Programming Project 6:

Multiprogramming with Fork

Due Date: ______________________________

Project Duration: One week

Overview and Goal

In this project, you will implement the following syscalls: Fork, Join, Exit, and Yield. After completing the previous project, the OS could support a single user-level process. In this project, you will add the necessary functionality to run many user-level processes at once.

Download New Files

The files for this project are available in:

http://www.cs.pdx.edu/~harry/Blitz/OSProject/p6/

The following files are new to this project:

TestProgram3.h

TestProgram3.c

The following files have been modified from the last project:

makefile

DISK

The makefile has been modified to compile TestProgram3. The DISK file has been enlarged, since the previous version was too small to accommodate TestProgram3.

All remaining files are unchanged from the last project.

Task 1:

Implement the Fork syscall.

Task 2:

Implement the following syscalls:

Join

Exit

Yield

Changes to Kernel.h and Kernel.c:

Please change Kernel.h

from:

NUMBER_OF_PHYSICAL_PAGE_FRAMES = 100

to:

NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Please change your InitFirstProcess routine in Kernel.c

from:

th.Fork (StartUserProcess, "TestProgram1" asInteger)

to:

th.Fork (StartUserProcess, "TestProgram3" asInteger)

The Fork Syscall

When a user-level process wishes to create another process, it invokes the Fork syscall. The kernel will then create a new process and assign it a process ID (a “pid”). This will involve creating a new logical address space, loading this address space with bytes from the current address space, and creating a single thread to run in the new space. The kernel must also copy the CPU machine state (i.e., the registers) of the current process and start the new process off running with this machine state. Thus, the newly created process is an exact clone of the first process.

While we don’t have file I/O yet, the Fork syscall must also ensure that all files that are open in the parent will also be open in the child. You will need to add this functionality to Fork in a later project, when file I/O is implemented.

The initial “parent” process is the one invoking the Fork. Its thread will do the work of creating the new process. The new process and its thread are immediately runnable, so the new thread should be placed on the ready list.

In the parent, after creating the new process, the thread returns to the user-level code. It returns the pid of the child process. In the child process, since it has exactly the same state, once it runs, it too should return from the Fork syscall. However, the value returned should be zero.

If we make an exact copy of the machine state, an exact copy of the system stack, an exact copy of the virtual address space, and an exact copy of the user stack, it is easy to resume the execution of the child process. However, it will do exactly what the parent process did. It is rather difficult to make it do anything different, such as returning zero instead of returning the pid.

The Parent-Child Relationship

Each process will have a process ID and these are unique across all processes, past, present and future. (Of course, with a finite sized integer, there is the possibility that the counter will wrap around; this would be disastrous, but we’ll ignore the possibility.) A new pid should have been assigned in the ProcessManager.GetANewProcess method you wrote earlier.

Each time a process does a Fork, it creates a child process. A single process may create zero, one, or many children. The children may go on to create other children, which are called “descendants” of the original “ancestor.” The parent may terminate before its children, or some or all of its children may terminate first.

Given a process, you will occasionally need to know which process is its parent. The ProcessControlBlock contains a field called parentsPid, which you can use. The simplest approach is to search the processManager.processTable array, looking for a process with that pid. This linear search will take a little while, but not too long. Likewise, if you want the child of a process P, you can do a linear search over the array looking for a process whose parentsPid is P.

In general, linear searches are to be avoided in OS kernels, but in our simplified OS, a linear search of PCBs is acceptable.

You might think that it would be smarter to store a pointer to the PCB of the parent right in the PCB of the child.

class ProcessControlBlock

fields

...

parent: ptr to ProcessControlBlock -- an idea???

...

endClass

Unfortunately, there is a problem with this approach. PCBs are recycled when a process terminates and are then used for other processes. If we store a pointer to a PCB in the parent field, when we need the parent and we follow this pointer, we might get a PCB that now holds a completely unrelated process.

Although the linear search approach is just fine, a better design would be to store both a pointer to the parent’s PCB and the parent’s pid. Then you can follow the pointer to a PCB and then check to make sure that pid of this PCB is the pid of the parent.

