Transcript 下載/瀏覽Download

The Critical-Section(臨界區)Problem
 n processes all competing to use some
shared data.
 Each process has a code segment, called
critical section, in which the shared data is
accessed.
 Problem – ensure that when one process is
executing in its critical section, no other
process is allowed to execute in its critical
section.
EOS STUT
1
進入臨界區(Critical-Section)間的條件
1. 互斥(Mutual Exclusion)
當某個處理元在臨界區間內存取某些資源，則其他處
理元不能進入臨界區間去存取相同資源。
2. 行進(Progress)
當臨界區間內沒有處理元在執行，若有多個處理元要
求進入臨界區間執行，則只有那些不在執行剩餘程
式碼（Remainder Code）的處理元，才有資格被挑
選為下一個進入臨界區間的處理元，而且這個挑選
工作不能無限期的延遲下去。
3. 有限等待(Bounded Waiting)
當某個處理元要求進入臨界區間，一直到它獲得進入
臨界區間這段時間，允許其他處理元進入臨界區間
的次數有限制。
EOS STUT
2
Initial Attempts to Solve Problem –
2 processes
 Only 2 processes, P0 and P1
 General structure of process Pi (other
process Pj)
do {
entry section
//critical section
exit section
//remainder section
} while (1);
 Processes may share some common
variables to synchronize their actions.
EOS STUT
3
Ref. only Algorithm 1 - (2 processes)
 Shared variables:


Ref. only
int turn;
initially turn = 0
turn = i  Pi can enter its critical section
 Process Pi
do {
while (turn != i) ;
//critical section
turn = j;
//reminder section
} while (1);
 Wrong way! : Satisfies mutual exclusion, but not
progress
EOS STUT
4
Algorithm 2 - (2 processes)
 Shared variables


boolean flag[2];
initially flag [0] = flag [1] = false.
flag [i] = true Pi ready to enter its critical section
 Process Pi
do {
flag[i] = true;
while (flag[j]) ;
//critical section
flag [i] = false;
//remainder section
} while (1);
 Satisfies mutual exclusion, but not progress
requirement.
EOS STUT
5
Ref. only Algorithm 2 - (2 processes)
 Consider the following trace:






P0 sets flag[0] to true
A context-switch occurs
P1 sets flag[1] to true
P1 loops in while
A context-switch occurs
P0 loops in while
 Both P0 and P1 loop forever. This is the
livelock
EOS STUT
6
Algorithm 3 - Peterson's Algorithm
 Combined shared variables of algorithms 1 and 2.
 Process Pi
Ref. only
do {
flag [i] = true;
turn = j;
while (flag [j] && turn == j) ;
//critical section
flag [i] = false;
//remainder section
} while (1);
 Meets all three requirements; solves the criticalsection problem for two processes.
EOS STUT
7
Bakery Algorithm
Ref. only
 A Multiple-Process Solution.
 Pertain to a centralized environment.
 Notation <≡ lexicographical order (ticket #, process id
#)


(a,b) < (c,d) if a < c or if a = c and b < d
max (a0,…, an-1) is a number, k, such that k ≡ ai for i = 0, …, n –
1
 Shared data
boolean choosing[n];
int number[n];
Data structures are initialized to false and 0 respectively
EOS STUT
8
Bakery Algorithm
Ref. only
 保證達到互斥、行進、及有限等待條件。
 就如同到銀行或商店接受服務一樣，任何一個顧
客欲接受服務，均需依序取一張號碼牌，並依號
碼牌由小至大依序接受服務；但是萬一很多人均
有相同號碼牌時，則依客戶姓名排序依序接受服
務。
EOS STUT
9
Bakery Algorithm
do {
Ref. only
choosing[i] = true;
number[i] = max(number[0], number[1], …, number [n – 1])+1;
choosing[i] = false;
for (j = 0; j < n; j++) {
while (choosing[j]) ;
while ((number[j] != 0) && ( number[j], j ) < ( number[i], i ) ) ;
}
//critical section
number[i] = 0;
//remainder section
} while (1);
//errata
 If Pi is in its critical section and Pk (k != i) has already chosen its
number k != 0, then (number[i], i) < number[k], k)
EOS STUT
10
Bakery演算法
 互斥
Ref. only

任何要求進入臨界區間的處理元，均會使用while迴圈查核
(number[j] != 0) && ( number[j], j ) < ( number[i], i )。

若最小號碼牌（number）的處理元僅有一個，則輪由這個處
理元進入臨界區間，其他處理元只好繼續在迴圈內等待；

若有多個處理元擁有最小號碼牌（均同號碼），則因為每個
處理元編號不同，故僅會有一個處理元進入臨界區間，其他
處理元繼續在迴圈中等待。因此這些處理元進入臨界區間是
互斥的。
EOS STUT
11
Bakery演算法
行進
Ref. only
 由互斥的證明中，我們可以知道只有一個處
理元進入臨界區間，其他在迴圈等待之處理
元因為各擁有自己的號碼牌，且之後欲進入
臨界區間之處理元，其號碼牌均較大，所以
處理元進入臨界區間的次序是完全順序
（Totally Order），因此漸漸會有機會進
入臨界區間。
EOS STUT
12
Bakery演算法
有限等待
Ref. only
 當某個處理元要求進入臨界區間時，此處理
元取得號碼牌具有完全順序之後，它在迴圈
中等待時，若有別的處理元亦要進入臨界區
間，則別的處理元取到的號碼牌一定比此處
理元大，別的處理元不可能插隊在完全順序
之內，因此這是一種先到先服務（First
Come First Serve）的排程，也就是在此處
理元之前進入臨界區間的其他處理元，總共
進入臨界區間的次數是有限的。
EOS STUT
13
Synchronization Hardware
 Many systems provide hardware support for critical
section code
 Uniprocessors – could disable interrupts


Currently running code would execute without preemption
Generally too inefficient on multiprocessor systems

Operating systems using this not broadly scalable
 Modern machines provide special atomic hardware
instructions



Atomic = non-interruptable
Either test memory word and set value
Or swap contents of two memory words
EOS STUT
14
TestAndSet Instruction
 Definition:
Ref. only
boolean TestAndSet (boolean *target)
{
boolean rv = *target;
*target = TRUE;
return rv:
}
EOS STUT
15