Forking & Process Scheduling Vivek Pai / Kai Li Princeton University

2 Mechanics  Exams not graded –Hopefully, this week  Filesystem project –Extension + bonus? –This wasn’t supposed to happen  Today: forking + scheduling

3 A Quick Review  What have we covered –How to store data via files –How to virtualize memory Every process gets uniform address space Lots of tricks can be played to share memory  What’s left –How to share/virtualize the processor –Having processes communicate/cooperate

4 How To Launch a New Process?  Obvious choice: “start process” system call  But not all processes start the same –“testprogram” versus “testprogram >& outfile” versus “testprogram arg1 arg2 >& outfile”  The “parent” process wants to specify various aspects of the child’s “environment” –Next step: add more parameters to specify environment

5 Can We Generalize?  What happens as more information gets added to the process’s “environment” – more parameters? New system calls? This gets ugly  What’s the most general way of setting up all of the environment?  So, why not allow process setup at any point? –This is the exec( ) system call (and its variants)

6 But We Want a Parent and a Child  The exec call “destroys” the current process  So, instead, destroy a copy of the process –The fork( ) call duplicates the current process –Better yet, don’t tightly couple fork and exec This way, you can customize the child’s environment  So what does fork( ) entail? –Making a copy of everything about the process –Ouch!

7 What Gets Copied  So far, we’ve covered the following: –VM system –File system –Signals  How do we go about copying this information?  What parts are easy to copy, and what’s hard?  What’s the common case with fork/exec? –What needs to get preserved in this scenario?

8 Shared Memory  How to destroy a virtual address space? –Link all PTEs –Reference count  How to swap out/in? –Link all PTEs –Operation on all entries  How to pin/unpin? –Link all PTEs –Reference count Process 1 Process 2 w w Page table Physical pages

Copy-On-Write  Child’s virtual address space uses the same page mapping as parent’s  Make all pages read-only  Make child process ready  On a read, nothing happens  On a write, generates an access fault –map to a new page frame –copy the page over –restart the instruction Parent process Child process r r r r Page table Physical pages

10 Issues of Copy-On-Write  How to destroy an address space –Same as shared memory case?  How to swap in/out? –Same as shared memory  How to pin/unpin –Same as shared memory

11 Process Creation & Termination Four primitives:  Fork – create a copy of this process  Exec – replace this process with this program  Wait – wait for child process to finish  Kill – (potentially) end a running process Processes form a tree – what happens when parent disappears?

12 Signals Asynchronous event delivery mechanism  Examples – FPE, segv, ctrl- c, hang up, resume  Default actions – ignore, abort, core dump  Handler – program- specified routine for signal

13 A Signaling Sidebar  What’s wrong with this program: int randVal; void SigHand(void) { printf(“your rand val is %d\n”, randVal); } int main(int argc, char *argv[]) { set up ctrl-c handler; while (1) { randVal = 0; randVal = 1; … randVal = 9; }

14 Scheduling Primitives  Block – wait on some event/resource –Network packet arrival –Keyboard, mouse input –Disk activity completion  Yield – give up running for now –Directed (let my friend run) –Undirected (let any process run)  Synchronization –We will talk about this later

15 Our Friend, The Transition Diagram Running Blocked Ready Scheduler dispatch Wait for resource Resource becomes available Create a process terminate

16 Process State Transition of Non-Preemptive Scheduling Running Blocked Ready Resource becomes available (move to ready queue) Create a process Terminate (call scheduler) Yield (call scheduler) Block for resource (call scheduler) Scheduler dispatch

17 Who’s Happy Right Now

Scheduler  A non-preemptive scheduler invoked by explicit block or yield calls  The simplest form Scheduler: save current process state (into PCB) choose next process to run dispatch (load PCB and run)  Does this work?

More on Scheduler  Should the scheduler use a special stack? –Yes, because a user process can overflow and it would require another stack to deal with stack overflow  Should the scheduler simply be a kernel process? –You can view it that way because it has a stack, code and its data structure –This process always runs when there is no user process

20 Where Should PCB Be Saved?  Save the PCB on its user stack –Many processors have a special instruction to do it “efficiently” –But, need to deal with the overflow problem –When the process terminates, the PCB vanishes  Save the PCB on the kernel heap data structure –May not be as efficient as saving it on stack –But, it is very flexible and no other problems

21 Physical Memory & Multiprogramming  Memory is a scarce resource  Want to run many programs  Programs need memory to run  What happens when M(a) + M(b) + M(c) > physical mem? Answer: paging. But what if no paging?

22 Job Swapping Partially executed swapped-out processes Ready Queue CPU I/O Waiting queues I/O Terminate Swap out Swap in

23 Add Job Swapping to State Transition Diagram Running Blocked Ready Resource becomes available (move to ready queue) Create a process Terminate (call scheduler) Yield (call scheduler) Block for resource (call scheduler) Scheduler dispatch Swap out Swap in Swap

24 Think About Swapping Is swapping  Necessary  Desirable  Good  Ideal Things to consider  Performance  Complexity  Efficiency Moreover, what decides swapping versus paging? Is each appropriate somewhere?