[Our OS will differ a little from Unix/Linux in how we deal with a parent which has terminated. In our OS, the parent (P) of process (C) may have terminated, and you will need to check for this possibility. In Unix/Linux, when a process P terminates, all of its children processes are given a new parent. In particular, all children are “reparented” to become children of whichever process was previously the parent of P. In other words, the “grandparent” takes over as parent when the parent dies. Thus, in Unix/Linux, every process will always have a parent.]

Implementing Handle-Sys-Fork

The code that implements the Fork syscall needs to do (more-or-less) the following:

1. Allocate and set up new Thread and ProcessControlBlock objects.

2. Make a copy of the address space.

3. Invoke Thread.Fork to start up the new process’s thread.

4. Return the child’s pid.

The newly started thread needs to do the following:

A. Initialize the user registers.

B. Initialize the user and system stacks.

C. Figure out where in the user’s address space to “return” to.

D. Invoke BecomeUserThread to jump into the new user-level process.

Let’s call the initial function to be executed by the newly created thread, ResumeChildAfterFork. This is a kernel function you’ll need to add to Kernel.c. The Handle_Sys_Fork function will perform steps 1 through 4 above, which will include creating a new thread. The new thread will begin by executing the ResumeChildAfterFork, which will perform steps A through D, completing the operation of starting the new process.

There are many details, so next we will go through these steps more carefully.

The first thing to do in Handle_Sys_Fork is to obtain a new ProcessControlBlock and a new Thread object.

Next, you’ll need to initialize the following fields in the PCB: myThread and parentsPid. (The pid field should have been initialized in GetANewProcess.)

You’ll also need to initialize the following fields in the new Thread object: name, status, and myProcess.

Recall that a user level program was executing and it did a Fork syscall. At that point, the state of the user-level process was contained in the user registers. These registers have not changed since the system call. (Well, maybe the call to GetANewThread or GetANewProcess caused this thread to be suspended for a while. But during any intervening process switches, the user registers would have been saved and subsequently restored.)

Recall that the BLITZ CPU has 2 sets of registers: system and user registers. Some of the time, threads run in system mode (when handling an exception or a syscall) and some of the time, they run in user mode. Sometimes we talk about the top-half and the bottom-half of a thread. The top-half is the kernel part, the part that runs in system mode. The bottom-half is the part of the thread that runs in user mode.

Next you must grab the values in the user registers and store a copy of them in the new Thread object. You can use SaveUserRegs to do this.

Recall that all syscall handlers begin running with interrupts disabled. After getting the user registers, it would be a good idea to re-enable interrupts so that other threads can share the CPU.

We are in the middle of starting a new thread and this new thread will need a system stack. In terms of the bottom-half, the new thread must be a duplicate of the bottom-half of the current thread, but the top-half need not be the same. In a few instructions, we are going to do a Fork on this Thread. We don’t need anything from the system stack of the current thread. So there is no reason to copy the system stack. The systemStack array has already been initialized so there is no need to do that again. You can simply leave the contents (left over from some previous thread) in the array. All you’ll need to do is initialize the stackTop pointer, which should be initialized to point to the very last word in the systemStack array, just as it was in the Thread.Init method.

newThrd.stackTop = & (newThrd.systemStack[SYSTEM_STACK_SIZE-1])

There is no reason to initialize the system registers for the new top-half. They can just pick up leftover values from some previous thread.

In this project, we have not yet implemented the file-related syscalls (Open, Read, Write, Close, etc.), so we don’t have to do anything related to open files. However, in a future project, each process will have some open files. The child process should share the open files of the parent. In other words, if a file is open in the parent before the Fork, it should be open in the child after the Fork.

So at this point in Handle_Sys_Fork, it is recommended that you add the following line:

-- Don’t forget to copy the fileDescriptor array here...

This comment will make sense later.

Next, you’ll need to make a copy of the parent’s virtual address space. You’ll need to see how many pages are in the parent’s address space and call frameManager.GetNewFrames. Then you’ll need to run through each and copy the page. You can use MemoryCopy to do this efficiently. You can use AddrSpace.ExtractFrameAddr to determine where the frames are in physical memory. You’ll also need to set the “writable” bit in the child’s frame to whatever it was set to in the parent’s frame. (See AddrSpace.IsWritable, AddrSpace.SetWritable, and AddrSpace.ClearWritable.)

