1 @node Project 1--Threads, Project 2--User Programs, Pintos Tour, 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.
14 Before you read the description of this project, you should read all
15 of the following sections: @ref{Introduction}, @ref{Coding Standards},
16 @ref{Project Documentation}, @ref{Debugging Tools}, and
17 @ref{Development Tools}. You should at least skim the material in
18 @ref{Threads Tour}. To complete this project you will also need to
19 read @ref{Multilevel Feedback Scheduling}.
22 * Understanding Threads::
24 * Debugging versus Testing::
26 * Problem 1-1 Alarm Clock::
28 * Problem 1-3 Priority Scheduling::
29 * Problem 1-4 Advanced Scheduler::
33 @node Understanding Threads
34 @section Understanding Threads
36 The first step is to read and understand the initial thread system.
37 Pintos, by default, implements thread creation and thread completion,
38 a simple scheduler to switch between threads, and synchronization
39 primitives (semaphores, locks, and condition variables).
41 However, there's a lot of magic going on in some of this code, so if
42 you haven't already compiled and run the base system, as described in
43 the introduction (@pxref{Introduction}), you should do so now. You
44 can read through parts of the source code by hand to see what's going
45 on. If you like, you can add calls to @func{printf} almost
46 anywhere, then recompile and run to see what happens and in what
47 order. You can also run the kernel in a debugger and set breakpoints
48 at interesting spots, single-step through code and examine data, and
49 so on. @xref{i386-elf-gdb}, for more information.
51 When a thread is created, you are creating a new context to be
52 scheduled. You provide a function to be run in this context as an
53 argument to @func{thread_create}. The first time the thread is
54 scheduled and runs, it will start from the beginning of that function
55 and execute it in the context. When that function returns, that thread
56 completes. Each thread, therefore, acts like a mini-program running
57 inside Pintos, with the function passed to @func{thread_create}
58 acting like @func{main}.
60 At any given time, Pintos is running exactly one thread, with the
61 others switched out. The scheduler decides which thread to run next
62 when it needs to switch between them. (If no thread is ready to run
63 at any given time, then the special ``idle'' thread runs.) The
64 synchronization primitives are used to force context switches when one
65 thread needs to wait for another thread to do something.
67 The exact mechanics of a context switch are pretty gruesome and have
68 been provided for you in @file{threads/switch.S} (this is 80@var{x}86
69 assembly; don't worry about understanding it). It involves saving the
70 state of the currently running thread and restoring the state of the
71 thread we're switching to.
73 Using the @command{gdb} debugger, slowly trace through a context
74 switch to see what happens (@pxref{i386-elf-gdb}). You can set a
75 breakpoint on the @func{schedule} function to start out, and then
76 single-step from there.@footnote{@command{gdb} might tell you that
77 @func{schedule} doesn't exist, which is arguably a @command{gdb} bug.
78 You can work around this by setting the breakpoint by filename and
79 line number, e.g.@: @code{break thread.c:@var{ln}} where @var{ln} is
80 the line number of the first declaration in @func{schedule}.} Be sure
81 to keep track of each thread's address
82 and state, and what procedures are on the call stack for each thread.
83 You will notice that when one thread calls @func{switch_threads},
84 another thread starts running, and the first thing the new thread does
85 is to return from @func{switch_threads}. We realize this comment will
86 seem cryptic to you at this point, but you will understand threads
87 once you understand why the @func{switch_threads} that gets called is
88 different from the @func{switch_threads} that returns.
