---
# System prepended metadata
title: Event-Based concurrency
tags: [OS]
---
# Event-Based concurrency
## **Intro**
### **How I/O work**
* **System call**
* `需請求Kernal`
* **Kernal maintain a buffer for each I/O device**
* 
### **Five I/O mode (based on wait)**
* **blocking**
* `-`
* **non-blocking**
* `if no data in buffer, return
EWOULDBLOCK immediately`
* **I/O multiplexing**
* `select() and poll()`
* `blocked in both wait and copy`
* **Signal driven I/O**
* `SIGIO`
* `nonblocked in wait but blocked in copy`
* **ascynchrnous I/O**
* `aio()`
* `nonblocked all the way.`
* `Signal as I/O completed`
## **I/O multiplexing**
* I/O multiplexing is to be notified, by kernel, if one or more I/O conditions are ready.
* Use select() and poll()
### **Scenario**
* **Single thread client** vs.`descriptors`, `sockets`
* **Single thread server** vs.
* `both TCP and UDP`
* `services and protocals`
* `TCP server: listening socket & its connected socket`
### **Select**
* **Code**
* 
* 
* **Ready condition 1** : 請求 kernal 測試 input device
* `connected socket ready for reading `
* ==socket receive buffer== 達 low-water mark，可以呼叫read()
* `listening socket is ready for reading`
* 已有新的client連線建立，在complete connection queue，可以呼叫 accept() 取回此新 socket 的 file descriptor
* **Ready condition 2** : 請求 kernal 測試 output device
* `socket ready for writing`
* ==socket send buffer== 的 available space > low-water mark
* **summary**
* 
### **Poll**
* **system call**
* 
* **設定**
* `bitwise OR AND`
* client[0].events = POLLRDNORM | POLLRDBAND;
* client[0].revents & POLLRDNORM
* 
* **3 Classes of Data Identified by poll**
* 
## **Programming Example**
### **Echo Client**
* **問題** ：同時使用 fgets, readline 兩個 blocking I/O
* `stdin 等 user input，readline 等 socket input`
* **解法** : 用 select 同時等 (fgets)stdin, (readline)socket
* 
* why :`read/write/exceptset 是雙向變數，每次return
，值已被修改，故每次都要reset`
* 
### **Concurrent TCP server by select**
* **keep track of client[] and rset**
* `client[]` : client descriptor array
* `rset` : readdescriptorset
* 
* **Code**
* `Initialization`
* 
* `Loop` : 把 connfd 放進 client[] 和 &allset (FD_SET descriptor)
* 
* 從 client[]裡找到 sockfd 並 readline(sockfd) 和 Write()
* 
### **Concurrent TCP server by poll**
* **keep track of client[] and pollfd**
* `client[]` : client of ==pollfd== structure
* struct pollfd = client[OPEN_MAX]
* **Code**
* `Initialization`
* 
* `Loop` : 把 connfd 放進 client[] 和 &allset (FD_SET descriptor)
* 
* 從 client[]裡找到 sockfd 並 readline(sockfd) 和 Write()
* 
## **Ascynchrnous I/O**
* Issue an I/O request 後 returns immediately
* Uses another call to determine whether the request completes
### **使用**
* **To issue an asynchronous read to a file**
* `issue `==io_read==`(struct aiocb *aiocbp);`
* **struct aiocb**
* 
* **To check whether the request has completed**
* `call `==io_error==`(const struct aiocb *aiocbp);`
* Returns 0 if completed
* Returns EINPROGRESS if not yet
* Use `signals` (非periodly check) to inform applications when an asynchronous I/O completes
* **summary**
* 
###### tags: `OS`