Assignment2 - CSci 4061: Intro to Operating Systems...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: CSci 4061: Intro to Operating Systems Assignment 2: Multi-process Web Browser Due: March 10th by 10pm. You may work in a group of 2 or 3. 1. Purpose: Most of the traditional web browsers (like Firefox) run as a multi‐threaded process, rendering web‐ pages on user request. However, a multi‐process architecture for web‐browsers is emerging, e.g. Internet Explorer 8.0 and Google Chrome. Each browser process manages one or more tabs. This design architecture provides two primary benefits: 1. Robustness: Web‐page isolation via separate processes, makes the web‐browser immune to crashes by providing address‐space isolation to the rendered web‐pages. Each web‐page hence is rendered by a separate process and has its own address space. A plug‐in crash (one of main reasons for browser‐crashes) in one web‐page hence will have (theoretically) no effect on other web‐pages being managed by other processes. 2. Parallelism: Web‐pages are becoming ever more complex with contents ranging from javascript code to full‐blown web‐applications like google docs. The rendering engine too has grown in complexity in order to provide this functionality. With the functionality of a web‐browser slowly transforming into a platform where web‐applications can run independently and in parallel and hardware is multicore/multimode, a multi‐process architecture for the browser is attractive. In this project, we will explore this new architecture for a web‐browser while learning the various mechanisms for inter‐process communication. Next follows a brief description of your multi‐process web‐browser. Note: we will provide all browser graphics and event handling (e.g. clicks and input retrieval). 2.0 Description: The multi‐process web‐browser uses a separate process for each of its tab to guard against the glitches in the web‐page rendering engine. You will design a multi‐process web‐browser in this assignment. Your designed web‐browser (called the multi‐proc‐web‐browser) will consist of one MAIN‐PARENT process (called the ROUTER process), one CONTROLLER child process and zero or more URL‐RENDERING child process(es) corresponding to each open tab. 2.1.ROUTER process: The ROUTER process is the main parent process which runs when the web‐browser program is started. The ROUTER process is responsible for forking the CONTROLLER process. Once done, it waits for request from its child processes (CONTROLLER/URL‐RENDERING processes) in a loop (polling the pipe descriptors via non‐blocking read). Note that the ROUTER process is not involved in any of the browser rendering graphics and is only intended to function as a router between the CONTROLLER process and URL‐ RENDERING processes. 2.2 CONTROLLER process: The CONTROLLER process is the child of the ROUTER process. It renders the window (shown in Fig.1) to the user. Following is the description of the various regions that appear in the window: 1. URL region: A text field where the destination url (eg., is entered. 2. Tab‐selector region: The tab where the url is to be rendered. It accepts an integer value > 0 and less than the number of opened tabs. 3. Create‐new‐tab button: The control for creating new tab. 4. Close‐tab button: The control for closing the tab. Whenever a user wants to open a url in a new tab, he first clicks on the ‘create‐new‐tab’ button. This opens a new URL‐RENDERING window (shown in Fig.2). The user then enters the url in the ‘URL region’ of the CONTROLLER followed by the tab‐number where the requested url is to be rendered. Once done, the user hits the enter key in the ‘URL region’ and the page is rendered in the specified tab. Note that the user must hit the enter key in the ‘URL region’ for the web‐page to be rendered in the specified tab. Hitting the enter key anywhere else will not work. 2.3 URL­RENDERING process: The URL‐RENDERING tab process(es) are also children of the ROUTER process. They render the requested url (specified in the CONTROLLER windows). Fig 2. shows the layout of the URL‐RENDERING tab. Note that the URL‐RENDERING tab does not has any control where the url can be entered. All urls rendered in the URL‐RENDERING tab are specified in the CONTROLLER with appropriate tab index. Create‐new‐ URL‐region tab button 1 Tab‐selector region Tab‐close button Fig 1. Controller Tab <Contents of web page> Fig 2. URL‐RENDERING tab 3 Forms of IPC: You will implement the following forms of IPC while implementing this multi‐process browser. 1. Pipes: Pipe is a uni‐directional channel of communication between parent and child process. Earlier described messaging infrastructure between the CONTROLLER, ROUTER and URL‐ RENDERING processes will be implemented using pipes. The IPC in section 3.1 is to be implemented using ‘pipes’. 2. EXTRA‐CREDIT: Shared memory is an efficient way for sharing data among different processes. You will use the shared‐memory to implement the caching of web pages. Description of the caching feature will be posted later. 