU queues,

metadata stored in page structure

127
Mass-Storage Systems
 Chapter 10: Mass-Storage Systems

 Overview of Mass Storage Structure

 Disk Structure
 Disk Attachment
 Disk Scheduling
 Disk Management
 Swap-Space Management
 RAID Structure
 Stable-Storage Implementation

129
Objectives

To describe the physical structure of secondary storage devices

and its effects on the uses of the devices
To explain the performance characteristics of mass-storage
devices
To evaluate disk scheduling algorithms
To discuss operating-system services provided for mass storage,
including RAID

130
 Overview of Mass Storage Structure
 Magnetic disks provide bulk of secondary storage of modern computers
 Drives rotate at 60 to 250 times per second
 Transfer rate is rate at which data flow between drive and computer
 Positioning time (random-access time) is time to move disk arm to desired
cylinder (seek time) and time for desired sector to rotate under the disk
head (rotational latency)
 Head crash results from disk head making contact with the disk surface --
That’s bad
 Disks can be removable
 Drive attached to computer via I/O bus
 Busses vary, including EIDE, ATA, SATA, USB, Fibre Channel, SCSI, SAS,
Firewire
 Host controller in computer uses bus to talk to disk controller built into
drive or storage array

131
Moving-head Disk Mechanism

132
Hard Disks
 Platters range from .85” to 14”
(historically)
 Commonly 3.5”, 2.5”, and 1.8”
 Range from 30GB to 3TB per drive
 Performance
 Transfer Rate – theoretical – 6 Gb/sec
 Effective Transfer Rate – real – 1Gb/sec
 Seek time from 3ms to 12ms – 9ms common
for desktop drives
 Average seek time measured or calculated
based on 1/3 of tracks
 Latency based on spindle speed
 1 / (RPM / 60) = 60 / RPM
(From Wikipedia)
 Average latency = ½ latency

133
Hard Disk Performance
 Access Latency = Average access time = average seek time +
average latency
 For fastest disk 3ms + 2ms = 5ms
 For slow disk 9ms + 5.56ms = 14.56ms
 Average I/O time = average access time + (amount to transfer /
transfer rate) + controller overhead
 For example to transfer a 4KB block on a 7200 RPM disk with a
5ms average seek time, 1Gb/sec transfer rate with a .1ms
controller overhead =
 5ms + 4.17ms + 0.1ms + transfer time =
 Transfer time = 4KB / 1Gb/s * 8Gb / GB * 1GB / 10242KB = 32 /
(10242) = 0.031 ms
 Average I/O time for 4KB block = 9.27ms + .031ms = 9.301ms

134
The First Commercial Disk Drive

1956
IBM RAMDAC computer
included the IBM Model
350 disk storage system

5M (7 bit) characters
50 x 24” platters
Access time = < 1 second

135
Solid-State Disks
 Nonvolatile memory used like a hard drive
 Many technology variations
 Can be more reliable than HDDs
 More expensive per MB
 Maybe have shorter life span
 Less capacity
 But much faster
 Busses can be too slow -> connect directly to PCI for example
 No moving parts, so no seek time or rotational latency

136
 Magnetic Tape
 Was early secondary-storage medium
 Evolved from open spools to cartridges
 Relatively permanent and holds large quantities of data
 Access time slow
 Random access ~1000 times slower than disk
 Mainly used for backup, storage of infrequently-used data,
transfer medium between systems
 Kept in spool and wound or rewound past read-write head
 Once data under head, transfer rates comparable to disk
 140MB/sec and greater
 200GB to 1.5TB typical storage
 Common technologies are LTO-{3,4,5} and T10000

137
Disk Structure
 Disk drives are addressed as large 1-dimensional arrays of logical
blocks, where the logical block is the smallest unit of transfer
 Low-level formatting creates logical blocks on physical media
 The 1-dimensional array of logical blocks is mapped into the
sectors of the disk sequentially
 Sector 0 is the first sector of the first track on the outermost
cylinder
 Mapping proceeds in order through that track, then the rest of the
tracks in that cylinder, and then through the rest of the cylinders
from outermost to innermost
 Logical to physical address should be easy
 Except for bad sectors
 Non-constant # of sectors per track via constant angular velocity

138
Disk Attachment
 Host-attached storage accessed through I/O ports talking to I/O
busses
 SCSI itself is a bus, up to 16 devices on one cable, SCSI initiator
requests operation and SCSI targets perform tasks
 Each target can have up to 8 logical units (disks attached to device
controller)
 Fibre Channel is high-speed serial architecture
 Can be switched fabric with 24-bit address space – the basis of storage
area networks (SANs) in which many hosts attach to many storage
units
 I/O directed to bus ID, device ID, logical unit (LUN)