90 @strong{Warning}: In Pintos, each thread is assigned a small,
91 fixed-size execution stack just under @w{4 kB} in size. The kernel
92 does try to detect stack overflow, but it cannot always succeed. You
93 may cause bizarre problems, such as mysterious kernel panics, if you
94 declare large data structures as non-static local variables,
95 e.g. @samp{int buf[1000];}. Alternatives to stack allocation include
96 the page allocator in @file{threads/palloc.c} and the block allocator
97 in @file{threads/malloc.c}. Note that the page allocator doles out
98 @w{4 kB} chunks and that @func{malloc} has a @w{2 kB} block size
99 limit. If you need larger chunks, consider using a linked structure
105 Here is a brief overview of the files in the @file{threads}
106 directory. You will not need to modify most of this code, but the
107 hope is that presenting this overview will give you a start on what
113 The kernel loader. Assembles to 512 bytes of code and data that the
114 PC BIOS loads into memory and which in turn loads the kernel into
115 memory, does basic processor initialization, and jumps to the
116 beginning of the kernel. You should not need to look at this code or
120 The linker script used to link the kernel. Sets the load address of
121 the kernel and arranges for @file{start.S} to be at the very beginning
122 of the kernel image. Again, you should not need to look at this code
123 or modify it, but it's here in case you're curious.
126 Jumps to @func{main}.
130 Kernel initialization, including @func{main}, the kernel's ``main
131 program.'' You should look over @func{main} at least to see what
136 Basic thread support. Much of your work will take place in these
137 files. @file{thread.h} defines @struct{thread}, which you will
138 modify in the first three projects.
142 Assembly language routine for switching threads. Already discussed
147 Page allocator, which hands out system memory in multiples of 4 kB
152 A very simple implementation of @func{malloc} and @func{free} for
157 Basic interrupt handling and functions for turning interrupts on and
162 A Perl program that outputs assembly for low-level interrupt handling.
166 Basic synchronization primitives: semaphores, locks, and condition
167 variables. You will need to use these for synchronization through all
172 Test code. For project 1, you will replace this file with your test
176 Functions for I/O port access. This is mostly used by source code in
177 the @file{devices} directory that you won't have to touch.
180 Functions and macros related to memory management, including page
181 directories and page tables. This will be more important to you in
182 project 3. For now, you can ignore it.
191 @subsection @file{devices} code
193 The basic threaded kernel also includes these files in the
194 @file{devices} directory:
199 System timer that ticks, by default, 100 times per second. You will
200 modify this code in Problem 1-1.
204 VGA display driver. Responsible for writing text to the screen.
205 You should have no need to look at this code. @func{printf} will
206 call into the VGA display driver for you, so there's little reason to
207 call this code yourself.
211 Serial port driver. Again, @func{printf} calls this code for you,
212 so you don't need to do so yourself. Feel free to look through it if
217 Supports reading and writing sectors on up to 4 IDE disks. This won't
218 actually be used until project 2.
222 Interrupt queue, for managing a circular queue that both kernel
223 threads and interrupt handlers want to access. Used by the keyboard
228 @subsection @file{lib} files
230 Finally, @file{lib} and @file{lib/kernel} contain useful library
231 routines. (@file{lib/user} will be used by user programs, starting in
232 project 2, but it is not part of the kernel.) Here's a few more
249 Implementation of the standard C library. @xref{C99}, for information
250 on a few recently introduced pieces of the C library that you might
251 not have encountered before. @xref{Unsafe String Functions}, for
252 information on what's been intentionally left out for safety.
256 Functions and macros to aid debugging. @xref{Debugging Tools}, for
261 Pseudo-random number generator.
267 System call numbers. Not used until project 2.
271 Doubly linked list implementation. Used all over the Pintos code, and
272 you'll probably want to use it a few places yourself in project 1.
274 @item kernel/bitmap.c
275 @itemx kernel/bitmap.h
276 Bitmap implementation. You can use this in your code if you like, but
277 you probably won't have any need for project 1.
