10 Ben Pfaff <blp@stanford.edu>
12 ---- PRELIMINARIES ----
14 >> If you have any preliminary comments on your submission, notes for
15 >> the TAs, or extra credit, please give them here.
17 (This is a sample design document.)
19 >> Please cite any offline or online sources you consulted while
20 >> preparing your submission, other than the Pintos documentation,
21 >> course text, and lecture notes.
28 ---- DATA STRUCTURES ----
30 >> Copy here the declaration of each new or changed `struct' or `struct'
31 >> member, global or static variable, `typedef', or enumeration.
32 >> Identify the purpose of each in 25 words or less.
34 A "latch" is a new synchronization primitive. Acquires block
35 until the first release. Afterward, all ongoing and future
36 acquires pass immediately.
41 bool released; /* Released yet? */
42 struct lock monitor_lock; /* Monitor lock. */
43 struct condition rel_cond; /* Signaled when released. */
46 Added to struct thread:
48 /* Members for implementing thread_join(). */
49 struct latch ready_to_die; /* Release when thread about to die. */
50 struct semaphore can_die; /* Up when thread allowed to die. */
51 struct list children; /* List of child threads. */
52 list_elem children_elem; /* Element of `children' list. */
56 >> Briefly describe your implementation of thread_join() and how it
57 >> interacts with thread termination.
59 thread_join() finds the joined child on the thread's list of
60 children and waits for the child to exit by acquiring the child's
61 ready_to_die latch. When thread_exit() is called, the thread
62 releases its ready_to_die latch, allowing the parent to continue.
64 ---- SYNCHRONIZATION ----
66 >> Consider parent thread P with child thread C. How do you ensure
67 >> proper synchronization and avoid race conditions when P calls wait(C)
68 >> before C exits? After C exits? How do you ensure that all resources
69 >> are freed in each case? How about when P terminates without waiting,
70 >> before C exits? After C exits? Are there any special cases?
72 C waits in thread_exit() for P to die before it finishes its own
73 exit, using the can_die semaphore "down"ed by C and "up"ed by P as
74 it exits. Regardless of whether whether C has terminated, there
75 is no race on wait(C), because C waits for P's permission before
78 Regardless of whether P waits for C, P still "up"s C's can_die
79 semaphore when P dies, so C will always be freed. (However,
80 freeing C's resources is delayed until P's death.)
82 The initial thread is a special case because it has no parent to
83 wait for it or to "up" its can_die semaphore. Therefore, its
84 can_die semaphore is initialized to 1.
88 >> Critique your design, pointing out advantages and disadvantages in
89 >> your design choices.
91 This design has the advantage of simplicity. Encapsulating most
92 of the synchronization logic into a new "latch" structure
93 abstracts what little complexity there is into a separate layer,
94 making the design easier to reason about. Also, all the new data
95 members are in `struct thread', with no need for any extra dynamic
96 allocation, etc., that would require extra management code.
98 On the other hand, this design is wasteful in that a child thread
99 cannot free itself before its parent has terminated. A parent
100 thread that creates a large number of short-lived child threads
101 could unnecessarily exhaust kernel memory. This is probably
102 acceptable for implementing kernel threads, but it may be a bad
103 idea for use with user processes because of the larger number of
104 resources that user processes tend to own.