1 @node Project 1--Threads, Project 2--User Programs, Introduction, Top
2 @chapter Project 1: Threads
4 In this assignment, we give you a minimally functional thread system.
5 Your job is to extend the functionality of this system to gain a
6 better understanding of synchronization problems. Additionally, you
7 will use at least part of this increased functionality in future
10 You will be working in primarily in the @file{threads} directory for
11 this assignment, with some work in the @file{devices} directory on the
12 side. Compilation should be done in the @file{threads} directory.
15 * Understanding Threads::
17 * Debugging versus Testing::
19 * Problem 1-1 Alarm Clock::
21 * Problem 1-3 Priority Scheduling::
22 * Problem 1-4 Advanced Scheduler::
26 @node Understanding Threads
27 @section Understanding Threads
29 The first step is to read and understand the initial thread system.
30 Pintos, by default, implements thread creation and thread completion,
31 a simple scheduler to switch between threads, and synchronization
32 primitives (semaphores, locks, and condition variables).
34 However, there's a lot of magic going on in some of this code, so if
35 you haven't already compiled and run the base system, as described in
36 the introduction (@pxref{Introduction}), you should do so now. You
37 can read through parts of the source code by hand to see what's going
38 on. If you like, you can add calls to @code{printf()} almost
39 anywhere, then recompile and run to see what happens and in what
40 order. You can also run the kernel in a debugger and set breakpoints
41 at interesting spots, single-step through code and examine data, and
42 so on. @xref{i386-elf-gdb}, for more information.
44 When a thread is created, you are creating a new context to be
45 scheduled. You provide a function to be run in this context as an
46 argument to @code{thread_create()}. The first time the thread is
47 scheduled and runs, it will start from the beginning of that function
48 and execute it in the context. When that function returns, that thread
49 completes. Each thread, therefore, acts like a mini-program running
50 inside Pintos, with the function passed to @code{thread_create()}
51 acting like @code{main()}.
53 At any given time, Pintos is running exactly one thread, with the
54 others switched out. The scheduler decides which thread to run next
55 when it needs to switch between them. (If no thread is ready to run
56 at any given time, then the special ``idle'' thread runs.) The
57 synchronization primitives are used to force context switches when one
58 thread needs to wait for another thread to do something.
60 The exact mechanics of a context switch are pretty gruesome and have
61 been provided for you in @file{threads/switch.S} (this is 80@var{x}86
62 assembly; don't worry about understanding it). It involves saving the
63 state of the currently running thread and restoring the state of the
64 thread we're switching to.
66 Using the @command{gdb} debugger, slowly trace through a context
67 switch to see what happens (@pxref{i386-elf-gdb}). You can set a
68 breakpoint on the @code{schedule()} function to start out, and then
69 single-step from there. Be sure to keep track of each thread's
70 address and state, and what procedures are on the call stack for each
71 thread. You will notice that when one thread calls
72 @code{switch_threads()}, another thread starts running, and the first
73 thing the new thread does is to return from
74 @code{switch_threads()}. We realize this comment will seem cryptic to
75 you at this point, but you will understand threads once you understand
76 why the @code{switch_threads()} that gets called is different from the
77 @code{switch_threads()} that returns. @c FIXME
79 @strong{Warning}: In Pintos, each thread is assigned a small,
80 fixed-size execution stack just under @w{4 kB} in size. The kernel
81 does try to detect stack overflow, but it cannot always succeed. You
82 ma cause bizarre problems, such as mysterious kernel panics, if you
83 declare large data structures as non-static local variables,
84 e.g. @samp{int buf[1000];}. Alternatives to stack allocation include
85 the page allocator in @file{threads/palloc.c} and the block allocator
86 in @file{threads/malloc.c}. Note that the page allocator doles out
87 @w{4 kB} chunks and that @code{malloc()} has a @w{2 kB} block size
88 limit. If you need larger chunks, consider using a linked structure
94 Here is a brief overview of the files in the @file{threads}
95 directory. You will not need to modify most of this code, but the
96 hope is that presenting this overview will give you a start on what
102 The kernel loader. Assembles to 512 bytes of code and data that the
103 PC BIOS loads into memory and which in turn loads the kernel into
104 memory, does basic processor initialization, and jumps to the
105 beginning of the kernel. You should not need to look at this code or
109 The linker script used to link the kernel. Sets the load address of
110 the kernel and arranges for @file{start.S} to be at the very beginning
111 of the kernel image. Again, you should not need to look at this code
112 or modify it, but it's here in case you're curious.