281 Hash table implementation. Likely to come in handy for project 3.
283 @item kernel/console.c
284 @itemx kernel/console.h
285 Implements @func{printf} and a few other functions.
288 @node Debugging versus Testing
289 @section Debugging versus Testing
291 When you're debugging code, it's useful to be able to be able to run a
292 program twice and have it do exactly the same thing. On second and
293 later runs, you can make new observations without having to discard or
294 verify your old observations. This property is called
295 ``reproducibility.'' The simulator we use, Bochs, can be set up for
296 reproducibility, and that's the way that @command{pintos} invokes it
299 Of course, a simulation can only be reproducible from one run to the
300 next if its input is the same each time. For simulating an entire
301 computer, as we do, this means that every part of the computer must be
302 the same. For example, you must use the same disks, the same version
303 of Bochs, and you must not hit any keys on the keyboard (because you
304 could not be sure to hit them at exactly the same point each time)
307 While reproducibility is useful for debugging, it is a problem for
308 testing thread synchronization, an important part of this project. In
309 particular, when Bochs is set up for reproducibility, timer interrupts
310 will come at perfectly reproducible points, and therefore so will
311 thread switches. That means that running the same test several times
312 doesn't give you any greater confidence in your code's correctness
313 than does running it only once.
315 So, to make your code easier to test, we've added a feature, called
316 ``jitter,'' to Bochs, that makes timer interrupts come at random
317 intervals, but in a perfectly predictable way. In particular, if you
318 invoke @command{pintos} with the option @option{-j @var{seed}}, timer
319 interrupts will come at irregularly spaced intervals. Within a single
320 @var{seed} value, execution will still be reproducible, but timer
321 behavior will change as @var{seed} is varied. Thus, for the highest
322 degree of confidence you should test your code with many seed values.
324 On the other hand, when Bochs runs in reproducible mode, timings are not
325 realistic, meaning that a ``one-second'' delay may be much shorter or
326 even much longer than one second. You can invoke @command{pintos} with
327 a different option, @option{-r}, to make it set up Bochs for realistic
328 timings, in which a one-second delay should take approximately one
329 second of real time. Simulation in real-time mode is not reproducible,
330 and options @option{-j} and @option{-r} are mutually exclusive.
337 There should be no busy-waiting in any of your solutions to this
341 Do your best to resist the temptation to directly disable interrupts
342 in your solution by calling @func{intr_disable} or
343 @func{intr_set_level}, although you may find doing so to be useful
344 while debugging. Instead, use semaphores, locks and condition
345 variables to solve synchronization problems. Read the tour section on
346 synchronization (@pxref{Synchronization}) or the comments in
347 @file{threads/synch.h} if you're unsure what synchronization
348 primitives may be used in what situations.
350 Given some designs of some problems, there may be one or two instances
351 in which it is appropriate to directly change the interrupt levels
352 instead of relying on the given synchronization primitives. This must
353 be justified in your @file{DESIGNDOC} file. If you're not sure you're
357 All parts of this assignment are required if you intend to earn full
358 credit on this project. However, some will be more important in
363 Problem 1-1 (Alarm Clock) could be handy for later projects, but it is
364 not strictly required.
367 Problem 1-2 (Join) will be needed for future projects. We don't give
368 out solutions, so to avoid extra work later you should make sure that
369 your implementation of @func{thread_join} works correctly.
372 Problems 1-3 and 1-4 won't be needed for later projects.
376 Problem 1-4 (MLFQS) builds on the features you
377 implement in Problem 1-3. To avoid unnecessary code duplication, it
378 would be a good idea to divide up the work among your team members
379 such that you have Problem 1-3 fully working before you begin to tackle
383 @node Problem 1-1 Alarm Clock
384 @section Problem 1-1: Alarm Clock
386 Improve the implementation of the timer device defined in
387 @file{devices/timer.c} by reimplementing @func{timer_sleep}.
