5 - In Project 2, we're missing tests that pass arguments to system calls
6 that span multiple pages, where some are mapped and some are not.
7 An implementation that only checks the first page, rather than all pages
8 that can be touched during a call to read()/write() passes all tests.
10 - In Project 2, we're missing a test that would fail if they assumed
11 that contiguous user-virtual addresses are laid out contiguously
12 in memory. The loading code should ensure that non-contiguous
13 physical pages are allocated for the data segment (at least.)
15 - Need some tests that test that illegal accesses lead to process
16 termination. I have written some, will add them. In P2, obviously,
17 this would require that the students break this functionality since
18 the page directory is initialized for them, still it would be good
21 - There does not appear to be a test that checks that they close all
22 fd's on exit. Idea: add statistics & self-diagnostics code to palloc.c
23 and malloc.c. Self-diagnostics code could be used for debugging.
24 The statistics code would report how much kernel memory is free.
25 Add a system call "get_kernel_memory_information". User programs
26 could engage in a variety of activities and notice leaks by checking
27 the kernel memory statistics.
29 From: "Godmar Back" <godmar@gmail.com>
31 For reasons I don't currently understand, some of our students seem
32 hesitant to include each thread in a second "all-threads" list and are
33 looking for ways to implement the advanced scheduler without one.
35 Currently, I believe, all tests for the mlfqs are such that all
36 threads are either ready or sleeping in timer_sleep(). This allows for
37 an incorrect implementation in which recent-cpu and priorities are
38 updated only for those threads that are on the alarm list or the ready
41 The todo item would be a test where a thread is blocked on a
42 semaphore, lock or condition variable and have its recent_cpu decay to
43 zero, and check that it's scheduled right after the unlock/up/signal.
45 From: "Godmar Back" <godmar@gmail.com>
46 Subject: set_priority & donation - a TODO item
47 To: "Ben Pfaff" <blp@cs.stanford.edu>
48 Date: Mon, 20 Feb 2006 22:20:26 -0500
52 it seems that there are currently no tests that check the proper
53 behavior of thread_set_priority() when called by a thread that is
54 running under priority donation. The proper behavior, I assume, is to
55 temporarily drop the donation if the set priority is higher, and to
56 reassume the donation should the thread subsequently set its own
57 priority again to a level that's lower than a still active donation.
61 From: Godmar Back <godmar@gmail.com>
62 Subject: on caching in project 4
63 To: Ben Pfaff <blp@cs.stanford.edu>
64 Date: Mon, 9 Jan 2006 20:58:01 -0500
66 here's an idea for future semesters.
68 I'm in the middle of project 4, I've started by implementing a buffer
69 cache and plugging it into the existing filesystem. Along the way I
70 was wondering how we could test the cache.
72 Maybe one could adopt a similar testing strategy as in project 1 for
73 the MLQFS scheduler: add a function that reads "get_cache_accesses()"
74 and a function "get_cache_hits()". Then create a version of pintos
75 that creates access traces for a to-be-determined workload. Run an
76 off-line analysis that would determine how many hits a perfect cache
77 would have (MAX), and how much say an LRU strategy would give (MIN).
78 Then add a fudge factor to account for different index strategies and
79 test that the reported number of cache hits/accesses is within (MIN,
80 MAX) +/- fudge factor.
82 (As an aside - I am curious why you chose to use a clock-style
83 algorithm rather than the more straightforward LRU for your buffer
84 cache implementation in your sample solution. Is there a reason for
85 that? I was curious to see if it made a difference, so I implemented
86 LRU for your cache implementation and ran the test workload of project
87 4 and printed cache hits/accesses.
88 I found that for that workload, the clock-based algorithm performs
89 almost identical to LRU (within about 1%, but I ran nondeterministally
90 with QEMU). I then reduced the cache size to 32 blocks and found again
91 the same performance, which raises the suspicion that the test
92 workload might not force any cache replacement, so the eviction
93 strategy doesn't matter.)