At this point, you are almost ready to invoke Thread.Fork to start the new thread, but there is one more number you’ll need first. The new thread needs to know where (in the user-level address space) to resume execution and the parent’s top-half must determine that value.

To approach this, ask how does execution return to the bottom-half after any syscall? In the case of a Fork syscall, how will the current thread (the parent process) perform its return to some instruction in the parent’s virtual address space?

When the syscall instruction was executed, the assembly language SyscallTrapHandler was called. It invoked the high-level SyscallTrapHandler function, which in turn called Handle_Sys_Fork. When we are ready to return, the Handle_Sys_Fork function will return to the high-level SyscallTraphandler, which will return to the assembly SyscallTrapHandler, which will execute the “reti” instruction.

At the very beginning of the Fork processing, a “sycall” instruction was executed. When the “syscall” instruction was executed, the CPU pushed an “exception block” onto the system stack. Directly before the CPU jumped to the assembly SyscallTrapHandler, the system stack looked like this:

| |

r15-->| Return Address |

| Status Register |

| Except. Info Word |

| . |

| . |

| . |

---------------------

Subsequently, more stuff was pushed onto the stack when the high-level SyscallTrapHandler was called and more stuff was pushed when the Handle_Sys_Fork function was called, but by the time we return to the point directly before the “reti” instruction, everything that got pushed will have been popped. The “reti” instruction will then pop the 3 words of the exception block and will use the “return address” to determine where to resume user-mode execution.

You’ll need to get that return address from the system stack and you’ll need to get it from within Handle_Sys_Fork. Unfortunately, it will be buried somewhere below the top of the system stack.

Fortunately, there is a function called GetOldUserPCFromSystemStack in Switch.s, which will do exactly what you need.

external GetOldUserPCFromSystemStack () returns int

Here is the assembly code:

!

! =============== GetOldUserPCFromSystemStack ===============

!

! external GetOldUserPCFromSystemStack () returns int

!

! This routine is called by the kernel after a syscall has

! occurred. It expects the assembly SyscallTrapHandler to have

! called the high-level SyscallTrapHandler, which then called

! Handle_Sys_Fork. It expects to be called from Handle_Sys_Fork,

! and will not work properly otherwise.

!

! This routine looks down into stuff buried in the system stack

! and finds the exception block that was pushed onto the stack

! at the time of the syscall. From that, it retrieves the user-mode

! PC, which points to the instruction the kernel must return to

! after the syscall.

!

GetOldUserPCFromSystemStack:

load [r14],r1 ! r1 = ptr to frame of SyscallTrapHandler

load [r1+28],r1 ! r1 = pc from interrupt block

store r1,[r15+4]

ret

You’ll need to call this function; let’s call the value it returns, oldUserPC. You’ll need to pass this address to the ResumeChildAfterFork function so that the child can use it.

You are now ready to fork the new thread. As with any fork, you need to provide a pointer to a function and a single integer argument. As an argument, you can pass oldUserPC.

newThrd.Fork (ResumeChildAfterFork, oldUserPC)

Once you have called Thread.Fork, the new thread will finish the work of returning to the child process and will become the thread of the newly created process. The parent thread is now ready to return from Handle_Sys_Fork to the parent user-level process.

Next, let’s look at what you’ll need to do in the ResumeChildAfterFork function.

The key piece of info ResumeChildAfterFork needs (besides the info stored in the Thread) is the address in the user program to return to. This is just the value returned from GetOldUserPCFromSystemStack which was passed as an argument to ResumeChildAfterFork.

Basically, ResumeChildAfterFork needs to switch into user mode and jump to this address. Fortunately, we have an assembly routine that does just this: BecomeUserThread.

Notice that ResumeChildAfterFork will bear a strong resemblance to the code in StartUserProcess.

Every thread begins with interrupts enabled. Since you will need to do things that might involve race conditions, you should begin by disabling interrupts.

Next, you’ll need to initialize the page table registers to point to the page table for the child process, so invoke AddrSpace.SetToThisPageTable.

Then, you’ll need to set the user registers before returning to user mode. You have the values of the registers (stored in the Thread object) but you need to copy these values into the registers. This is exactly what the RestoreUserRegs function does.