3.1 Inter­Process Communication Infrastructure: Fig. 3 demonstrates the IPC among the browser processes. The messages follow the following format: There are three kinds of messages exchanged among the various browser processes: CREATE_NEW_TAB, NEW_URL_ENTERED Router Process TAB_KILLED CONTROLLER Process URL‐RENDERING Process 1 CREATE_NEW_TAB, NEW_URL_ENTERED URL‐RENDERING Process 2 URL‐RENDERING Process N Fig 3. IPC among various browser process 1. CREATE_NEW_TAB: This request is sent by the CONTROLLER process to the ROUTER process when the user clicks the create‐new‐tab button (see fig. 1). See Appendix A for request format and Section 4 on how each of the involved processes handles this message. 2. NEW_URL_ENTERED: The CONTROLLER process passes this message to the ROUTER process when the user hits the enter button after entering the URL in the URL region (see fig.1). See Appendix A for the request format and Section 4 on how each of the involved processes handles this message. 3. TAB_KILLED: Individual URL‐RENDERING/CONTROLLER processes pass this message to the ROUTER process when the user closes the tab by clicking the ‘cross’ at the top‐right corner of the tab‐window (see fig. 1). See Appendix A for request format and Section 4 on how each of the involved processes handles this message. 4 Program Flow: This section describes the flow of each of the involved processes. 4.1 ROUTER Process: 1. When you invoke the multi‐proc‐web‐browser, the main() function of your program starts executing. At this stage the only process that exists is the main‐parent process which is the ROUTER process. Upon its invocation, the ROUTER process performs the following additional tasks: 2. fork the CONTROLLER process: a. First, the ROUTER process creates two pipes for bi‐directional communication with the CONTROLLER process. b. Then fork the new child process, the CONTROLLER process. See Section 4.2 for a description of the CONTROLLER. 3. Wait for requests from child processes: a. The ROUTER process then polls (via non‐blocking read) on the set of open pipe file descriptors created before forking the child process (see step 2 above), in a loop. At this stage, there is at least one child process (the CONTROLLER process), which the ROUTER polls for messages. The termination condition for the loop is when there exist no child process for the ROUTER process i.e., all child processes have finished their execution. b. The non‐blocking read will read data from the pipe. If no data is available, the read system call will return immediately with return value of ‐1 and errno set to EAGAIN. In that case, simply continue with the polling (see step 3(a)). c. In case the read returns with some data, read the data into a buffer of type child_req_to_parent (see Appendix A for its definition). There can be three kinds of messages that the ROUTER process may read from the pipe. i. CREATE_TAB: Upon receiving this message, the ROUTER process forks a URL‐ RENDERING process. Again, the ROUTER process first creates 2 pipes for bidirectional communication with the child process and then forks to create the URL‐RENDERING child process. See Section 4.3 for the description of the URL‐ RENDERING process’ functionality. ii. NEW_URI_ENTERED: This message specifies the URL to be rendered along with tab‐ index on which the URL should be rendered. Upon receiving this message, the ROUTER process simply writes this message on the pipe connecting the ROUTER process with the corresponding URL‐RENDERING process whose index is specified in the message. iii. TAB_KILLED: This message contains the tab‐index of the recently closed tab. Upon receiving this message, the ROUTER process simply closes the file descriptors of the corresponding pipe which was created for communication with the closed tab. Note: This change should be now be reflected in the list of pipe descriptors which the ROUTER process polls for messages. 4. When all browsing windows (including the CONTROLLER window) are closed (which implies there are no child processes for the ROUTER process), the ROUTER process exits with a return status of 0 (for success), marking the successful completion of the program. 4.2 CONTROLLER Process: 1. The CONTROLLER process is created by the ROUTER as described in Section 4.1. 2. Upon its creation, it invokes the show_browser() function of the wrapper‐library that we provide to you (see Appendex B). This is a blocking call which returns when the user closes the CONTROLLER window. The CONTROLLER process exits with the return status of 0 (for success) when the user closes the CONTROLLER window. How does the CONTROLLER respond to requests if it is blocked? It uses callbacks described now. 3. Two callback functions are registered when creating the browser‐window (CONTROLLER window) e.g.., new_tab_created_cb and uri_entered_cb. These get executed in the context of CONTROLLER process. Following is what each of these callbacks are supposed to do: a. uri_entered_cb: Basic template for this function is available in Appendix C. You should append your code in the end of the template for this callback. Basically, this callback function receives two input parameters: (1) the URL that the user has entered in url‐region (see Fig.1) and (2) auxiliary data (a void* which should be type‐casted to a pointer of type browser_window*). The template code extracts the pipe descriptor from the browser_window data‐structure. Your supplied code should prepare and send a NEW_URI_ENTERED request message to the ROUTER process). The child_to_parent_fd field of browser_window structure stores the pipe descriptor through which CONTROLLER communicates with the ROUTER - this