388 Threads call @code{timer_sleep(@var{x})} to suspend execution until
389 time has advanced by at least @w{@var{x} timer ticks}. This is
390 useful for threads that operate in real-time, for example, for
391 blinking the cursor once per second. There is no requirement that
392 threads start running immediately after waking up; just put them on
393 the ready queue after they have waited for approximately the right
396 A working implementation of this function is provided. However, the
397 version provided is poor, because it ``busy waits,'' that is, it spins
398 in a tight loop checking the current time until the current time has
399 advanced far enough. This is undesirable because it wastes time that
400 could potentially be used more profitably by another thread. Your
401 solution should not busy wait.
403 The argument to @func{timer_sleep} is expressed in timer ticks, not in
404 milliseconds or any another unit. There are @code{TIMER_FREQ} timer
405 ticks per second, where @code{TIMER_FREQ} is a macro defined in
406 @code{devices/timer.h}.
408 Separate functions @func{timer_msleep}, @func{timer_usleep}, and
409 @func{timer_nsleep} do exist for sleeping a specific number of
410 milliseconds, microseconds, or nanoseconds, respectively, but these will
411 call @func{timer_sleep} automatically when necessary. You do not need
414 If your delays seem too short or too long, reread the explanation of the
415 @option{-r} option to @command{pintos} (@pxref{Debugging versus
418 @node Problem 1-2 Join
419 @section Problem 1-2: Join
421 Implement @code{thread_join(tid_t)} in @file{threads/thread.c}. There
422 is already a prototype for it in @file{threads/thread.h}, which you
423 should not change. This function causes the currently running thread
424 to block until the thread whose thread id is passed as an argument
425 exits. If @var{A} is the running thread and @var{B} is the argument,
426 then we say that ``@var{A} joins @var{B}.''
428 Incidentally, we don't use @code{struct thread *} as
429 @func{thread_join}'s parameter type because a thread pointer is not
430 unique over time. That is, when a thread dies, its memory may be,
431 whether immediately or much later, reused for another thread. If
432 thread A over time had two children B and C that were stored at the
433 same address, then @code{thread_join(@var{B})} and
434 @code{thread_join(@var{C})} would be ambiguous. Introducing a thread
435 id or @dfn{tid}, represented by type @code{tid_t}, that is
436 intentionally unique over time solves the problem. The provided code
437 uses an @code{int} for @code{tid_t}, but you may decide you prefer to
440 The model for @func{thread_join} is the @command{wait} system call
441 in Unix-like systems. (Try reading the manpages.) That system call
442 can only be used by a parent process to wait for a child's death. You
443 should implement @func{thread_join} to have the same restriction.
444 That is, a thread may only join its immediate children.
446 A thread need not ever be joined. Your solution should properly free
447 all of a thread's resources, including its @struct{thread},
448 whether it is ever joined or not, and regardless of whether the child
449 exits before or after its parent. That is, a thread should be freed
450 exactly once in all cases.
452 Joining a given thread is idempotent. That is, joining a thread T
453 multiple times is equivalent to joining it once, because T has already
454 exited at the time of the later joins. Thus, joins on T after the
455 first should return immediately.
457 Calling @func{thread_join} on an thread that is not the caller's
458 child should cause the caller to return immediately.
460 Consider all the ways a join can occur: nested joins (@var{A} joins
461 @var{B}, then @var{B} joins @var{C}), multiple joins (@var{A} joins
462 @var{B}, then @var{A} joins @var{C}), and so on. Does your join work
463 if @func{thread_join} is called on a thread that has not yet been
464 scheduled for the first time? You should handle all of these cases.
465 Write test code that demonstrates the cases your join works for.
466 Don't overdo the output volume, please!
468 Be careful to program this function correctly. You will need its
469 functionality for project 2.
471 Once you've implemented @func{thread_join}, define
472 @code{THREAD_JOIN_IMPLEMENTED} in @file{constants.h}.
473 @xref{Conditional Compilation}, for more information.
