Project 1: Threads

Due: Sunday 09/26 11:59 pm EDT Tuesday 09/28 11:59 pm EDT

In this assignment, we give you a minimally functional thread system. Your job is to extend the functionality of this system to gain a better understanding of synchronization problems.

You will be working primarily in the threads directory for this assignment, with some work in the devices directory on the side. Compilation should be done in the threads directory.

Before you read the description of this project, you should read all of the following sections: 1. Introduction, C. Coding Standards, Testing and Debugging. You should at least skim the material from A.1 Loading through A.5 Memory Allocation, especially A.3 Synchronization. To complete this project you will also need to read B. 4.4BSD Scheduler.


Background


Understanding Threads

The first step is to read and understand the code for the initial thread system. Pintos already implements thread creation and thread completion, a simple scheduler to switch between threads, and synchronization primitives (semaphores, locks, condition variables, and optimization barriers).

Some of this code might seem slightly mysterious. If you haven't already compiled and run the base system, as described in the introduction (see section 1. Introduction), you should do so now. You can read through parts of the source code to see what's going on. If you like, you can add calls to printf() almost anywhere, then recompile and run to see what happens and in what order. You can also run the kernel in a debugger and set breakpoints at interesting spots, single-step through code and examine data, and so on.

When a thread is created, you are creating a new context to be scheduled. You provide a function to be run in this context as an argument to thread_create(). The first time the thread is scheduled and runs, it starts from the beginning of that function and executes in that context. When the function returns, the thread terminates. Each thread, therefore, acts like a mini-program running inside Pintos, with the function passed to thread_create() acting like main().

At any given time, exactly one thread runs and the rest, if any, become inactive. The scheduler decides which thread to run next. (If no thread is ready to run at any given time, then the special "idle" thread, implemented in idle(), runs.) Synchronization primitives can force context switches when one thread needs to wait for another thread to do something.

The mechanics of a context switch are in threads/switch.S, which is 80x86 assembly code. (You don't have to understand it.) It saves the state of the currently running thread and restores the state of the thread we're switching to.

Using the GDB debugger, slowly trace through a context switch to see what happens (see section E.5 GDB). You can set a breakpoint on schedule() to start out, and then single-step from there.(1) Be sure to keep track of each thread's address and state, and what procedures are on the call stack for each thread. You will notice that when one thread calls switch_threads(), another thread starts running, and the first thing the new thread does is to return from switch_threads(). You will understand the thread system once you understand why and how the switch_threads() that gets called is different from the switch_threads() that returns. See section A.2.3 Thread Switching, for more information.

  Warning
In Pintos, each thread is assigned a small, fixed-size execution stack just under 4 kB in size. The kernel tries to detect stack overflow, but it cannot do so perfectly. You may cause bizarre problems, such as mysterious kernel panics, if you declare large data structures as non-static local variables, e.g. int buf[1000];. Alternatives to stack allocation include the page allocator and the block allocator (see section A.5 Memory Allocation).


Source Files

Listing 1 provides a brief overview of the files in the threads directory. You will not need to modify most of this code, but the hope is that presenting this overview will give you a start on what code to look at.


Synchronization

Proper synchronization is an important part of the solutions to these problems. Any synchronization problem can be easily solved by turning interrupts off: while interrupts are off, there is no concurrency, so there's no possibility for race conditions. Therefore, it's tempting to solve all synchronization problems this way, but don't. Instead, use semaphores, locks, and condition variables to solve the bulk of your synchronization problems. Read the tour section on synchronization (see section A.3 Synchronization) or the comments in threads/synch.c if you're unsure what synchronization primitives may be used in what situations.

In the Pintos projects, the only class of problem best solved by disabling interrupts is coordinating data shared between a kernel thread and an interrupt handler. Because interrupt handlers can't sleep, they can't acquire locks. This means that data shared between kernel threads and an interrupt handler must be protected within a kernel thread by turning off interrupts.