139
Storage Array
 Can just attach disks, or arrays of disks
 Storage Array has controller(s), provides features to attached
host(s)
 Ports to connect hosts to array
 Memory, controlling software (sometimes NVRAM, etc)
 A few to thousands of disks
 RAID, hot spares, hot swap (discussed later)
 Shared storage -> more efficiency
 Features found in some file systems
 Snaphots, clones, thin provisioning, replication, deduplication, etc

140
Storage Area Network

 Common in large storage environments

 Multiple hosts attached to multiple storage arrays - flexible

141
Storage Area Network (Cont.)

 SAN is one or more storage arrays

 Connected to one or more Fibre Channel switches
 Hosts also attach to the switches
 Storage made available via LUN Masking from specific arrays
to specific servers
 Easy to add or remove storage, add new host and allocate it
storage
 Over low-latency Fibre Channel fabric
 Why have separate storage networks and communications
networks?
 Consider iSCSI, FCOE

142
Network-Attached Storage
 Network-attached storage (NAS) is storage made available
over a network rather than over a local connection (such as a
bus)
 Remotely attaching to file systems
 NFS and CIFS are common protocols
 Implemented via remote procedure calls (RPCs) between host
and storage over typically TCP or UDP on IP network
 iSCSI protocol uses IP network to carry the SCSI protocol
 Remotely attaching to devices (blocks)

143
Disk Scheduling
 The operating system is responsible for using hardware
efficiently — for the disk drives, this means having a fast
access time and disk bandwidth
 Minimize seek time
 Seek time  seek distance
 Disk bandwidth is the total number of bytes transferred,
divided by the total time between the first request for service
and the completion of the last transfer

144
Disk Scheduling (Cont.)
 There are many sources of disk I/O request
 OS
 System processes
 Users processes
 I/O request includes input or output mode, disk address, memory
address, number of sectors to transfer
 OS maintains queue of requests, per disk or device
 Idle disk can immediately work on I/O request, busy disk means
work must queue
 Optimization algorithms only make sense when a queue exists

145
Disk Scheduling (Cont.)
 Note that drive controllers have small buffers and can manage a
queue of I/O requests (of varying “depth”)
 Several algorithms exist to schedule the servicing of disk I/O
requests
 The analysis is true for one or many platters
 We illustrate scheduling algorithms with a request queue (0-199)

98, 183, 37, 122, 14, 124, 65, 67

Head pointer 53

146
FCFS
Illustration shows total head movement of 640 cylinders

147
SSTF
 Shortest Seek Time First selects the request with the minimum
seek time from the current head position
 SSTF scheduling is a form of SJF scheduling; may cause
starvation of some requests
 Illustration shows total head movement of 236 cylinders

148
SCAN

 The disk arm starts at one end of the disk, and moves toward
the other end, servicing requests until it gets to the other end
of the disk, where the head movement is reversed and servicing
continues.
 SCAN algorithm Sometimes called the elevator algorithm
 Illustration shows total head movement of 208 cylinders
 But note that if requests are uniformly dense, largest density at
other end of disk and those wait the longest

149
SCAN (Cont.)

150
C-SCAN
 Provides a more uniform wait time than SCAN
 The head moves from one end of the disk to the other,
servicing requests as it goes
 When it reaches the other end, however, it immediately returns to
the beginning of the disk, without servicing any requests on the
return trip
 Treats the cylinders as a circular list that wraps around from
the last cylinder to the first one
 Total number of cylinders?

151
C-SCAN (Cont.)

152
C-LOOK
 LOOK a version of SCAN, C-LOOK a version of C-SCAN
 Arm only goes as far as the last request in each direction,
then reverses direction immediately, without first going all
the way to the end of the disk
 Total number of cylinders?

153
C-LOOK (Cont.)

154
 Selecting a Disk-Scheduling Algorithm
 SSTF is common and has a natural appeal
 SCAN and C-SCAN perform better for systems that place a heavy load on the
disk
 Less starvation
 Performance depends on the number and types of requests
 Requests for disk service can be influenced by the file-allocation method
 And metadata layout
 The disk-scheduling algorithm should be written as a separate module of
the operating system, allowing it to be replaced with a different algorithm
if necessary
 Either SSTF or LOOK is a reasonable choice for the default algorithm
 What about rotational latency?
 Difficult for OS to calculate
 How does disk-based queueing effect OS queue ordering efforts?