95 Godmar Back <godmar@gmail.com> writes:
97 > in your sample solution to P4, dir_reopen does not take any locks when
98 > changing a directory's open_cnt. This looks like a race condition to
99 > me, considering that dir_reopen is called from execute_process without
100 > any filesystem locks held.
102 * Get rid of rox--causes more trouble than it's worth
104 * Reconsider command line arg style--confuses everyone.
106 * Finish writing tour.
110 * Get rid of mmap syscall, add sbrk.
112 * page-linear, page-shuffle VM tests do not use enough memory to force
113 eviction. Should increase memory consumption.
115 * Add FS persistence test(s).
117 * process_death test needs improvement
121 * Improve automatic interpretation of exception messages.
125 - Mark read-only pages as actually read-only in the page table. Or,
126 since this was consistently rated as the easiest project by the
127 students, require them to do it.
129 - Don't provide per-process pagedir implementation but only
130 single-process implementation and require students to implement
131 the separation? This project was rated as the easiest after all.
132 Alternately we could just remove the synchronization on pid
133 selection and check that students fix it.
137 - Need a better way to measure performance improvement of buffer
138 cache. Some students reported that their system was slower with
139 cache--likely, Bochs doesn't simulate a disk with a realistic
144 - Add "Digging Deeper" sections that describe the nitty-gritty x86
145 details for the benefit of those interested.
147 - Add explanations of what "real" OSes do to give students some
154 . Low-level x86 stuff, like paged page tables.
156 . Specifics on how to implement sbrk, malloc.
160 . opendir/readdir/closedir
162 . everything needed for getcwd()
164 To add partition support:
166 - Find four partition types that are more or less unused and choose to
167 use them for Pintos. (This is implemented.)
169 - Bootloader reads partition tables of all BIOS devices to find the
170 first that has the "Pintos kernel" partition type. (This is
171 implemented.) Ideally the bootloader would make sure there is
172 exactly one such partition, but I didn't implement that yet.
174 - Bootloader reads kernel into memory at 1 MB using BIOS calls. (This
177 - Kernel arguments have to go into a separate sector because the
178 bootloader is otherwise too big to fit now? (I don't recall if I
179 did anything about this.)
181 - Kernel at boot also scans partition tables of all the disks it can
182 find to find the ones with the four Pintos partition types (perhaps
183 not all exist). After that, it makes them available to the rest of
184 the kernel (and doesn't allow access to other devices, for safety).
186 - "pintos" and "pintos-mkdisk" need to write a partition table to the
187 disks that they create. "pintos-mkdisk" will need to take a new
188 parameter specifying the type. (I might have partially implemented
189 this, don't remember.)
191 - "pintos" should insist on finding a partition header on disks handed
194 - Need some way for "pintos" to assemble multiple disks or partitions
195 into a single image that can be copied directly to a USB block
196 device. (I don't know whether I came up with a good solution yet or
197 not, or whether I implemented any of it.)
201 - Needs to be able to scan PCI bus for UHCI controller. (I
202 implemented this partially.)
204 - May want to be able to initialize USB controllers over CardBus
205 bridges. I don't know whether this requires additional work or if
206 it's useful enough to warrant extra work. (It's of special interest
207 for me because I have a laptop that only has USB via CardBus.)
209 - There are many protocol layers involved: SCSI over USB-Mass Storage
210 over USB over UHCI over PCI. (I may be forgetting one.) I don't
211 know yet whether it's best to separate the layers or to merge (some
212 of) them. I think that a simple and clean organization should be a
215 - VMware can likely be used for testing because it can expose host USB
216 devices as guest USB devices. This is safer and more convenient
217 than using real hardware for testing.
219 - Should test with a variety of USB keychain devices because there
220 seems to be wide variation among them, especially in the SCSI
221 protocols they support. Should try to use a "lowest-common
222 denominator" SCSI protocol if any such thing really exists.
224 - Might want to add a feature whereby kernel arguments can be given
225 interactively, rather than passed on-disk. Needs some though.