This project only requires accessing a little bit of thread state from interrupt handlers. For the alarm clock, the timer interrupt needs to wake up sleeping threads. In the advanced scheduler, the timer interrupt needs to access a few global and per-thread variables. When you access these variables from kernel threads, you will need to disable interrupts to prevent the timer interrupt from interfering.

When you do turn off interrupts, take care to do so for the least amount of code possible, or you can end up losing important things such as timer ticks or input events. Turning off interrupts also increases the interrupt handling latency, which can make a machine feel sluggish if taken too far.

The synchronization primitives themselves in synch.c are implemented by disabling interrupts. You may need to increase the amount of code that runs with interrupts disabled here, but you should still try to keep it to a minimum.

Disabling interrupts can be useful for debugging, if you want to make sure that a section of code is not interrupted. You should remove debugging code before turning in your project. (Don't just comment it out, because that can make the code difficult to read.)

There should be no busy waiting in your submission. A tight loop that calls thread_yield() is one form of busy waiting.


Development Suggestions

In the past, many groups divided the assignment into pieces, then each group member worked on his or her piece until just before the deadline, at which time the group reconvened to combine their code and submit. This is a bad idea. We do not recommend this approach. Groups that do this often find that two changes conflict with each other, requiring lots of last-minute debugging. Some groups who have done this have turned in code that did not even compile or boot, much less pass any tests.

Instead, we recommend integrating your team's changes early and often, using a source code control system such as Git (see section F.3 Git). This is less likely to produce surprises, because everyone can see everyone else's code as it is written, instead of just when it is finished. These systems also make it possible to review changes and, when a change introduces a bug, drop back to working versions of code.

You should expect to run into bugs that you simply don't understand while working on this and subsequent projects. When you do, reread the appendix on debugging tools, which is filled with useful debugging tips that should help you to get back up to speed (see section E. Debugging Tools). Be sure to read the section on backtraces (see section E.4 Backtraces), which will help you to get the most out of every kernel panic or assertion failure.

  Tips
We also suggest you to read the requirement description as well as the FAQs carefully---perhaps multiple times. A lot of the confusions come from not reading them carefully.
  Tips
make check runs all tests. If you are debugging one test, you can run just that test with make tests/threads/<test_case>.result. Read the Testing part for more details.

Suggested Order of Implementation

We suggest first implementing the following, which can happen in parallel:

  • Optimized timer_sleep() in Exercise 1.1.

  • Basic support for priority scheduling in Exercise 2.1 when a new thread is created in thread_create().

  • Implement fixed-point real arithmetic routines (B.6 Fixed-Point Real Arithmetic) that are needed for MLFQS in exercise 3.1.

Then you can add full support for priority scheduling in Exercise 2.1 by considering all other possible scenarios and the synchronization primitives.

Then you can tackle either Exercise 3.1 first or Exercise 2.2 first.


Project 1 Requirements


0. Design Document

Before you turn in your project, you must copy the project 1 design document template into your source tree under the name pintos/src/threads/DESIGNDOC and fill it in. We recommend that you read the design document template before you start working on the project. See section D. Project Documentation, for a sample design document that goes along with a fictitious project.


1. Alarm Clock

Exercise 1.1

  Exercise 1.1
Reimplement timer_sleep(), defined in devices/timer.c. Although a working implementation is provided, it "busy waits," that is, it spins in a loop checking the current time and calling thread_yield() until enough time has gone by. Reimplement it to avoid busy waiting.

Function: void timer_sleep (int64_t ticks)
Suspends execution of the calling thread until time has advanced by at least x timer ticks. Unless the system is otherwise idle, the thread need not wake up after exactly x ticks. Just put it on the ready queue after they have waited for the right amount of time.

timer_sleep() is useful for threads that operate in real-time, e.g. for blinking the cursor once per second.

The argument to timer_sleep() is expressed in timer ticks, not in milliseconds or any another unit. There are TIMER_FREQ timer ticks per second, where TIMER_FREQ is a macro defined in devices/timer.h. The default value is 100. We don't recommend changing this value, because any change is likely to cause many of the tests to fail.