115 Jumps to @code{main()}.
119 Kernel initialization, including @code{main()}, the kernel's ``main
120 program.'' You should look over @code{main()} at least to see what
125 Basic thread support. Much of your work will take place in these
126 files. @file{thread.h} defines @code{struct thread}, which you will
127 modify in the first three projects.
131 Assembly language routine for switching threads. Already discussed
136 Page allocator, which hands out system memory one 4 kB page at a time.
140 A very simple implementation of @code{malloc()} and @code{free()} for
141 the kernel. The largest block that can be allocated is 2 kB.
145 Basic interrupt handling and functions for turning interrupts on and
150 A Perl program that outputs assembly for low-level interrupt handling.
154 Basic synchronization primitives: semaphores, locks, and condition
155 variables. You will need to use these for synchronization through all
160 Test code. For project 1, you will replace this file with your test
164 Functions for I/O port access. This is mostly used by source code in
165 the @file{devices} directory that you won't have to touch.
168 Functions and macros related to memory management, including page
169 directories and page tables. This will be more important to you in
170 project 3. For now, you can ignore it.
173 FIXME devices and lib directories?
175 @node Debugging versus Testing
176 @section Debugging versus Testing
178 When you're debugging code, it's useful to be able to be able to run a
179 program twice and have it do exactly the same thing. On second and
180 later runs, you can make new observations without having to discard or
181 verify your old observations. This property is called
182 ``reproducibility.'' The simulator we use, Bochs, can be set up for
183 reproducibility. If you use the Bochs configuration files we provide,
184 which specify @samp{ips: @var{n}} where @var{n} is a number of
185 simulated instructions per second, your simulations can be
188 Of course, a simulation can only be reproducible from one run to the
189 next if its input is the same each time. For simulating an entire
190 computer, as we do, this means that every part of the computer must be
191 the same. For example, you must use the same disks, the same version
192 of Bochs, and you must not hit any keys on the keyboard (because you
193 could not be sure to hit them at exactly the same point each time)
196 While reproducibility is useful for debugging, it is a problem for
197 testing thread synchronization, an important part of this project. In
198 particular, when Bochs is set up for reproducibility, timer interrupts
199 will come at perfectly reproducible points, and therefore so will
200 thread switches. That means that running the same test several times
201 doesn't give you any greater confidence in your code's correctness
202 than does running it only once.
204 So, to make your code easier to test, we've added a feature to Bochs
205 that makes timer interrupts come at random intervals, but in a
206 perfectly predictable way. In particular, if you invoke
207 @command{pintos} with the option @option{-j @var{seed}}, timer
208 interrupts will come at irregularly spaced intervals. Within a single
209 @var{seed} value, execution will still be reproducible, but timer
210 behavior will change as @var{seed} is varied. Thus, for the highest
211 degree of confidence you should test your code with many seed values.
216 There should be no busy-waiting in any of your solutions to this
217 assignment. Furthermore, resist the temptation to directly disable
218 interrupts in your solution by calling @code{intr_disable()} or
219 @code{intr_set_level()}, although you may find doing so to be useful
220 while debugging. Instead, use semaphores, locks and condition
221 variables to solve synchronization problems. Hint: read the comments
222 in @file{threads/synch.h} if you're unsure what synchronization
223 primitives may be used in what situations.