155
Disk Management
 Low-level formatting, or physical formatting — Dividing a disk into
sectors that the disk controller can read and write
 Each sector can hold header information, plus data, plus error
correction code (ECC)
 Usually 512 bytes of data but can be selectable
 To use a disk to hold files, the operating system still needs to record its
own data structures on the disk
 Partition the disk into one or more groups of cylinders, each treated
as a logical disk
 Logical formatting or “making a file system”
 To increase efficiency most file systems group blocks into clusters
 Disk I/O done in blocks
 File I/O done in clusters

156
Disk Management (Cont.)

 Raw disk access for apps that want to do their own block
management, keep OS out of the way (databases for example)
 Boot block initializes system
 The bootstrap is stored in ROM
 Bootstrap loader program stored in boot blocks of boot partition
 Methods such as sector sparing used to handle bad blocks

157
Booting from a Disk in Windows

158
 Swap-Space Management

 Swap-space — Virtual memory uses disk space as an extension of main memory

 Less common now due to memory capacity increases
 Swap-space can be carved out of the normal file system, or, more commonly, it can be
in a separate disk partition (raw)
 Swap-space management
 4.3BSD allocates swap space when process starts; holds text segment (the
program) and data segment
 Kernel uses swap maps to track swap-space use
 Solaris 2 allocates swap space only when a dirty page is forced out of physical
memory, not when the virtual memory page is first created
 File data written to swap space until write to file system requested
 Other dirty pages go to swap space due to no other home
 Text segment pages thrown out and reread from the file system as needed
 What if a system runs out of swap space?
 Some systems allow multiple swap spaces

159
Data Structures for Swapping on Linux Systems

160
RAID Structure
 RAID – redundant array of inexpensive disks
 multiple disk drives provides reliability via redundancy
 Increases the mean time to failure
 Mean time to repair – exposure time when another failure could
cause data loss
 Mean time to data loss based on above factors
 If mirrored disks fail independently, consider disk with 1300,000
mean time to failure and 10 hour mean time to repair
 Mean time to data loss is 100, 0002 / (2 ∗ 10) = 500 ∗ 106 hours, or
57,000 years!
 Frequently combined with NVRAM to improve write performance
 Several improvements in disk-use techniques involve the use of
multiple disks working cooperatively

161
RAID (Cont.)
 Disk striping uses a group of disks as one storage unit
 RAID is arranged into six different levels
 RAID schemes improve performance and improve the reliability
of the storage system by storing redundant data
 Mirroring or shadowing (RAID 1) keeps duplicate of each disk
 Striped mirrors (RAID 1+0) or mirrored stripes (RAID 0+1) provides
high performance and high reliability
 Block interleaved parity (RAID 4, 5, 6) uses much less redundancy
 RAID within a storage array can still fail if the array fails, so
automatic replication of the data between arrays is common
 Frequently, a small number of hot-spare disks are left
unallocated, automatically replacing a failed disk and having
data rebuilt onto them

162
RAID Levels

163
RAID (0 + 1) and (1 + 0)

164
Other Features
 Regardless of where RAID implemented, other useful features
can be added
 Snapshot is a view of file system before a set of changes take
place (i.e. at a point in time)
 More in Ch 12
 Replication is automatic duplication of writes between separate
sites
 For redundancy and disaster recovery
 Can be synchronous or asynchronous
 Hot spare disk is unused, automatically used by RAID production
if a disk fails to replace the failed disk and rebuild the RAID set
if possible
 Decreases mean time to repair

165
File-System Interface
 Chapter 11: File-System Interface

 File Concept
 Access Methods
 Disk and Directory Structure
 File-System Mounting
 File Sharing
 Protection

167
Objectives

 To explain the function of file systems

 To describe the interfaces to file systems
 To discuss file-system design tradeoffs, including access
methods, file sharing, file locking, and directory structures
 To explore file-system protection

168
File Concept
 Contiguous logical address space
 Types:
 Data
 numeric
 character
 binary

 Program
 Contents defined by file’s creator
 Many types
 Consider text file, source file, executable file

169
File Attributes
 Name – only information kept in human-readable form
 Identifier – unique tag (number) identifies file within file system
 Type – needed for systems that support different types
 Location – pointer to file location on device
 Size – current file size
 Protection – controls who can do reading, writing, executing
 Time, date, and user identification – data for protection, security,
and usage monitoring
 Information about files are kept in the directory structure, which is
maintained on the disk
 Many variations, including extended file attributes such as file
checksum
 Information kept in the directory structure

