School of Computer Science
COMP 3000 (WINTER 2022) OPERATING SYSTEMS ASSIGNMENT 2
Please submit the answers to the following questions in Brightspace by the due time indicated in the submission entry. There are 20 points in total (weight: 0.25).
Submit your answers as a gzipped tarball “username-comp3000-assign2.tar.gz” (where username is your MyCarletonOne username). Do NOT just submit a file of another format (e.g., txt or zip) by renaming it
Copyright By PowCoder代写 加微信 powcoder
to .tar.gz. Unlike tutorials, assignments are graded for the correctness of the answers.
The tarball you submit must contain the following:
1. Aplaintextfilecontainingyoursolutionstoallquestions,includingexplanations.
2. AREADME.txtfilelistingthecontentsofyoursubmissionaswellasanyinformationtheTAsshould
know when grading your assignment. Without this file, grading will be based on the TA’s
understanding.
3. Foreachquestion,whereapplicable,aCfileforyourmodifiedversionoftheprovidedsourcecode.
This should include all required changes for that question. This way, you can avoid something
wrong with one question affecting another question.
4. Difffilesshowingthemodifications,bycomparingeachsubmittedCfileaboveandtheoriginal:for
example, diff -c 3000test.c 3000test_modifiedQ1.c > Q1.diff. Avoid moving around or changing existing code (unless necessary) which may be distracting.
You only need to submit code for the following questions:
Part 1: Q1 (3000test) and Part 2: Q7 (3000userlogin.patched).
You can use this command to create the tarball: tar zcvf username-comp3000-assign2.tar.gz your_assignment_directory. **Don’t forget to include your plaintext file!!**
No other formats will be accepted. Submitting in another format will likely result in your assignment not being graded and you receiving no marks for this assignment. In particular, do not submit an MS Word, OpenOffice, or PDF file as your answers document!
Empty or corrupted tarballs may be given a grade of zero, so please double check your submission by downloading and extracting it.
Don’t forget to include what outside resources you used to complete each of your answers, including other students, and web resources. You do not need to list help from the instructor, TA, or information found in the textbook or man pages.
Use of any outside resources verbatim as your answer (like copy-paste or quotation) is not allowed, and will
be treated as unauthorized collaboration (and reported as plagiarism).
Please do NOT post assignment solutions on MS Teams or Brightspace (or other platforms not used in the course such as Discord) or it will be penalized. Moreover, posting the assignment questions to any external forums/websites is NOT permitted and will also be penalized/reported.
Questions – part 1 [14]
The following questions (where applicable) will be based on the original 3000test.c
in Tutorial 5:
Note: in this part, always use “ls -ls” to see both the logical size and the physical size.
Calling an external program explicitly in the code is not allowed.
To save you from having to read the superblock to determine filesystem block size, you can find it out using “sudo blockdev –getbsz [block_device_name]” and hardcode it in your code.
1. Make your 3000test output one of the following lines, when the provided file is a symbolic link (dereferenced) or a regular file:
– Sparse file: YES; Internal fragmentation: YES
– Sparse file: NO; Internal fragmentation: YES – Sparse file: NO; Internal fragmentation: NO – Sparse file: YES; Internal fragmentation: NO This must reflect the actual situation of the file [3].
Definition: a sparse file is a file with “holes”, which means not all data blocks are allocated (blocks containing just 0’s will not be allocated until written to; the corresponding data pointers in the inode just point to NULL). Internal fragmentation in the context of this course refers to wasted/unused space (not covered by the file’s actual size) within allocated blocks.
[1.5] Explain why you made this code change, with a reasoning like “why checking this can lead to the conclusion of this YES/NO”.
You can make sure it works by testing it with a few files of the four cases above.
Hint: make good use of the C struct in Tutorial 5.
2. YoumayhavenoticedinTutorial3that,theC alsocontainsthefiletype information d_type (see ) such as for regular files. This way, you can avoid reading the actual file (i.e., its inode by ) unless needed, which may improve performance when processing a large number of files (not our case though) in the same directory. For instance, you will only touch the actual file when you want to search for ‘a’ in it but not before that. [2] State the main reason why it will not be a good idea to use d_type in place of st_mode in 3000test from the perspective of programming/implementation.
