User-level programs

Our System/161 simulator can run normal C programs if they are compiled with a cross-compiler, os161-gcc. A cross compiler runs on a host (e.g., a Linux x86 machine) and produces MIPS executables; it is the same compiler used to compile the OS/161 kernel. Various user-level programs already exist in userland/bin, userland/testbin, and userland/sbin. Note: that only a small subset these programs will execute successfully due to OS/161 only supporting a small subset of the system call interface.

To create new user programs (for testing purposes), you need to edit the Makefile in bin, sbin, or testbin (depending on where you put your programs) and then create a directory similar to those that already exist. Use an existing program and its Makefile as a template. In assignment 1, we did this for you to produce the starting state of test programs such as rot13 and boxes.

Design

In the beginning, you should tackle this assignment by producing a DESIGN. The design should clearly reflect the development of your solution. The design should not merely be what you programmed. If you try to code first and design later, or even if you design hastily and rush into coding, you will most certainly end up in a software "tar pit". Don't do it! Plan everything you will do. Don't even think about coding until you can precisely explain to your partner what problems you need to solve and how the pieces relate to each other. Note that it can often be hard to write (or talk) about new software design, you are facing problems that you have not seen before, and therefore even finding terminology to describe your ideas can be difficult. There is no magic solution to this problem; but it gets easier with practice. The important thing is to go ahead and try. Always try to describe your ideas and designs to someone else. In order to reach an understanding, you may have to invent terminology and notation, this is fine. If you do this, by the time you have completed your design, you will find that you have the ability to efficiently discuss problems that you have never seen before. Why do you think that CS is filled with so much jargon? To help you get started, we have provided the following questions as a guide for reading through the code to comprehend what is already provided.

To get a feel for what problems you need to solve, review the design questions and design document section later in this specification.

Setup

We assume after ASST0 and ASST1 that you now have some familiarity with setting up for OS/161 development. If you need more detail, refer back to ASST0.

You will soon be able to clone the ASST2 source repository from CSE's cloud gitlab instance, replacing the XXX with your 3 digit group number:

% cd ~/cs3231
% git clone https://zNNNNNNN@nw-syd-gitlab.cseunsw.tech/COMP3231/25T2/grpXXX-asst2.git asst2-src

The above requires a group number! If you haven't been issued a group yet, see the group nomination site here.

You should receive an email from the gitlab instance once your group repository has been set up. This will typically happen overnight once your group nomination is finished.

If you're still waiting for your group gitlab repository, you can fetch the assignment 2 sources prior to your group being formed via a read-only link as in assignment 1:

% cd ~/cs3231
% git clone https://www.cse.unsw.edu.au/~cs3231/25T2/assignments/asst2-src

Notes:

• The nominations to group ID and group ID to gitlab project scripts run from time to time. There might be a modest delay between nominating your partner and receiving an email about a gitlab repository. If you have been waiting for much more than 24h, ask on the forum.
• The gitlab repository above is shared between you and your partner. You can both push and pull changes to and from the repository to cooperate on the assignment. If you are not familiar with cooperative software development and git you should consider spending a little time familiarising yourself with git.

Building and Testing the Provided Code

% cd ~/cs3231/asst2-src
% ./configure

% cd ~/cs3231/asst2-src
% bmake
% bmake install

% cd ~/cs3231/asst2-src/kern/conf
% ./config ASST2

% cd ../compile/ASST2
% bmake depend
% bmake
% bmake install

% cd ~/cs3231/root
% wget http://cgi.cse.unsw.edu.au/~cs3231/25T2/assignments/asst2/sys161.conf

Unknown syscall 55
Unknown syscall 55
Unknown syscall 55
Unknown syscall 55
:
:
Unknown syscall 55
Unknown syscall 55
Unknown syscall 3
exit() was called, but it's unimplemented.
This is expected if your user-level program has finished.
panic: Can't continue further until sys_exit() is implemented

OS/161 kernel [? for menu]: p /testbin/asst2
Operation took 0.000212160 seconds
OS/161 kernel [? for menu]:

**********
* File Tester
**********
* write() works for stdout
**********
* write() works for stderr
**********
* opening new file "test.file"
* open() got fd 3
* writing test string
* wrote 45 bytes
* writing test string again
* wrote 45 bytes
* closing file
**********
* opening old file "test.file"
* open() got fd 3
* reading entire file into buffer
* attempting read of 500 bytes
* read 90 bytes
* attempting read of 410 bytes
* read 0 bytes
* reading complete
* file content okay
**********
* testing lseek
* reading 10 bytes of file into buffer
* attempting read of 10 bytes
* read 10 bytes
* reading complete
* file lseek  okay
* closing file
exit() was called, but it's unimplemented.
This is expected if your user-level program has finished.
panic: Can't continue further until sys_exit() is implemented

The Assignment Task: File System Calls

Of the full range of system calls that is listed in kern/include/kern/syscall.h, your task is to implement the following file-based system calls: open, read, write, lseek, close, dup2, and document your design. You need to perform this assignment in three steps, see the specs of Parts 1-3 below. As always, it's important to read the entire specification document before you get started.