descriptor gets set when you invoke the create_browser function of the wrapper library. See Appendix B for details of the create_browser function). Note that the template code will only work if you invoked the create_browser wrapper‐library function as per recommendation. b. new_tab_created_cb: Basic template for this function is also available in Appendix C. You should append your code in the end of provided template for this callback. The template code extracts the URL along with the tab‐index where user wishes to render the URL. The CONTROLLER process simply sends a CREATE_TAB message to the ROUTER process. You should supply the code for creating this message and writing it to the pipe descriptor connecting the CONTROLLER to the ROUTER. Code for extracting the pipe descriptor is provided in the template. 4.3 URL­RENDERING Process: 1. URL‐RENDERING process is the child of the ROUTER process and is created when the user clicks the ‘create‐new‐tab’ button in the CONTROLLER window. 2. Upon its creation, it performs the following two tasks in a loop. The termination condition for the loop is the TAB_KILLED message from the ROUTER process. Task 1. Wait for messages from the ROUTER: the CONTROLLER URL‐RENDERING process reads (non‐blocking) the pipe descriptor (that the ROUTER process created for communication with the child processes as explained in bullet 2(a) of Section 4.1) for incoming messages from the ROUTER process. There are three kinds of messages that it might receive: i. CREATE_TAB: CONTROLLER URL‐RENDERING process ignores this message. ii. NEW_URI_ENTERED: Upon receiving this message, the URL‐RENDERING process renders the requested url by invoking render_web_page_in_tab()of the wrapper‐library (see Appendix B). iii. TAB_KILLED: Invoke process_all_gtk_events()function of the wrapper‐ library (see Appendix B) and exit the process with the return status of 0 (for success). iv. If it reads a message of any other type, interpret it as a bogus message and do some error handling (for example, you may ignore the read message, or you may print an error message on the screen etc.). You have the choice of deciding upon the recovery strategy. Task 2: Process GTK events: Invoke process_single_gtk_event()wrapper‐library function. 3. As mentioned earlier, the URL‐RENDERING process exits on receiving the TAB_KILLED message from the ROUTER process. 5.0 Error Handling You are expected to check the return value of all system calls that you use in your program to check for error conditions. If your program encounters an error (for example, invalid tab‐index supplied in tab‐ selector region), useful error message should be printed to the screen. Your program should be robust; it should try to recover from errors if possible (one such instance is given in Point 2 (Task 1, sub‐point iv) of Section 4.3). If the error prevents your program from functioning normally, then it should exit after printing the error message. (The use of the perror() function for printing error messages is encouraged.) Make sure, while your program executes, there should not be any zombie or/and orphaned processes. Upon closing all the tabs (of the web‐browser), main process must exit properly, cleaning‐up any used resources. We will soon update the Grading Criteria for the assignment. 6.0 Documentation You must include a README file which describes your program. It needs to contain the following: • • • • • • The purpose of your program How to compile the program How to use the program from the shell (Syntax) What exactly your program does Any explicit assumptions you have made Your strategies for error handling The README file does not have to be very long, as long as it properly describes the above points. Proper in this case means that a first-time user will be able to answer the above questions without any confusion. Within your code you should use one or two sentences to describe each function that you write. You do not need to comment every line of your code. However, you might want to comment portions of your code to increase readability. At the top of your README file and main C source file please include the following comment: /* CSci4061 S2010 Assignment 2 * section: one_digit_number * login: itlabs_login_name (login used to submit) * date: mm/dd/yy * name: full_name1, full_name2 (for partner) * id: id_for_first_name, id_for_second_name */ 7.0 Deliverables: Files containing your code • A README file (readme and c code should indicate this is assignment #2). • A makefile that will compile your code and produce a program called multi-pro-web-browser. Note: this makefile will be used by us to compile your program with the make utility. All files should be submitted using the SUBMIT utility. You can find a link to it on the class website. This is your official submission that we will grade. Please note that future submissions under the same homework title OVERWRITE previous submissions; we can only grade the most recent submission. Therefore, once you have submitted your final code before the deadline, DO NOT accidently resubmit your code again after the deadline because we can only grade the most recent submission. ...
View Full Document

This note was uploaded on 10/21/2011 for the course CSCI 4061 taught by Professor Tripathi during the Spring '08 term at Minnesota.

Ask a homework question - tutors are online