225 Given some designs of some problems, there may be one or two instances
226 in which it is appropriate to directly change the interrupt levels
227 instead of relying on the given synchroniztion primitives. This must
228 be justified in your @file{DESIGNDOC} file. If you're not sure you're
231 While all parts of this assignment are required if you intend to earn
232 full credit on this project, keep in mind that Problem 1-2 (Join) will
233 be needed for future assignments, so you'll want to get this one
234 right. We don't give out solutions, so you're stuck with your Join
235 code for the whole quarter. Problem 1-1 (Alarm Clock) could be very
236 handy, but not strictly required in the future. The upshot of all
237 this is that you should focus heavily on making sure that your
238 implementation of @code{thread_join()} works correctly, since if it's
239 broken, you will need to fix it for future assignments. The other
240 parts can be turned off in the future if you find you can't make them
243 Also keep in mind that Problem 1-4 (the MLFQS) builds on the features you
244 implement in Problem 1-3, so to avoid unnecessary code duplication, it
245 would be a good idea to divide up the work among your team members
246 such that you have Problem 1-3 fully working before you begin to tackle
249 @node Problem 1-1 Alarm Clock
250 @section Problem 1-1: Alarm Clock
252 Improve the implementation of the timer device defined in
253 @file{devices/timer.c} by reimplementing @code{timer_sleep()}.
254 Threads call @code{timer_sleep(@var{x})} to suspend execution until
255 time has advanced by at least @w{@var{x} timer ticks}. This is
256 useful for threads that operate in real-time, for example, for
257 blinking the cursor once per second. There is no requirement that
258 threads start running immediately after waking up; just put them on
259 the ready queue after they have waited for approximately the right
262 A working implementation of this function is provided. However, the
263 version provided is poor, because it ``busy waits,'' that is, it spins
264 in a tight loop checking the current time until the current time has
265 advanced far enough. This is undesirable because it wastes time that
266 could potentially be used more profitably by another thread. Your
267 solution should not busy wait.
269 The argument to @code{timer_sleep()} is expressed in timer ticks, not
270 in milliseconds or some other unit.
272 @node Problem 1-2 Join
273 @section Problem 1-2: Join
275 Implement @code{thread_join(tid_t)} in @file{threads/thread.c}. There
276 is already a prototype for it in @file{threads/thread.h}, which you
277 should not change. This function causes the currently running thread
278 to block until the thread whose thread id is passed as an argument
279 exits. If A is the running thread and B is the argument, then we say
280 that ``A joins B'' in this case.
282 Incidentally, we don't use @code{struct thread *} as
283 @file{thread_join()}'s parameter type because a thread pointer is not
284 unique over time. That is, when a thread dies, its memory may be,
285 whether immediately or much later, reused for another thread. If
286 thread A over time had two children B and C that were stored at the
287 same address, then @code{thread_join(@r{B})} and
288 @code{thread_join(@r{C})} would be ambiguous. Introducing a thread id
289 or @dfn{tid}, represented by type @code{tid_t}, that is intentionally
290 unique over time solves the problem. The provided code uses an
291 @code{int} for @code{tid_t}, but you may decide you prefer to use some
294 The model for @code{thread_join()} is the @command{wait} system call
295 in Unix-like systems. (Try reading the manpages.) That system call
296 can only be used by a parent process to wait for a child's death. You
297 should implement @code{thread_join()} to have the same restriction.
298 That is, a thread may only join its immediate children.
300 A thread need not ever be joined. Your solution should properly free
301 all of a thread's resources, including its @code{struct thread},
302 whether it is ever joined or not, and regardless of whether the child
303 exits before or after its parent. That is, a thread should be freed
304 exactly once in all cases.
306 Joining a given thread is idempotent. That is, joining a thread T
307 multiple times is equivalent to joining it once, because T has already
308 exited at the time of the later joins. Thus, joins on T after the
309 first should return immediately.
311 Calling @code{thread_join()} on an thread that is not the caller's
312 child should cause the caller to return immediately.
314 Consider all the ways a join can occur: nested joins (A joins B when B
315 is joined on C), multiple joins (A joins B, then A joins C), and so
316 on. Does your join work if @code{thread_join()} is called on a thread
317 that has not yet been scheduled for the first time? You should handle
318 all of these cases. Write test code that demonstrates the cases your
319 join works for. Don't overdo the output volume, please!
321 Be careful to program this function correctly. You will need its
322 functionality for project 2.
324 @node Problem 1-3 Priority Scheduling
325 @section Problem 1-3: Priority Scheduling
327 Implement priority scheduling in Pintos. Priority scheduling is a key
328 building block for real-time systems. Implement functions
329 @code{thread_set_priority()} to set the priority of the running thread
330 and @code{thread_get_priority()} to get the running thread's priority.