Separate functions timer_msleep(), timer_usleep(), and timer_nsleep() do exist for sleeping a specific number of milliseconds, microseconds, or nanoseconds, respectively, but these will call timer_sleep() automatically when necessary. You do not need to modify them.

If your delays seem too short or too long, reread the explanation of the -r option to pintos (see section 1.1.4 Debugging versus Testing).

  Hint
You may find struct list and the provided functions to be useful for this exercise. Read the comments in lib/kernel/list.h carefully, since this list design/usage is different from the typical linked list you are familiar with (actually, Linux kernel uses a similar design). Searching the Pintos codebase to see how struct list is used may also give you some inspirations.
  Hint
You may want to leverage some synchronization primitive that provides some sort of thread "waiting" functionality, e.g., semaphore. You do not have to wait for the Synchronization Lecture to be able to use these primitives. Reading through section A.3 Synchronization is sufficient. In addition, when modifying some global variable, e.g., a global list, you will need to use some synchronization primitive as well to ensure it is not modified or read concurrently (e.g., a timer interrupt occurs during the modification and we switch to run another thread).
  Hint
You need to decide where to check whether the elapsed time exceeded the sleep time. The original timer_sleep implementation calls timer_ticks, which returns the current ticks. Check where the static ticks variable is updated. You can search with grep or use cscope to help you find this out: type make cscope, then type cscope -d, in the Find this C symbol: type ticks.

The alarm clock implementation is not needed for later projects, although it could be useful for project 4.


2. Priority Scheduling

Exercise 2.1

  Exercise 2.1
Implement priority scheduling in Pintos. When a thread is added to the ready list that has a higher priority than the currently running thread, the current thread should immediately yield the processor to the new thread. Similarly, when threads are waiting for a lock, semaphore, or condition variable, the highest priority waiting thread should be awakened first. A thread may raise or lower its own priority at any time, but lowering its priority such that it no longer has the highest priority must cause it to immediately yield the CPU.

Thread priorities range from PRI_MIN (0) to PRI_MAX (63). Lower numbers correspond to lower priorities, so that priority 0 is the lowest priority and priority 63 is the highest. The initial thread priority is passed as an argument to thread_create(). If there's no reason to choose another priority, use PRI_DEFAULT (31). The PRI_ macros are defined in threads/thread.h, and you should not change their values.

  Hint
For this exercise, you need to consider all the scenarios where the priority must be enforced. For example, when an alarm clock for a thread fires off, that thread should be made ready again, which entails a priority check. You can find some of these scenarios by looking for places that modify ready_list (directly and indirectly, cscope can be helpful).
  Hint
To yield the CPU, you can check the thread APIs in threads/thread.h. Read the comment and implementation of the corresponding thread function in threads/thread.c. That function may not be used in interrupt context (i.e., should not call it inside an interrupt handler). To yield the CPU in the interrupt context, you can take a look at functions in threads/interrupt.c. Again, if you want to use an existing function, you can use cscope to find the references to get a sense of how that function is used.

One issue with priority scheduling is "priority inversion". Consider high, medium, and low priority threads H, M, and L, respectively. If H needs to wait for L (for instance, for a lock held by L), and M is on the ready list, then H will never get the CPU because the low priority thread will not get any CPU time. A partial fix for this problem is for H to "donate" its priority to L while L is holding the lock, then recall the donation once L releases (and thus H acquires) the lock.

Exercise 2.2

  Exercise 2.2
Implement priority donation. You will need to account for all different situations in which priority donation is required. You must implement priority donation for locks. You need not implement priority donation for the other Pintos synchronization constructs. You do need to implement priority scheduling in all cases. Be sure to handle multiple donations, in which multiple priorities are donated to a single thread.