475 @node Problem 1-3 Priority Scheduling
476 @section Problem 1-3: Priority Scheduling
478 Implement priority scheduling in Pintos. Priority scheduling is a key
479 building block for real-time systems. Implement functions
480 @func{thread_set_priority} to set the priority of the running thread
481 and @func{thread_get_priority} to get the running thread's priority.
482 (This API only allows a thread to examine and modify its own
483 priority.) There are already prototypes for these functions in
484 @file{threads/thread.h}, which you should not change.
486 Thread priority ranges from @code{PRI_MIN} (0) to @code{PRI_MAX} (59).
487 The initial thread priority is passed as an argument to
488 @func{thread_create}. If there's no reason to choose another
489 priority, use @code{PRI_DEFAULT} (29). The @code{PRI_} macros are
490 defined in @file{threads/thread.h}, and you should not change their
493 When a thread is added to the ready list that has a higher priority
494 than the currently running thread, the current thread should
495 immediately yield the processor to the new thread. Similarly, when
496 threads are waiting for a lock, semaphore or condition variable, the
497 highest priority waiting thread should be woken up first. A thread
498 may set its priority at any time.
500 One issue with priority scheduling is ``priority inversion'': if a
501 high priority thread needs to wait for a low priority thread (for
502 instance, for a lock held by a low priority thread, or in
503 @func{thread_join} for a thread to complete), and a middle priority
504 thread is on the ready list, then the high priority thread will never
505 get the CPU because the low priority thread will not get any CPU time.
506 A partial fix for this problem is to have the waiting thread
507 ``donate'' its priority to the low priority thread while it is holding
508 the lock, then recall the donation once it has acquired the lock.
511 You will need to account for all different orders in which priority
512 donation and inversion can occur. Be sure to handle multiple
513 donations, in which multiple priorities are donated to a thread. You
514 must also handle nested donation: given high, medium, and low priority
515 threads @var{H}, @var{M}, and @var{L}, respectively, if @var{H} is
516 waiting on a lock that @var{M} holds and @var{M} is waiting on a lock
517 that @var{L} holds, then both @var{M} and @var{L} should be boosted to
520 You only need to implement priority donation when a thread is waiting
521 for a lock held by a lower-priority thread. You do not need to
522 implement this fix for semaphores, condition variables, or joins,
523 although you are welcome to do so. However, you do need to implement
524 priority scheduling in all cases.
526 You may assume a static priority for priority donation, that is, it is
527 not necessary to ``re-donate'' a thread's priority if it changes
528 (although you are free to do so).
530 @node Problem 1-4 Advanced Scheduler
531 @section Problem 1-4: Advanced Scheduler
533 Implement Solaris's multilevel feedback queue scheduler (MLFQS) to
534 reduce the average response time for running jobs on your system.
535 @xref{Multilevel Feedback Scheduling}, for a detailed description of
536 the MLFQS requirements.
538 Demonstrate that your scheduling algorithm reduces response time
539 relative to the original Pintos scheduling algorithm (round robin) for
540 at least one workload of your own design (i.e.@: in addition to the
543 You must write your code so that we can turn the MLFQS on and off at
544 compile time. By default, it must be off, but we must be able to turn
545 it on by inserting the line @code{#define MLFQS 1} in
546 @file{constants.h}. @xref{Conditional Compilation}, for details.
553 @b{I am adding a new @file{.h} or @file{.c} file. How do I fix the
554 @file{Makefile}s?}@anchor{Adding c or h Files}
556 To add a @file{.c} file, edit the top-level @file{Makefile.build}.
557 You'll want to add your file to variable @samp{@var{dir}_SRC}, where
558 @var{dir} is the directory where you added the file. For this
559 project, that means you should add it to @code{threads_SRC}, or
560 possibly @code{devices_SRC} if you put in the @file{devices}
561 directory. Then run @code{make}. If your new file doesn't get
562 compiled, run @code{make clean} and then try again.