You’ll also need to set isUserThread to true. [The isUserThread field in Thread is used for one thing: to determine whether the user registers should be saved and restored every time a context switch occurs. This variable is consulted only in the Run function just before and after calling Switch.]

Once you begin executing the user-level code, you’ll want an empty system stack. You can compute the initial value for the system stack top just as you did in StartUserProcess.

The initial value for the user stack top has been stored in user register r15, which was stored in the Thread object by Handle_Sys_Fork.

Finally, you can jump into the user-level process with the following:

BecomeUserThread (initUserStackTop, -- Initial Stack

initPC, -- Initial PC

initSystemStackTop) -- Initial System Stack

Recall that we need to return pid=0 to the child process. To see how to do this, we need to see how any value is returned from any syscall.

Look at it from the user-level code’s point-of-view. The user-level code executed a “syscall” instruction and, after the kernel has returned, the user code will expect the return value to have been placed in user register r1. See DoSyscall in UserRuntime.s for details.)

When you invoke BecomeUserThread, a jump will be made to the instruction immediately after the syscall. All the registers and the entire address space will be identical to what they were before the syscall was executed (in the parent), so the child will see this as a normal “return” from a kernel syscall. The instructions just after the syscall will fetch the value in r1 and return it to the high-level KPL Sys_Fork function. So all you have to do in the kernel is make sure the right value (namely, zero) is in user register r1.

Luckily for you, BecomeUserThread just happens to store a zero in user register r1 right before making the jump into user-land. (Do you suppose this was a coincidence???) This will cause the Sys_Fork user-level function to get a returned result of zero, which it returns as a pid to distinguish the child from the parent.

Next, let’s talk about two subtleties in the implementation of the Handle_Sys_Fork.

First, consider this race possibility: Let’s say your code in Handle_Sys_Fork ends by invoking Fork to start the new thread running and then returning newThrd.pid. So the last lines in Handle_Sys_Fork might be something like this:

newThrd.Fork (ResumeChildAfterFork, oldUserPC)

return newPCB.pid

Now suppose that right after Fork is invoked, the child gets scheduled before the parent gets to do the return. What if the child starts up, finishes the syscall, returns to the user program, and then the user-level child process runs all the way to completion and the child process terminates altogether? Could the PCB be returned to the free pool, then reallocated to some other process and have a new pid value stored in it, before the parent gets to execute its return statement? When the parent finally resumes, might it grab the pid out of the PCB (getting a wrong value!) and return that (wrong) pid to the user-level code?

No, this cannot happen. The child may terminate before the parent fetches the pid out of the PCB but child PCB will become a zombie and will not be given to another process until after the parent terminates. See the discussion of Zombies below for details.

Second, consider in more detail the call to RestoreUserRegs and the assignment of true to isUserThread.

Once you set isUserThread to true, any time there is a context switch, the user registers will be copied to the Thread object (overwriting anything stored there!) as part of a switch from one thread to another. Therefore, you must call RestoreUserRegisters before setting isUserThread to true, or else a timer interrupt might cause a thread switch, which would wipe out the user register data stored in the Thread object, if it happened to occur before the call to RestoreUserRegs completed.

One the other hand, if you call RestoreUserRegs before setting isUserThread to true, it is possible that a context switch could occur after initializing the user registers but before you set isUserThread to true. The thread scheduler will see that isUserThread is false and will not save the user registers. Any intervening processes might change the user registers. Again, the register values get lost!

It seems that both orders are subject to a race bug, but there is a simple solution.

Recall that every thread initially begins with interrupts enabled. The solution is to disable interrupts in ResumeChildAfterFork. Then there is no possibility of a context switch between the call to RestoreUserRegs and setting isUserThread to true. Also recall that BecomeUserThread will reenable interrupts as part of the process of resuming execution in user mode.

The Semantics of Join and Exit: Why Zombies are Needed

Just as in Unix, when a process terminates, it provides an exit code. This is provided in the call to Sys_Exit. For a “normal” termination, the convention is that zero is returned. The PCB contains a field to contain this value, called exitStatus. Your kernel must keep the PCB around to hold this value until it is no longer needed.