Challenge (Extra Credit)
Support nested priority donation: if H is waiting on a lock that M holds and M is waiting on a lock that L holds, then both M and L should be boosted to H's priority. If necessary, you may impose a reasonable limit on depth of nested priority donation, such as 8 levels.
Note: if you decide not to work on this extra credit exercise, you do not need to pass the priority-donate-nest and priority-donate-chain tests.

Exercise 2.3

  Exercise 2.3
Implement the following functions that allow a thread to examine and modify its own priority. Skeletons for these functions are provided in threads/thread.c.

Function: void thread_set_priority (int new_priority)
Sets the current thread's priority to new_priority. If the current thread no longer has the highest priority, yields.

Function: int thread_get_priority (void)
Returns the current thread's priority. In the presence of priority donation, returns the higher (donated) priority.

You need not provide any interface to allow a thread to directly modify other threads' priorities.

The priority scheduler is not used in any later project.


3. Advanced Scheduler

Exercise 3.1

  Exercise 3.1
Implement a multilevel feedback queue scheduler similar to the 4.4BSD scheduler to reduce the average response time for running jobs on your system. See section B. 4.4BSD Scheduler, for detailed requirements.

Like the priority scheduler, the advanced scheduler chooses the thread to run based on priorities. However, the advanced scheduler does not do priority donation. Thus, we recommend that you have the priority scheduler working, except possibly for priority donation, before you start work on the advanced scheduler.

You must write your code to allow us to choose a scheduling algorithm policy at Pintos startup time. By default, the priority scheduler must be active, but we must be able to choose the 4.4BSD scheduler with the -mlfqs kernel option. Passing this option sets thread_mlfqs, declared in threads/thread.h, to true when the options are parsed by parse_options(), which happens early in main().

When the 4.4BSD scheduler is enabled, threads no longer directly control their own priorities. The priority argument to thread_create() should be ignored, as well as any calls to thread_set_priority(), and thread_get_priority() should return the thread's current priority as set by the scheduler.

  Hint
Double check the implementations of your fixed-point arithmetic routines (and ideally have some unit test for them). Some simple mistake in these routines could result in mysterious issue in your scheduler.
  Hint
Efficiency matters a lot for the MLFQS exercise. An inefficient implementation can distort the system. Read the comment in test case mlfqs-load-avg.c. In fact, the inefficiency in your alarm clock implementation can also influence your MLFQS behavior. So double check if your implementation there can be optimized.

The advanced scheduler is not used in any later project.


Submission Instruction

  Git Branch

We will collect your solution automatically through GitHub by taking a snapshot by the deadline. Thus, be sure to commit your changes and do a git push to GitHub, especially in the last few minutes! Your submission must reside in a branch called lab1-handin. You can create this branch with git checkout -b lab1-handin. You can use other branches (e.g., master or lab1-dev) during development, but be sure to sync these changes to the submission branch with git checkout lab1-handin then git merge <branch_name>.

Double check that your submission resides in the correct branch lab1-handin (note the middle part is a dash -, not underscore) by the deadline. Using a different branch name will result in failure to collect and grade your submission in time.

  Late Hours
If you decide to use the late hour tokens, fill out this form before the deadline, so that we won't be collecting and grading your solution immediately. When you finish (within the token limit), fill out this form again to indicate you are done. Don't forget to fill out the form for the second time to avoid leaking your late tokens.


A. FAQ

How much code will I need to write?

Here's a summary of our reference solution, produced by the diffstat program. The final row gives total lines inserted and deleted; a changed line counts as both an insertion and a deletion.

The reference solution represents just one possible solution. Many other solutions are also possible and many of those differ greatly from the reference solution. Some excellent solutions may not modify all the files modified by the reference solution, and some may modify files not modified by the reference solution.

 
 devices/timer.c       |   42 +++++-
 threads/fixed-point.h |  120 ++++++++++++++++++
 threads/synch.c       |   88 ++++++++++++-
 threads/thread.c      |  196 ++++++++++++++++++++++++++----
 threads/thread.h      |   23 +++
 5 files changed, 440 insertions(+), 29 deletions(-)

fixed-point.h is a new file added by the reference solution.