Note that both fields will be available for our ext4 (or neither otherwise) so no compatibility issues.
3. Ifyouwanttoavoidthecalltommap()(line59)withoutaffectingthecurrentfunctionality(i.e., 3000test still does all it can do now), 1) [2] what standard C functions can you use instead? (mention all needed) 2) what will likely be a design challenge (or a value to determine) to trade off between performance and space? [1] Please be specific.
From here, we can see the advantages of mmap() which we will discuss further later in the course. The size of the file to read cannot be predetermined. You do not need to implement it this time (no code submission; just written answers).
struct dirent
man readdir
4. Let’sgenerateafilefilledwithall1’susingthiscommand:
LC_CTYPE=C tr ‘\0’ ‘\377’ < /dev/zero | dd of=myfile count=200
(‘\377’ is 0xff in octal)
Of course, myfile cannot be made sparse as it does not contain any 0’s. Next, make a few copies of it using the cp command:
for i in {1..3}; do cp myfile "myfile$i"; done
The next few commands will add a chunk of 0’s in different ways:
Then, try to optimize them using this command:
dd if=/dev/zero of=myfile1 conv=notrunc count=6
dd if=/dev/zero of=myfile2 conv=notrunc count=8 seek=1
dd if=/dev/zero of=myfile3 conv=notrunc count=8
You can use either ls -ls or your updated 3000test in Q1 to check, and you will see although we have injected 0’s to all three files, only myfile3.processed was optimized (smaller physical size).
[3] Explain why myfile1.processed and myfile2.processed did not get optimized (no change), by showing some calculation with numbers.
5. In the question above, we used a translator to convert the 0’s from to 1’s. [0.5] Explain why we cannot simply use to create a file
filled with 1’s of any size (the 200 is an example). Nevertheless, inspired by Q4, you can still adjust it to make it work. Give it a try [1] (no need to consider performance).
You should not change the commands used here, i.e., echo and dd, or introduce new commands.
echo -n -e "\xff" | dd of=myfile count=200
for i in {1..3}; do cp "myfile$i" "myfile$i.processed" --sparse=always;
Questions – part 2 [6]
The following questions (where applicable) will be based on the patched version of 3000userlogin.c in the posted explanation of Tutorial 4. The diff file is also here for
your reference.
Keep in mind each time you update the code, you will need to follow the same steps in Tutorial 4 to make it “setuid root”. You can put them in a script or a Makefile if you want to.
Calling an external program explicitly in the code is not allowed.
6. [1] Without “setuid root”, why does 3000userlogin.patched report “ ” even for a valid existing user, instead of “ ” as was seen with the original 3000userlogin? Make sure to mention the root cause.
7. You may have noticed the behavior of 3000userlogin resembles that of the su command to some extent (although still different). The main difference is whether the whole environment is reinitialized for the specified user (login) or just temporarily running something on its behalf (UID/GID switch). We can implement both in a way similar to busybox.
In the directory where you have built and prepared the downloaded 3000userlogin.patched:
Then you can simply use and ./login someuser
You should also be able to use it anywhere else by specifying the full pathname, e.g., /home/student/su someuser.
Make changes to 3000userlogin.patched to support this [3].
Requirements:
- When called as ./login, the current behavior of 3000userlogin.patched must be preserved.
- When called as ./su, 1) the current directory should not be changed; 2) only , and
are set to that of “someuser”, other environment variables preserved; 3) run
ln -s 3000userlogin.patched login
Could not find user
Failed to change to gid
ln -s 3000userlogin.patched su
./su someuser
(instead of the hardcoded ) as the new shell; 4) obviously UID and GID should be set for “someuser”.
In both cases, you should be able to return using exit like before.
8. [1] In the question above, explain why you do not need to follow the same steps to “setuid root” the two links (./su and ./login) but they still work.
9. Canweuselnwithout“-s”toachievethesamepurpose?[1]Whynot(ifNO)orwhatwillbethe implications (if YES)? (note the mark is not for a YES/NO answer but the explanation)
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com