Once a process terminates, the kernel can release all of its resources, such as the Thread object and any OpenFile or FCB objects that are no longer needed. Only the PCB needs to remain around. The PCB then becomes a “zombie.” A zombie is a creature that has stopped living but is not quite dead. (In real life, zombies roam the earth at night terrorizing teenagers staying in remote cabins.)

When can a zombie PCB be freed? Whenever (1) its parent either dies or becomes a zombie itself, or (2) its parent executes a join and takes the exit status, or (3) its parent doesn’t even exist.

The following approach is recommended: First, ignore the ProcessManager.FreeProcess method. In other words, either get rid of it or just don’t ever call it. (We asked you to write it so it could be used to test ProcessManager.GetANewProcess.)

[By the way, whenever you have a routine you wish to keep but will not ever use, I recommend placing a line like this as the first statement:

method FreeProcess (p: ptr to ProcessControlBlock)

FatalError ("Never called")

...

endMethod

This way, you still have the code in case you change your mind, but it is clear when reading your code that it is not used. Also, if you make a mistake and try to use the routine, you’ll get an immediate error.]

Second, add two new methods to ProcessManager: TurnIntoZombie and WaitForZombie. Also, you’ll need to implement the ProcessFinish function.

function ProcessFinish (exitStatus: int)

This function is called when a process is to be terminated. This function is called by the process’s thread. It will free all resources held by this process and will terminate the current thread. The PCB will be turned into a zombie. This method will have to do the following...

First, save the exitStatus in the PCB.

Next, disable interrupts.

Next, disconnect the PCB and the Thread, i.e., set myProcess and myThread to null. Also, set isUserThread to false.

Next, reenable interrupts.

Next, close any open files. (For the next project.)

Next, return all page frames to the free pool, by calling frameManager.ReturnAllFrames.

Next, invoke TurnIntoZombie on this PCB.

Finally, invoke ThreadFinish.

method TurnIntoZombie (p: ptr to ProcessControlBlock)

This method is passed p, a pointer to a process; It turns it into a zombie - dead but not gone! - so that its exitStatus can be retrieved if needed by its parent.

First, lock the process manager since you’ll be messing with other PCBs.

Next, identify all children of this process who are zombies; These children are now no longer needed so for each zombie child, change its status to FREE and add it back to the PCB free list. Don’t forget to signal the aProcessBecameFree condition variable, since other threads may be waiting for free PCBs.

Next, identify p’s parent. (Note, the parent may have already terminated, so there might not be a parent.)

If p’s parent is ACTIVE, then this method must turn p into a zombie. Execute a Broadcast on the aProcessDied condition variable, because the parent of p may be waiting for p to exit.

Otherwise (i.e., if our parent is a zombie or is non-existent) then we do not need to turn p into a zombie, so just change p’s status to FREE, add it to the PCB free list, and signal the aProcessBecameFree condition variable.

Finally, unlock the process manager.

method WaitForZombie (proc: ptr to ProcessControlBlock) returns int

This method is passed a pointer to a process; It waits for that process to turn into a zombie. Then it saves its exitStatus and adds the PCB back to the free list. Finally, it returns the exitStatus.

First, lock the process manager.

Next, wait until the status of proc is ZOMBIE, using a while loop and the aProcessDied condition variable.

Next, fetch proc’s exitStatus.

Next, change proc’s status to FREE, add it to the PCB free list, and signal the aProcessBecameFree condition variable.

Finally, unlock the process manager and return the exit status.

Handle_Sys_Exit is now straightforward to implement: just call ProcessFinish, which will store the exit status in the PCB and free all resources except the PCB. ProcessFinish will also free the PCB too if it is not needed.

Handle_Sys_Join is also fairly straightforward. First, you have to identify the child process and make sure that the pid that is passed in is the pid of a valid process and that it is truly a child of this process. (If not, the kernel should return –1 to the caller.) Then you can call WaitForZombie and return whatever it returns.

Implementing Sys-Yield

There is not really a need for a Yield syscall in any OS that has preemptive scheduling, but it is helpful in testing, to make sure that other processes are really running.

Within Handle_Sys_Yield, you can simply invoke Thread.Yield on the current thread and return.

When executed, the scheduler will be invoked and other threads will get a chance to run. Sometime later, this thread will run again (when the call to Yield returns) and a return will be made to the user-level code.