How do I update the Makefiles when I add a new source file?

To add a .c file, edit the top-level Makefile.build. Add the new file to variable dir_SRC, where dir is the directory where you added the file. For this project, that means you should add it to threads_SRC or devices_SRC. Then run make. If your new file doesn't get compiled, run make clean and then try again.

When you modify the top-level Makefile.build and re-run make, the modified version should be automatically copied to threads/build/Makefile. The converse is not true, so any changes will be lost the next time you run make clean from the threads directory. Unless your changes are truly temporary, you should prefer to edit Makefile.build.

A new .h file does not require editing the Makefiles.

What does warning: no previous prototype for `func' mean?

It means that you defined a non-static function without preceding it by a prototype. Because non-static functions are intended for use by other .c files, for safety they should be prototyped in a header file included before their definition. To fix the problem, add a prototype in a header file that you include, or, if the function isn't actually used by other .c files, make it static.

What is the interval between timer interrupts?

Timer interrupts occur TIMER_FREQ times per second. You can adjust this value by editing devices/timer.h. The default is 100 Hz.

We don't recommend changing this value, because any changes are likely to cause many of the tests to fail.

How long is a time slice?

There are TIME_SLICE ticks per time slice. This macro is declared in threads/thread.c. The default is 4 ticks.

We don't recommend changing this value, because any changes are likely to cause many of the tests to fail.

How do I run the tests?

See section Testing.

Why do I get a test failure in pass()?

You are probably looking at a backtrace that looks something like this:

 
0xc0108810: debug_panic (lib/kernel/debug.c:32)
0xc010a99f: pass (tests/threads/tests.c:93)
0xc010bdd3: test_mlfqs_load_1 (...threads/mlfqs-load-1.c:33)
0xc010a8cf: run_test (tests/threads/tests.c:51)
0xc0100452: run_task (threads/init.c:283)
0xc0100536: run_actions (threads/init.c:333)
0xc01000bb: main (threads/init.c:137)

This is just confusing output from the backtrace program. It does not actually mean that pass() called debug_panic(). In fact, fail() called debug_panic() (via the PANIC() macro). GCC knows that debug_panic() does not return, because it is declared NO_RETURN (see section E.3 Function and Parameter Attributes), so it doesn't include any code in fail() to take control when debug_panic() returns. This means that the return address on the stack looks like it is at the beginning of the function that happens to follow fail() in memory, which in this case happens to be pass().

See section E.4 Backtraces, for more information.

How do interrupts get re-enabled in the new thread following schedule()?

Every path into schedule() disables interrupts. They eventually get re-enabled by the next thread to be scheduled. Consider the possibilities: the new thread is running in switch_thread() (but see below), which is called by schedule(), which is called by one of a few possible functions:

  • thread_exit(), but we'll never switch back into such a thread, so it's uninteresting.

  • thread_yield(), which immediately restores the interrupt level upon return from schedule().

  • thread_block(), which is called from multiple places:

    • sema_down(), which restores the interrupt level before returning.

    • idle(), which enables interrupts with an explicit assembly STI instruction.

    • wait() in devices/intq.c, whose callers are responsible for re-enabling interrupts.

There is a special case when a newly created thread runs for the first time. Such a thread calls intr_enable() as the first action in kernel_thread(), which is at the bottom of the call stack for every kernel thread but the first.


A.1 Alarm Clock FAQ

Do I need to account for timer values overflowing?

Don't worry about the possibility of timer values overflowing. Timer values are expressed as signed 64-bit numbers, which at 100 ticks per second should be good for almost 2,924,712,087 years. By then, we expect Pintos to have been phased out of the Computer Science curriculum.


A.2 Priority Scheduling FAQ

Doesn't priority scheduling lead to starvation?

Yes, strict priority scheduling can lead to starvation because a thread will not run if any higher-priority thread is runnable. The advanced scheduler introduces a mechanism for dynamically changing thread priorities.