170
File info Window on Mac OS X

171
File Operations
 File is an abstract data type
 Create
 Write – at write pointer location
 Read – at read pointer location
 Reposition within file - seek
 Delete
 Truncate
 Open(Fi) – search the directory structure on disk for entry
Fi, and move the content of entry to memory
 Close (Fi) – move the content of entry Fi in memory to
directory structure on disk

172
Open Files
 Several pieces of data are needed to manage open files:
 Open-file table: tracks open files
 File pointer: pointer to last read/write location, per process
that has the file open
 File-open count: counter of number of times a file is open – to
allow removal of data from open-file table when last processes
closes it
 Disk location of the file: cache of data access information
 Access rights: per-process access mode information

173
Open File Locking
 Provided by some operating systems and file systems
 Similar to reader-writer locks
 Shared lock similar to reader lock – several processes can acquire
concurrently
 Exclusive lock similar to writer lock
 Mediates access to a file
 Mandatory or advisory:
 Mandatory – access is denied depending on locks held and
requested
 Advisory – processes can find status of locks and decide what to do

174
File Locking Example – Java API
import java.io.*;
import java.nio.channels.*;
public class LockingExample {
public static final boolean EXCLUSIVE = false;
public static final boolean SHARED = true;
public static void main(String arsg[]) throws IOException {
FileLock sharedLock = null;
FileLock exclusiveLock = null;
try {
RandomAccessFile raf = new RandomAccessFile("file.txt", "rw");
// get the channel for the file
FileChannel ch = raf.getChannel();
// this locks the first half of the file - exclusive
exclusiveLock = ch.lock(0, raf.length()/2, EXCLUSIVE);
/** Now modify the data . . . */
// release the lock
exclusiveLock.release();

175
 File Locking Example – Java API (Cont.)

// this locks the second half of the file - shared

sharedLock = ch.lock(raf.length()/2+1, raf.length(),
SHARED);
/** Now read the data . . . */
// release the lock
sharedLock.release();
} catch (java.io.IOException ioe) {
System.err.println(ioe);
}finally {
if (exclusiveLock != null)
exclusiveLock.release();
if (sharedLock != null)
sharedLock.release();
}
}
}

176
File Types – Name, Extension

177
File Structure
 None - sequence of words, bytes
 Simple record structure
 Lines
 Fixed length
 Variable length
 Complex Structures
 Formatted document
 Relocatable load file
 Can simulate last two with first method by inserting
appropriate control characters
 Who decides:
 Operating system
 Program

178
Sequential-access File

179
Access Methods
 Sequential Access
read next
write next
reset
no read after last write
(rewrite)
 Direct Access – file is fixed length logical records
read n
write n
position to n
read next
write next
rewrite n
n = relative block number

 Relative block numbers allow OS to decide where file should be placed

 See allocation problem in Ch 12

180
Simulation of Sequential Access on Direct-access File

181
Other Access Methods

 Can be built on top of base methods

 General involve creation of an index for the file
 Keep index in memory for fast determination of location of
data to be operated on (consider UPC code plus record of
data about that item)
 If too large, index (in memory) of the index (on disk)
 IBM indexed sequential-access method (ISAM)
 Small master index, points to disk blocks of secondary index
 File kept sorted on a defined key
 All done by the OS
 VMS operating system provides index and relative files as
another example (see next slide)

182
Example of Index and Relative Files

183
Directory Structure

 A collection of nodes containing information about all files

Directory

Files
F1 F2 F4
F3
Fn

Both the directory structure and the files reside on disk

184
Disk Structure
 Disk can be subdivided into partitions
 Disks or partitions can be RAID protected against failure
 Disk or partition can be used raw – without a file system, or
formatted with a file system
 Partitions also known as minidisks, slices
 Entity containing file system known as a volume
 Each volume containing file system also tracks that file
system’s info in device directory or volume table of contents
 As well as general-purpose file systems there are many
special-purpose file systems, frequently all within the same
operating system or computer

185
A Typical File-system Organization

186
Types of File Systems
 We mostly talk of general-purpose file systems
 But systems frequently have may file systems, some general- and
some special- purpose
 Consider Solaris has
 tmpfs – memory-based volatile FS for fast, temporary I/O
 objfs – interface into kernel memory to get kernel symbols for
debugging
 ctfs – contract file system for managing daemons
 lofs – loopback file system allows one FS to be accessed in place of
another
 procfs – kernel interface to process structures
 ufs, zfs – general purpose file systems