Optional Enhancement

You might want to consider modifying the exception handlers (such as AddressExceptionHandler) so that they print the pid of the offending process, instead of the thread name, which will always be “UserProgram.”)

The User-Level Programs

The p6 directory contains the following user-level programs:

MyProgram -- For you to use during debugging

TestProgram1 -- Do not modify

TestProgram2 -- Do not modify

TestProgram3 -- Do not modify

You may modify MyProgram any way you wish while testing.

The remaining three programs constitute my test suite. TestProgram1 and TestProgram2 are from the last project while TestProgram3 contains the new tests for this project.

What to Hand In

The TestProgram3 main function looks like this:

function main ()

-- SysExitTest ()

-- BasicForkTest ()

-- YieldTest ()

-- ForkTest ()

-- JoinTest1 ()

-- JoinTest2 ()

-- JoinTest3 ()

-- JoinTest4 ()

-- ManyProcessesTest1 ()

-- ManyProcessesTest2 ()

-- ManyProcessesTest3 ()

-- ErrorTest ()

Sys_Exit (0)

endFunction

An individual test can be run by uncommenting the appropriate line, compiling the code, and running it. For example, to run the “basic fork” test, uncomment the line that calls BasicForkTest, compile, and run. To run another test, go back and re-comment the BasicForkTest and uncomment some other line.

After you have finished coding, please run each test once and hand in the output from each test.

Use the same code to execute all tests. Please hand in only one copy of your Kernel.c file and do not hand in any output that was produced by a different version of your code! [You should never tweak your code to pass tests; this is a symptom of a non-functional program.] The goal is to create one program that can pass all the tests we can throw at it.

Do not change TestProgram3, except to uncomment one of the lines in the main function.

During your testing, it may be convenient to modify the tests as you try to see what is going on and get things to work. Before you make your final test runs, please recopy TestProgram3.c from our directory, so that you get a fresh, unaltered version.

Please hand in hardcopy of Kernel.c. You only need to hand in...

Handle_Sys_Exit

Handle_Sys_Join

Handle_Sys_Fork

Handle_Sys_Yield

ResumeChildAfterFork

ProcessFinish

TurnIntoZombie

WaitForZombie

and any other code you wrote / modified.

Sample Output

If your program works correctly, you should see something like this:

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

SysExitTest running.

About to terminate the only process; should cause the OS to stop on a 'wait' instruction.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

BasicForkTest running.

I am the parent

I am the child

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

YieldTest running.

This test involves calls to Fork, Yield, and Exit.

RUN ONE: You should see 10 'compiler' messages and 10 'OS' messages.

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

RUN TWO: You should see the same 20 messages, but the order should be different, due to the presence of 'Yield's.

Designing compilers is fun!

Designing compilers is fun!

Writing OS kernel code is a blast!

Designing compilers is fun!

Designing compilers is fun!

Writing OS kernel code is a blast!

Designing compilers is fun!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Designing compilers is fun!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Designing compilers is fun!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Writing OS kernel code is a blast!

Designing compilers is fun!

Designing compilers is fun!

Designing compilers is fun!

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

ForkTest running.

This test involves calls to Fork, Yield, and Exit.

There should be 26 columns (A-Z) printed. Each letter should be printed 5 times.

A

A

B

B

A

C

C

B

D

A

C

D

B

D

C

A

B

D

E

C

E

E

D

F

F

...A BUNCH OF STUFF, DELETED IN THIS DOCUMENT...

W

W

U

V

W

X

X

V

W

X

Y

Y

W

X

Y

Z

Z

X

Y

Z

Y

Z

Z

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

JoinTest 1 running.

This test involves calls to Fork, Yield, and Exit.

Running first test...

This line should print first.

This line should print second.

Done.

Running second test...

This line should print first.

This line should print second.

Done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

JoinTest 2 running.

This test involves calls to Fork, Yield, and Exit.

Creating 5 children...

Child 1 running...

Child 2 running...

Child 3 running...

Child 4 running...

Waiting for children in order 1, 2, 3, 4, 5...

Child 5 running...

Creating 5 more children...

Child 1 running...

Child 2 running...

Child 3 running...

Child 4 running...