331 (A thread can examine and modify only its own priority.) There are
332 already prototypes for these functions in @file{threads/thread.h},
333 which you should not change.
335 Thread priority ranges from @code{PRI_MIN} (0) to @code{PRI_MAX} (59).
336 The initial thread priority is passed as an argument to
337 @code{thread_create()}. If there's no reason to choose another
338 priority, use @code{PRI_DEFAULT} (29). The @code{PRI_} macros are
339 defined in @file{threads/thread.h}, and you should not change their
342 When a thread is added to the ready list that has a higher priority
343 than the currently running thread, the current thread should
344 immediately yield the processor to the new thread. Similarly, when
345 threads are waiting for a lock, semaphore or condition variable, the
346 highest priority waiting thread should be woken up first. A thread's
347 priority may be set at any time, including while the thread is waiting
348 on a lock, semaphore, or condition variable.
350 One issue with priority scheduling is ``priority inversion'': if a
351 high priority thread needs to wait for a low priority thread (for
352 instance, for a lock held by a low priority thread, or in
353 @code{thread_join()} for a thread to complete), and a middle priority
354 thread is on the ready list, then the high priority thread will never
355 get the CPU because the low priority thread will not get any CPU time.
356 A partial fix for this problem is to have the waiting thread
357 ``donate'' its priority to the low priority thread while it is holding
358 the lock, then recall the donation once it has acquired the lock.
361 You will need to account for all different orders that priority
362 donation and inversion can occur. Be sure to handle multiple
363 donations, in which multiple priorities are donated to a thread. You
364 must also handle nested donation: given high, medium, and low priority
365 threads H, M, and L, respectively, if H is waiting on a lock that M
366 holds and M is waiting on a lock that L holds, then both M and L
367 should be boosted to H's priority.
369 You only need to implement priority donation when a thread is waiting
370 for a lock held by a lower-priority thread. You do not need to
371 implement this fix for semaphores, condition variables or joins.
372 However, you do need to implement priority scheduling in all cases.
374 @node Problem 1-4 Advanced Scheduler
375 @section Problem 1-4: Advanced Scheduler
377 Implement Solaris's multilevel feedback queue scheduler (MLFQS) to
378 reduce the average response time for running jobs on your system.
379 @xref{Multilevel Feedback Scheduling}, for a detailed description of
380 the MLFQS requirements.
382 Demonstrate that your scheduling algorithm reduces response time
383 relative to the original Pintos scheduling algorithm (round robin) for
384 at least one workload of your own design (i.e.@: in addition to the
387 You may assume a static priority for this problem. It is not necessary
388 to ``re-donate'' a thread's priority if it changes (although you are
391 You must write your code so that we can turn the MLFQS on and off at
392 compile time. By default, it must be off, but we must be able to turn
393 it on by inserting the line @code{#define MLFQS 1} in
394 @file{constants.h}. @xref{Conditional Compilation}, for details.
404 @b{I am adding a new @file{.h} or @file{.c} file. How do I fix the
405 @file{Makefile}s?}@anchor{Adding c or h Files}
407 To add a @file{.c} file, edit the top-level @file{Makefile.build}.
408 You'll want to add your file to variable @samp{@var{dir}_SRC}, where
409 @var{dir} is the directory where you added the file. For this
410 project, that means you should add it to @code{threads_SRC}, or
411 possibly @code{devices_SRC} if you put in the @file{devices}
412 directory. Then run @code{make}. If your new file doesn't get
413 compiled, run @code{make clean} and then try again.
415 When you modify the top-level @file{Makefile.build}, the modified
416 version should be automatically copied to
417 @file{threads/build/Makefile} when you re-run make. The opposite is
418 not true, so any changes will be lost the next time you run @code{make
419 clean} from the @file{threads} directory. Therefore, you should
420 prefer to edit @file{Makefile.build} (unless your changes are meant to
423 There is no need to edit the @file{Makefile}s to add a @file{.h} file.
426 @b{How do I write my test cases?}
428 Test cases should be replacements for the existing @file{test.c}
429 file. Put them in a @file{threads/testcases} directory.
430 @xref{TESTCASE}, for more information.