Strict priority scheduling is valuable in real-time systems because it offers the programmer more control over which jobs get processing time. High priorities are generally reserved for time-critical tasks. It's not "fair," but it addresses other concerns not applicable to a general-purpose operating system.

What thread should run after a lock has been released?

When a lock is released, the highest priority thread waiting for that lock should be unblocked and put on the list of ready threads. The scheduler should then run the highest priority thread on the ready list.

If the highest-priority thread yields, does it continue running?

Yes. If there is a single highest-priority thread, it continues running until it blocks or finishes, even if it calls thread_yield(). If multiple threads have the same highest priority, thread_yield() should switch among them in "round robin" order.

What happens to the priority of a donating thread?

Priority donation only changes the priority of the donee thread. The donor thread's priority is unchanged. Priority donation is not additive: if thread A (with priority 5) donates to thread B (with priority 3), then B's new priority is 5, not 8.

Can a thread's priority change while it is on the ready queue?

Yes. Consider a ready, low-priority thread L that holds a lock. High-priority thread H attempts to acquire the lock and blocks, thereby donating its priority to ready thread L.

Can a thread's priority change while it is blocked?

Yes. While a thread that has acquired lock L is blocked for any reason, its priority can increase by priority donation if a higher-priority thread attempts to acquire L. This case is checked by the priority-donate-sema test.

Can a thread added to the ready list preempt the processor?

Yes. If a thread added to the ready list has higher priority than the running thread, the correct behavior is to immediately yield the processor. It is not acceptable to wait for the next timer interrupt. The highest priority thread should run as soon as it is runnable, preempting whatever thread is currently running.

How does thread_set_priority() affect a thread receiving donations?

It sets the thread's base priority. The thread's effective priority becomes the higher of the newly set priority or the highest donated priority. When the donations are released, the thread's priority becomes the one set through the function call. This behavior is checked by the priority-donate-lower test.

Doubled test names in output make them fail.

Suppose you are seeing output in which some test names are doubled, like this:

 
(alarm-priority) begin
(alarm-priority) (alarm-priority) Thread priority 30 woke up.
Thread priority 29 woke up.
(alarm-priority) Thread priority 28 woke up.

What is happening is that output from two threads is being interleaved. That is, one thread is printing "(alarm-priority) Thread priority 29 woke up.\n" and another thread is printing "(alarm-priority) Thread priority 30 woke up.\n", but the first thread is being preempted by the second in the middle of its output.

This problem indicates a bug in your priority scheduler. After all, a thread with priority 29 should not be able to run while a thread with priority 30 has work to do.

Normally, the implementation of the printf() function in the Pintos kernel attempts to prevent such interleaved output by acquiring a console lock during the duration of the printf call and releasing it afterwards. However, the output of the test name, e.g., (alarm-priority), and the message following it is output using two calls to printf, resulting in the console lock being acquired and released twice.


A.3 Advanced Scheduler FAQ

How does priority donation interact with the advanced scheduler?

It doesn't have to. We won't test priority donation and the advanced scheduler at the same time.

Can I use one queue instead of 64 queues?

Yes. In general, your implementation may differ from the description, as long as its behavior is the same.

Some scheduler tests fail and I don't understand why. Help!

If your implementation mysteriously fails some of the advanced scheduler tests, try the following:

  • Read the source files for the tests that you're failing, to make sure that you understand what's going on. Each one has a comment at the top that explains its purpose and expected results.

  • Double-check your fixed-point arithmetic routines and your use of them in the scheduler routines.

  • Consider how much work your implementation does in the timer interrupt. If the timer interrupt handler takes too long, then it will take away most of a timer tick from the thread that the timer interrupt preempted. When it returns control to that thread, it therefore won't get to do much work before the next timer interrupt arrives. That thread will therefore get blamed for a lot more CPU time than it actually got a chance to use. This raises the interrupted thread's recent CPU count, thereby lowering its priority. It can cause scheduling decisions to change. It also raises the load average.

Ryan Huang | Last updated 2021-12-02 14:48:25 -0500.