Waiting for children in order 5, 4, 1, 3, 2...

Child 5 running...

Done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

JoinTest3 running.

This test involves 5 illegal calls to Sys_Join, waiting on non-existent children.

In each case, it prints the return code, which should be -1.

Return code from 1st call = -1

Return code from 2nd call = -1

Return code from 3rd call = -1

Return code from 4th call = -1

Return code from 5th call = -1

Done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

JoinTest4 running.

This test forks a child process and then waits on it twice.

The first call to Sys_Join should return its error code; the

second call to Sys_Join should return -1.

The PID of the child = 2

This should print first.

This should print second.

Okay (1).

Okay (2).

This should print first.

The PID of the child = 3

This should print second.

Okay (3).

Okay (4).

In the next test, we create 2 children, and each creates 2 children,

giving 7 processes in all. Then each process attempts a Sys_Join on

every process except its own children, to make sure the result is -1.

Finally, each process with children waits on them.

A is running...

My first child is A.B pid1 = 4

My second child is A.C pid2 = 5

---------------

A.B.D is running...

---------------

A.C.F is running...

---------------

A.C is running...

My first child is A.C.F pid1 = 7

My second child is A.C.G pid2 = 9

---------------

A.C.G is running...

---------------

A.B is running...

My first child is A.B.D pid1 = 6

My second child is A.B.E pid2 = 8

---------------

A.B.E is running...

---------------

A done with error tests...

A.C.F done with error tests...

A.C done with error tests...

A.C.G done with error tests...

A.B.E done with error tests...

A.B done with error tests...

A.B.D done with error tests...

-----------------------------------A is waiting on A.B pid1 = 4

-----------------------------------A.C is waiting on A.C.F pid1 = 7

-----------------------------------A.B is waiting on A.B.D pid1 = 6

A.C.F is done.

-----------------------------------A.C is waiting on A.C.G pid2 = 9

A.C.G is done.

A.C is done.

A.B.E is done.

A.B.D is done.

-----------------------------------A.B is waiting on A.B.E pid2 = 8

A.B is done.

-----------------------------------A is waiting on A.C pid2 = 5

A is done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

ManyProcessesTest1 running.

This test should create 100 child processes.

It should print 100 lines of output.

Child 1

Child 2

Child 3

Child 4

Child 5

...A BUNCH OF STUFF, DELETED IN THIS DOCUMENT...

Child 92

Child 93

Child 94

Child 95

Child 96

Child 97

Child 98

Child 99

Child 100

Done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

ManyProcessesTest2 running.

This test attempts to create 9 new processes.

It should print a line for each process and then it should print 123.

Process 0

Process 1

Process 2

Process 3

Process 4

Process 5

Process 6

Process 7

Process 8

Process 9

Final return value = 123

Done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

ManyProcessesTest3 running.

This test attempts to create 10 new processes.

It should run out of resources and hang.

Process 0

Process 1

Process 2

Process 3

Process 4

Process 5

Process 6

Process 7

Process 8

Process 9

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****

=================== KPL PROGRAM STARTING ===================

Initializing Thread Scheduler...

Initializing Process Manager...

Initializing Thread Manager...

Initializing Frame Manager...

AllocateRandomFrames called. NUMBER_OF_PHYSICAL_PAGE_FRAMES = 512

Initializing Disk Driver...

Initializing Serial Driver...

Initializing File Manager...

Serial handler thread running...

Loading initial program...

ErrorTest running.

Should print "An AddressException exception has occured while in user mode"...

********** An AddressException exception has occured while in user mode **********

ProcessControlBlock

...A BUNCH OF STUFF, DELETED IN THIS DOCUMENT...

Thread "UserProgram"

...A BUNCH OF STUFF, DELETED IN THIS DOCUMENT...

Okay.

Should print "A PageReadonlyException exception has occured while in user mode"...

********** A PageReadonlyException exception has occured while in user mode **********

ProcessControlBlock

...A BUNCH OF STUFF, DELETED IN THIS DOCUMENT...

Thread "UserProgram"

...A BUNCH OF STUFF, DELETED IN THIS DOCUMENT...

Okay.

Done.

***** A 'wait' instruction was executed and no more interrupts are scheduled... halting emulation *****