433 @b{If a thread finishes, should its children be terminated immediately,
434 or should they finish normally?}
436 You should feel free to decide what semantics you think this
437 should have. You need only provide justification for your
441 @b{Why can't I disable interrupts?}
443 Turning off interrupts should only be done for short amounts of time,
444 or else you end up losing important things such as disk or input
445 events. Turning off interrupts also increases the interrupt handling
446 latency, which can make a machine feel sluggish if taken too far.
447 Therefore, in general, setting the interrupt level should be used
448 sparingly. Also, any synchronization problem can be easily solved by
449 turning interrupts off, since while interrupts are off, there is no
450 concurrency, so there's no possibility for race condition.
452 To make sure you understand concurrency well, we are discouraging you
453 from taking this shortcut at all in your solution. If you are unable
454 to solve a particular synchronization problem with semaphores, locks,
455 or conditions, or think that they are inadequate for a particular
456 reason, you may turn to disabling interrupts. If you want to do this,
457 we require in your design document a complete justification and
458 scenario (i.e.@: exact sequence of events) to show why interrupt
459 manipulation is the best solution. If you are unsure, the TAs can
460 help you determine if you are using interrupts too haphazardly. We
461 want to emphasize that there are only limited cases where this is
464 You might find @file{devices/intq.h} and its users to be an
465 inspiration or source of rationale.
468 @b{Where might interrupt-level manipulation be appropriate?}
470 You might find it necessary in some solutions to the Alarm problem.
472 You might want it at one small point for the priority scheduling
473 problem. Note that it is not required to use interrupts for these
474 problems. There are other, equally correct solutions that do not
475 require interrupt manipulation. However, if you do manipulate
476 interrupts and @strong{correctly and fully document it} in your design
477 document, we will allow limited use of interrupt disabling.
480 @item Alarm Clock FAQs
484 @b{Why can't I use most synchronization primitives in an interrupt
487 As you've discovered, you cannot sleep in an external interrupt
488 handler. Since many lock, semaphore, and condition variable functions
489 attempt to sleep, you won't be able to call those in
490 @code{timer_interrupt()}. You may still use those that never sleep.
492 Having said that, you need to make sure that global data does not get
493 updated by multiple threads simultaneously executing
494 @code{timer_sleep()}. Here are some pieces of information to think
499 Interrupts are turned off while @code{timer_interrupt()} runs. This
500 means that @code{timer_interrupt()} will not be interrupted by a
501 thread running in @code{timer_sleep()}.
504 A thread in @code{timer_sleep()}, however, can be interrupted by a
505 call to @code{timer_interrupt()}, except when that thread has turned
509 Examples of synchronization mechanisms have been presented in lecture.
510 Going over these examples should help you understand when each type is
515 @b{What about timer overflow due to the fact that times are defined as
516 integers? Do I need to check for that?}
518 Don't worry about the possibility of timer values overflowing. Timer
519 values are expressed as signed 63-bit numbers, which at 100 ticks per
520 second should be good for almost 2,924,712,087 years.
527 @b{Am I correct to assume that once a thread is deleted, it is no
528 longer accessible by the parent (i.e.@: the parent can't call
529 @code{thread_join(child)})?}
531 A parent joining a child that has completed should be handled
532 gracefully and should act as a no-op.
535 @item Priority Scheduling FAQs
539 @b{Doesn't the priority scheduling lead to starvation? Or do I have to
540 implement some sort of aging?}
542 It is true that strict priority scheduling can lead to starvation
543 because thread may not run if a higher-priority thread is runnable.
544 In this problem, don't worry about starvation or any sort of aging
545 technique. Problem 4 will introduce a mechanism for dynamically
546 changing thread priorities.
548 This sort of scheduling is valuable in real-time systems because it
549 offers the programmer more control over which jobs get processing
550 time. High priorities are generally reserved for time-critical
551 tasks. It's not ``fair,'' but it addresses other concerns not
552 applicable to a general-purpose operating system.