564 When you modify the top-level @file{Makefile.build}, the modified
565 version should be automatically copied to
566 @file{threads/build/Makefile} when you re-run make. The opposite is
567 not true, so any changes will be lost the next time you run @code{make
568 clean} from the @file{threads} directory. Therefore, you should
569 prefer to edit @file{Makefile.build} (unless your changes are meant to
572 There is no need to edit the @file{Makefile}s to add a @file{.h} file.
575 @b{How do I write my test cases?}
577 Test cases should be replacements for the existing @file{test.c}
578 file. Put them in a @file{threads/testcases} directory.
579 @xref{TESTCASE}, for more information.
582 @b{Why can't I disable interrupts?}
584 Turning off interrupts should only be done for short amounts of time,
585 or else you end up losing important things such as disk or input
586 events. Turning off interrupts also increases the interrupt handling
587 latency, which can make a machine feel sluggish if taken too far.
588 Therefore, in general, setting the interrupt level should be used
589 sparingly. Also, any synchronization problem can be easily solved by
590 turning interrupts off, since while interrupts are off, there is no
591 concurrency, so there's no possibility for race condition.
593 To make sure you understand concurrency well, we are discouraging you
594 from taking this shortcut at all in your solution. If you are unable
595 to solve a particular synchronization problem with semaphores, locks,
596 or conditions, or think that they are inadequate for a particular
597 reason, you may turn to disabling interrupts. If you want to do this,
598 we require in your design document a complete justification and
599 scenario (i.e.@: exact sequence of events) to show why interrupt
600 manipulation is the best solution. If you are unsure, the TAs can
601 help you determine if you are using interrupts too haphazardly. We
602 want to emphasize that there are only limited cases where this is
605 You might find @file{devices/intq.h} and its users to be an
606 inspiration or source of rationale.
609 @b{Where might interrupt-level manipulation be appropriate?}
611 You might find it necessary in some solutions to the Alarm problem.
613 You might want it at one small point for the priority scheduling
614 problem. Note that it is not required to use interrupts for these
615 problems. There are other, equally correct solutions that do not
616 require interrupt manipulation. However, if you do manipulate
617 interrupts and @strong{correctly and fully document it} in your design
618 document, we will allow limited use of interrupt disabling.
621 @b{What does ``warning: no previous prototype for `@var{function}''
624 It means that you defined a non-@code{static} function without
625 preceding it by a prototype. Because non-@code{static} functions are
626 intended for use by other @file{.c} files, for safety they should be
627 prototyped in a header file included before their definition. To fix
628 the problem, add a prototype in a header file that you include, or, if
629 the function isn't actually used by other @file{.c} files, make it
634 * Problem 1-1 Alarm Clock FAQ::
635 * Problem 1-2 Join FAQ::
636 * Problem 1-3 Priority Scheduling FAQ::
637 * Problem 1-4 Advanced Scheduler FAQ::
640 @node Problem 1-1 Alarm Clock FAQ
641 @subsection Problem 1-1: Alarm Clock FAQ
645 @b{Why can't I use most synchronization primitives in an interrupt
648 As you've discovered, you cannot sleep in an external interrupt
649 handler. Since many lock, semaphore, and condition variable functions
650 attempt to sleep, you won't be able to call those in
651 @func{timer_interrupt}. You may still use those that never sleep.
653 Having said that, you need to make sure that global data does not get
654 updated by multiple threads simultaneously executing
655 @func{timer_sleep}. Here are some pieces of information to think
660 Interrupts are turned off while @func{timer_interrupt} runs. This
661 means that @func{timer_interrupt} will not be interrupted by a
662 thread running in @func{timer_sleep}.
665 A thread in @func{timer_sleep}, however, can be interrupted by a
666 call to @func{timer_interrupt}, except when that thread has turned
670 Examples of synchronization mechanisms have been presented in lecture.
671 Going over these examples should help you understand when each type is
672 useful or needed. @xref{Synchronization}, for specific information
673 about synchronization in Pintos.