187
Operations Performed on Directory
 Search for a file

 Create a file

 Delete a file

 List a directory

 Rename a file

 Traverse the file system

188
 Directory Organization

The directory is organized logically to obtain

 Efficiency – locating a file quickly

 Naming – convenient to users
 Two users can have same name for different files
 The same file can have several different names
 Grouping – logical grouping of files by properties, (e.g.,
all Java programs, all games, …)

189
Single-Level Directory
 A single directory for all users

 Naming problem
 Grouping problem

190
Two-Level Directory
 Separate directory for each user

 Path name
 Can have the same file name for different user
 Efficient searching
 No grouping capability

191
Tree-Structured Directories

192
Tree-Structured Directories (Cont.)
 Efficient searching

 Grouping Capability

 Current directory (working directory)

 cd /spell/mail/prog
 type list

193
 Tree-Structured Directories (Cont)
 Absolute or relative path name
 Creating a new file is done in current directory
 Delete a file
rm <file-name>
 Creating a new subdirectory is done in current directory
mkdir <dir-name>
Example: if in current directory /mail
mkdir count

Deleting “mail”  deleting the entire subtree rooted by “mail”

194
Acyclic-Graph Directories
 Have shared subdirectories and files

195
Acyclic-Graph Directories (Cont.)
 Two different names (aliasing)
 If dict deletes list  dangling pointer
Solutions:
 Backpointers, so we can delete all pointers
Variable size records a problem
 Backpointers using a daisy chain organization
 Entry-hold-count solution
 New directory entry type
 Link – another name (pointer) to an existing file
 Resolve the link – follow pointer to locate the file

196
General Graph Directory

197
General Graph Directory (Cont.)
 How do we guarantee no cycles?
 Allow only links to file not subdirectories
 Garbage collection
 Every time a new link is added use a cycle detection algorithm to
determine whether it is OK

198
File System Mounting
 A file system must be mounted before it can be accessed
 A unmounted file system (i.e., Fig. 11-11(b)) is mounted at a
mount point

199
Mount Point

200
File Sharing
 Sharing of files on multi-user systems is desirable
 Sharing may be done through a protection scheme
 On distributed systems, files may be shared across a network
 Network File System (NFS) is a common distributed file-sharing
method
 If multi-user system
 User IDs identify users, allowing permissions and protections to be
per-user
Group IDs allow users to be in groups, permitting group access
rights
 Owner of a file / directory
 Group of a file / directory

201
File Sharing – Remote File Systems
 Uses networking to allow file system access between systems
 Manually via programs like FTP
 Automatically, seamlessly using distributed file systems
 Semi automatically via the world wide web
 Client-server model allows clients to mount remote file systems from
servers
 Server can serve multiple clients
 Client and user-on-client identification is insecure or complicated
 NFS is standard UNIX client-server file sharing protocol
 CIFS is standard Windows protocol
 Standard operating system file calls are translated into remote calls
 Distributed Information Systems (distributed naming services) such as
LDAP, DNS, NIS, Active Directory implement unified access to
information needed for remote computing

202
File Sharing – Failure Modes

 All file systems have failure modes

 For example corruption of directory structures or other non-user
data, called metadata
 Remote file systems add new failure modes, due to network
failure, server failure
 Recovery from failure can involve state information about
status of each remote request
 Stateless protocols such as NFS v3 include all information in
each request, allowing easy recovery but less security

203
File Sharing – Consistency Semantics
 Specify how multiple users are to access a shared file
simultaneously
 Similar to Ch 5 process synchronization algorithms
 Tend to be less complex due to disk I/O and network latency (for remote
file systems

 Andrew File System (AFS) implemented complex remote file sharing

semantics
 Unix file system (UFS) implements:
 Writes to an open file visible immediately to other users of the same open
file
 Sharing file pointer to allow multiple users to read and write concurrently

 AFS has session semantics

 Writes only visible to sessions starting after the file is closed

204
Protection
 File owner/creator should be able to control:
 what can be done
 by whom
 Types of access
 Read
 Write
 Execute
 Append
 Delete
 List

205
Access Lists and Groups
 Mode of access: read, write, execute
 Three classes of users on Unix / Linux
RWX
a) owner access 7  111
RWX
b) group access 6  110
RWX
c) public access 1  001
 Ask manager to create a group (unique name), say G, and add
some users to the group.
 For a particular file (say game) or subdirectory, define an
appropriate access.

Attach a group to a file

chgrp G game

206
Windows 7 Access-Control List Management

207
A Sample UNIX Directory Listing

208