Note: You will be writing the kernel code that implements part of the system call functionality within the kernel. You are not writing the C stubs that user-level applications call to invoke the system calls. The userland stubs are automatically generated when you build OS/161 in build/userland/lib/libc/syscalls.S which you should not modify.

It's crucial that your syscalls handle all error conditions gracefully (i.e., without crashing OS/161.) no matter what an application requests. Your code should also be memory leak free. You should consult the OS/161 man pages (also included in the distribution) and understand the system calls that you must implement. Your system calls must return the correct value (in case of success) or an appropriate error code (in case of failure) as specified in the man pages. Some of the auto-marking scripts rely on the return of error codes, however, we are lenient as to the specific code in the case of potential ambiguity as to which error code would be most appropriate. It's also not necessary to generate all error codes listed in the man pages.

The file userland/include/unistd.h contains the user-level interface definition of the system calls. This interface is different from that of the kernel functions that you will define to implement these calls. You need to design the kernel side of this interface. The function prototype for your interface can be put it in kern/include/syscall.h. The integer codes for the calls are defined in kern/include/kern/syscall.h.

Part 1: Implement the base system calls

Marks: Worth 45% of the marks for this assignment

Implement open(), close(), read() and write() as described above.

Once these are working, OS/161 can run various simple programs. The asst2 test we have provided you with should succeed in producing output and testing read and write operations, up until the point at which it attempts lseek.

Once this simple test passes, you have finished Part 1.

Once you have finished Part 1, commit your work using git. You will need to refer back to this commit in Part 3. Make sure you have committed everything that you have done. It would be a good idea if this commit message mentioned that you had finished Part 1.

You will need to find this revision number again for Part 3. We recommend you name the current revision as an additional branch, for instance like this:

% # add a branch "part1_done" at the current revision.
% git branch part1_done
% # list your branches. you should still be on the "main" branch.
% git branch -v

Part 2: Implement the unusual system calls

Marks: Worth 40% of the marks for this assignment

Implement dup2() and lseek() as described above.

The asst2 test will do some simple tests of lseek(). We do not provide a test for dup2(), and you might want to add one.

Once you are happy that these system calls are working as well, you have finished Part 2. Just like Part 1, make sure that your work is committed using git. If you plan to make further changes, e.g. in the advanced parts, make sure that you take a note of this revision number in the same way.

It is quite likely that you will have to change some of your Part 1 work to get the Part 2 system calls working, or that you will find a bug in your Part 1 work that you want to fix. That is OK, and there is no need to change the Part 1 work that you recorded above.

Part 3: Documenting your solution

This is a compulsory component of this assignment.

Marks: Notionally worth 15% of the marks for this assignment, but see the note at the end.

You must submit a small design document discussing how you have gone about the assignment. This part of the assignment is done individually, and this part does not need to be added to git or any bundle. If you are working with a partner, both you and your partner should be able to write this short document yourselves.

Your design document should be written as a markdown file named part3.md. We expect it to be 500 to 1500 words, i.e. short and to the point.

In this document, we expect you to discuss:

• What significant data structures did you add (in Part 1 or 2)? What function do they perform?
• What data structures are per-process and what structures are shared between processes?
• What are the main issues related to transferring data to and from applications?
• If you were to implement fork() (although you haven't yet) and introduce concurrency, where would concurrency issues appear in your implementation?
• How did your design change from Part 1 to Part 2? More on this below.

In particular, we are interested in how your design evolved from Part 1 to Part 2. Did you anticipate the requirements of Part 2 entirely during Part 1, or did you have to revise your design?

Recall the revision number you saved at the end of Part 1? You can view the difference in your sources between then and now with git diff:

% git diff -r part1_done..HEAD

The above assumes you saved the branch name part1_done in Part 1, and have your Part 2 work checked out now (HEAD). You can compute the change between different revision points in your work as well.

We have included an example markdown file with some example relevant syntax to get you started. The example is at submission/example.md in the released sources. Markdown is mostly a text format, and you don't need to use any fancy features. However, you should use at least one text snippet taken from a git diff command to describe the difference between your sources at your Part 1 checkpoint and your finished Part 2.

Such a diff might look like this:

diff --git a/src/userland/testbin/asst2/asst2.c b/src/userland/testbin/asst2/asst2.c
index 0ef2dc1..d074f18 100644
--- a/src/userland/testbin/asst2/asst2.c
+++ b/src/userland/testbin/asst2/asst2.c
@@ -77,7 +77,7 @@ main(int argc, char * argv[])
         r = strlen(teststr);
         do {
                 if (buf[k] != teststr[j]) {
-                        printf("ERROR  file contents mismatch\n");
+                        printf("ERROR  file contents mismatch at char %d.\n", k);
                         exit(1);
                 }
                 k++;

The above diff will render slightly better in markdown if the "diff" language is specified for the blockquote. See the example file submission/example.md for an example.