677 @b{What about timer overflow due to the fact that times are defined as
678 integers? Do I need to check for that?}
680 Don't worry about the possibility of timer values overflowing. Timer
681 values are expressed as signed 63-bit numbers, which at 100 ticks per
682 second should be good for almost 2,924,712,087 years.
685 @b{The test program mostly works but reports a few out-of-order
686 wake ups. I think it's a problem in the test program. What gives?}
687 @anchor{Out of Order 1-1}
689 This test is inherently full of race conditions. On a real system it
690 wouldn't work perfectly all the time either. There are a few ways you
691 can help it work more reliably:
695 Make time slices longer by increasing @code{TIME_SLICE} in
696 @file{timer.c} to a large value, such as 100.
699 Make the timer tick more slowly by decreasing @code{TIMER_FREQ} in
700 @file{timer.h} to its minimum value of 19.
703 The former two changes are only desirable for testing problem 1-1 and
704 possibly 1-3. You should revert them before working on other parts
705 of the project or turn in the project.
708 @b{Should @file{p1-1.c} be expected to work with the MLFQS turned on?}
710 No. The MLFQS will adjust priorities, changing thread ordering.
713 @node Problem 1-2 Join FAQ
714 @subsection Problem 1-2: Join FAQ
718 @b{Am I correct to assume that once a thread is deleted, it is no
719 longer accessible by the parent (i.e.@: the parent can't call
720 @code{thread_join(child)})?}
722 A parent joining a child that has completed should be handled
723 gracefully and should act as a no-op.
726 @node Problem 1-3 Priority Scheduling FAQ
727 @subsection Problem 1-3: Priority Scheduling FAQ
731 @b{Doesn't the priority scheduling lead to starvation? Or do I have to
732 implement some sort of aging?}
734 It is true that strict priority scheduling can lead to starvation
735 because thread may not run if a higher-priority thread is runnable.
736 In this problem, don't worry about starvation or any sort of aging
737 technique. Problem 4 will introduce a mechanism for dynamically
738 changing thread priorities.
740 This sort of scheduling is valuable in real-time systems because it
741 offers the programmer more control over which jobs get processing
742 time. High priorities are generally reserved for time-critical
743 tasks. It's not ``fair,'' but it addresses other concerns not
744 applicable to a general-purpose operating system.
747 @b{After a lock has been released, does the program need to switch to
748 the highest priority thread that needs the lock (assuming that its
749 priority is higher than that of the current thread)?}
751 When a lock is released, the highest priority thread waiting for that
752 lock should be unblocked and put on the ready to run list. The
753 scheduler should then run the highest priority thread on the ready
757 @b{If a thread calls @func{thread_yield} and then it turns out that
758 it has higher priority than any other threads, does the high-priority
759 thread continue running?}
761 Yes. If there is a single highest-priority thread, it continues
762 running until it blocks or finishes, even if it calls
766 @b{If the highest priority thread is added to the ready to run list it
767 should start execution immediately. Is it immediate enough if I
768 wait until next timer interrupt occurs?}
770 The highest priority thread should run as soon as it is runnable,
771 preempting whatever thread is currently running.
774 @b{What happens to the priority of the donating thread? Do the priorities
777 No. Priority donation only changes the priority of the low-priority
778 thread. The donating thread's priority stays unchanged. Also note
779 that priorities aren't additive: if thread A (with priority 5) donates
780 to thread B (with priority 3), then B's new priority is 5, not 8.
783 @b{Can a thread's priority be changed while it is sitting on the ready
786 Yes. Consider this case: low-priority thread L currently has a lock
787 that high-priority thread H wants. H donates its priority to L (the
788 lock holder). L finishes with the lock, and then loses the CPU and is
789 moved to the ready queue. Now L's old priority is restored while it
790 is in the ready queue.