555 @b{After a lock has been released, does the program need to switch to
556 the highest priority thread that needs the lock (assuming that its
557 priority is higher than that of the current thread)?}
559 When a lock is released, the highest priority thread waiting for that
560 lock should be unblocked and put on the ready to run list. The
561 scheduler should then run the highest priority thread on the ready
565 @b{If a thread calls @code{thread_yield()} and then it turns out that
566 it has higher priority than any other threads, does the high-priority
567 thread continue running?}
569 Yes. If there is a single highest-priority thread, it continues
570 running until it blocks or finishes, even if it calls
571 @code{thread_yield()}.
574 @b{If the highest priority thread is added to the ready to run list it
575 should start execution immediately. Is it immediate enough if I
576 wait until next timer interrupt occurs?}
578 The highest priority thread should run as soon as it is runnable,
579 preempting whatever thread is currently running.
582 @b{What happens to the priority of the donating thread? Do the priorities
585 No. Priority donation only changes the priority of the low-priority
586 thread. The donating thread's priority stays unchanged. Also note
587 that priorities aren't additive: if thread A (with priority 5) donates
588 to thread B (with priority 3), then B's new priority is 5, not 8.
591 @b{Can a thread's priority be changed while it is sitting on the ready
594 Yes. Consider this case: low-priority thread L currently has a lock
595 that high-priority thread H wants. H donates its priority to L (the
596 lock holder). L finishes with the lock, and then loses the CPU and is
597 moved to the ready queue. Now L's old priority is restored while it
598 is in the ready queue.
601 @b{Can a thread's priority change while it is sitting on the queue of a
604 Yes. Same scenario as above except L gets blocked waiting on a new
605 lock when H restores its priority.
608 @b{Why is pubtest3's FIFO test skipping some threads! I know my scheduler
609 is round-robin'ing them like it's supposed to! Our output is like this:}
623 @noindent @b{which repeats 5 times and then}
633 This happens because context switches are being invoked by the test
634 when it explicitly calls @code{thread_yield()}. However, the time
635 slice timer is still alive and so, every tick (by default), thread 1
636 gets switched out (caused by @code{timer_interrupt()} calling
637 @code{intr_yield_on_return()}) before it gets a chance to run its
638 mainline. It is by coincidence that Thread 1 is the one that gets
639 skipped in our example. If we use a different jitter value, the same
640 behavior is seen where a thread gets started and switched out
643 Solution: Increase the value of @code{TIME_SLICE} in
644 @file{devices/timer.c} to a very high value, such as 10000, to see
645 that the threads will round-robin if they aren't interrupted.
648 @b{What happens when a thread is added to the ready list which has
649 higher priority than the currently running thread?}
651 The correct behavior is to immediately yield the processor. Your
652 solution must act this way.
655 @b{What should @code{thread_get_priority()} return in a thread while
656 its priority has been increased by a donation?}
658 The higher (donated) priority.
661 @item Advanced Scheduler FAQs
665 @b{What is the interval between timer interrupts?}
667 Timer interrupts occur @code{TIMER_FREQ} times per second. You can
668 adjust this value by editing @file{devices/timer.h}. The default is
671 You can also adjust the number of timer ticks per time slice by
672 modifying @code{TIME_SLICE} in @file{devices/timer.c}.
675 @b{Do I have to modify the dispatch table?}
677 No, although you are allowed to. It is possible to complete
678 this problem (i.e.@: demonstrate response time improvement)
682 @b{When the scheduler changes the priority of a thread, how does this
683 affect priority donation?}
685 Short (official) answer: Don't worry about it. Your priority donation
686 code may assume static priority assignment.
688 Longer (unofficial) opinion: If you wish to take this into account,
689 however, your design may end up being ``cleaner.'' You have
690 considerable freedom in what actually takes place. I believe what
691 makes the most sense is for scheduler changes to affect the
692 ``original'' (non-donated) priority. This change may actually be
693 masked by the donated priority. Priority changes should only
694 propagate with donations, not ``backwards'' from donees to donors.
697 @b{What is meant by ``static priority''?}
699 Once thread A has donated its priority to thread B, if thread A's
700 priority changes (due to the scheduler) while the donation still
701 exists, you do not have to change thread B's donated priority.
702 However, you are free to do so.
705 @b{Do I have to make my dispatch table user-configurable?}
707 No. Hard-coding the dispatch table values is fine.