793 @b{Can a thread's priority change while it is sitting on the queue of a
796 Yes. Same scenario as above except L gets blocked waiting on a new
797 lock when H restores its priority.
800 @b{Why is @file{p1-3.c}'s FIFO test skipping some threads? I know my
801 scheduler is round-robin'ing them like it's supposed to. Our output
802 starts out okay, but toward the end it starts getting out of order.}
804 The usual problem is that the serial output buffer fills up. This is
805 causing serial_putc() to block in thread @var{A}, so that thread
806 @var{B} is scheduled. Thread @var{B} immediately tries to do output
807 of its own and blocks on the serial lock (which is held by thread
808 @var{A}). Now that we've wasted some time in scheduling and locking,
809 typically some characters have been drained out of the serial buffer
810 by the interrupt handler, so thread @var{A} can continue its output.
811 After it finishes, though, some other thread (not @var{B}) is
812 scheduled, because thread @var{B} was already scheduled while we
813 waited for the buffer to drain.
815 There's at least one other possibility. Context switches are being
816 invoked by the test when it explicitly calls @func{thread_yield}.
817 However, the time slice timer is still alive and so, every tick (by
818 default), a thread gets switched out (caused by @func{timer_interrupt}
819 calling @func{intr_yield_on_return}) before it gets a chance to run
820 @func{printf}, effectively skipping it. If we use a different jitter
821 value, the same behavior is seen where a thread gets started and
822 switched out completely.
824 Normally you can fix these problems using the same techniques
825 suggested on problem 1-1 (@pxref{Out of Order 1-1}).
828 @b{What happens when a thread is added to the ready list which has
829 higher priority than the currently running thread?}
831 The correct behavior is to immediately yield the processor. Your
832 solution must act this way.
835 @b{What should @func{thread_get_priority} return in a thread while
836 its priority has been increased by a donation?}
838 The higher (donated) priority.
841 @b{Should @file{p1-3.c} be expected to work with the MLFQS turned on?}
843 No. The MLFQS will adjust priorities, changing thread ordering.
846 @b{@func{printf} in @func{sema_up} or @func{sema_down} makes the
849 Yes. These functions are called before @func{printf} is ready to go.
850 You could add a global flag initialized to false and set it to true
851 just before the first @func{printf} in @func{main}. Then modify
852 @func{printf} itself to return immediately if the flag isn't set.
855 @node Problem 1-4 Advanced Scheduler FAQ
856 @subsection Problem 1-4: Advanced Scheduler FAQ
860 @b{What is the interval between timer interrupts?}
862 Timer interrupts occur @code{TIMER_FREQ} times per second. You can
863 adjust this value by editing @file{devices/timer.h}. The default is
866 You can also adjust the number of timer ticks per time slice by
867 modifying @code{TIME_SLICE} in @file{devices/timer.c}.
870 @b{Do I have to modify the dispatch table?}
872 No, although you are allowed to. It is possible to complete
873 this problem (i.e.@: demonstrate response time improvement)
877 @b{When the scheduler changes the priority of a thread, how does this
878 affect priority donation?}
880 Short (official) answer: Don't worry about it. Your priority donation
881 code may assume static priority assignment.
883 Longer (unofficial) opinion: If you wish to take this into account,
884 however, your design may end up being ``cleaner.'' You have
885 considerable freedom in what actually takes place. I believe what
886 makes the most sense is for scheduler changes to affect the
887 ``original'' (non-donated) priority. This change may actually be
888 masked by the donated priority. Priority changes should only
889 propagate with donations, not ``backwards'' from donees to donors.
892 @b{What is meant by ``static priority''?}
894 Once thread A has donated its priority to thread B, if thread A's
895 priority changes (due to the scheduler) while the donation still
896 exists, you do not have to change thread B's donated priority.
897 However, you are free to do so.
900 @b{Do I have to make my dispatch table user-configurable?}
902 No. Hard-coding the dispatch table values is fine.