If you view a markdown file on gitlab, it will be rendered into an HTML document. This is how a marker will inspect your document, and we recommend you check that any markdown features you have used render correctly. On that point, recall that you are not meant to be working with your partner? We recommend that you put this file in a private repository (they are easy to create) or otherwise manage the markdown->HTML rendering.

Note on marking: This part is worth 15% of the assignment. This will be marked by hand to check whether you followed the instructions above and produced a short note that is reasonably easy to read. The small amount of marks should motivate you to write something we can read. Clearly the point of the exercise is to check that you (and your partner, if you have one) understand what you are submitting. Evidence that this is not the case may, in rare cases, have greater consequences than the loss of 15% of the marks for the assignment.

Basic Assignment Submission

The standard submission instructions are available on the Wiki . Like ASST0 and ASST1, you will be submitting the git repository bundle via CSE's give system. For ASST2, the submission system will do a test build and run a simple test to confirm your bundle at least compiles. For the basic submission, you will also include a part3.md document separately.

Submit your bundle and your Part 3 document via give:

% cd ~
% give cs3231 asst2 asst2.bundle part3.md

You're now done.

Even though the generated bundle should represent all the changes you have made to the supplied code, occasionally students do something "ingenious". So always push your changes back to gitlab (and keep your git repository) so that you may recover your assignment should something go wrong.

Part A-1: Forking additional processes

Marks: worth the equivalent of 10% of this assignment

Implement the fork() and getpid() system calls.

This system call (together with execv() below) is probably the most difficult part of the whole assignment, but also the most rewarding. They enable multiprogramming and make OS/161 a much more useful entity. fork() is your mechanism for creating new processes. It should make a copy of the invoking process and make sure that the parent and child processes each observe the correct return value (that is, 0 for the child and the newly created pid for the parent).

You will want to think carefully through the design of fork. If you plan on doing Part A-4, consider it together with execv() to make sure that each system call is performing the correct functionality. More on that below.

Part A-2: Copying Data within the Kernel

Marks: worth the equivalent of 5% of this assignment

Part A-2 is unrelated to Part A-1.

Your task is to implement the copy_splice() system call, which is a hybrid of Linux's splice() and copy_file_range() system calls. The Linux variants specify which file objects they can be used on, splice() only for pipes and copy_file_range() only for regular files. Your copy_splice() should support both regular files and devices such as the console device "con:".

The intended signature is:

int copy_splice(int fd_in, int fd_out, size_t len);

This signature omits the offsets and flags in the Linux signatures for splice() etc, but has otherwise the same intention. The user program wants to copy up to len bytes from the input file to the output file without those bytes needing to be copied to and from user memory.

This is not a complex concept, and requires less fiddly user-space copies than the read() and write() system calls from the main part. However, there are some snags, and the Linux man pages mention buggy implementations. Think through all the possible cases before you start implementing.

Part A-3: Ending Processes

Marks: worth the equivalent of 5% of this assignment

This part depends on Part A-1.

Part A-4: Running new binaries

Marks: worth the equivalent of 5% of this assignment

This part depends on Part A-1.

Design Questions

Here are some additional questions and thoughts to aid in your design. They are not, by any means, meant to be a comprehensive list of all the issues you will want to consider. Your system must allow user programs to receive arguments from the command line. By the end of Parts A-3 and A-4, you should be capable of executing lines (in user programs) such as:

char *filename = "/bin/cp";
char *args[4];
pid_t pid;
args[0] = "cp";
args[1] = "file1";
args[2] = "file2";
args[3] = NULL;
pid = fork();
if (pid == 0)
execv(filename, argv);

which will load the executable file cp, install it as a new process, and execute it. The new process will then find file1 on the disk and copy it to file2. You can test your implementation using OS/161's shell, /bin/sh.

Some questions to think about:

• Passing arguments from one user program, through the kernel, into another user program, is a bit of a chore. What form does this take in C? This is rather tricky, and there are many ways to be led astray. You will probably find that very detailed pictures and several walk-throughs will be most helpful.
• How will you determine: (a) the stack pointer initial value; (b) the initial register contents; (c) the return value; (d) whether you can exec the program at all?
• What new data structures will you need to manage multiple processes?
• What relationships do these new structures have with the rest of the system?
• How will you manage file accesses? When we invoke the cat command, and it starts to read file1, what will happen if the shell also tries to read file1? What would you like to happen?
• How will you keep track of running processes. For which system calls is this useful?
• How will you implement the execv system call. How is the argument passing in this function different from that of other system calls?

Advanced Assignment Submission

Submission for the advanced assignment is similar to the basic assignment, except the advance component is given to a distinguished assignment name: asst2_adv. Again, you need to generate a bundle based on your repository. Note: Our marking scripts will switch to the asst2_adv branch prior to testing the advanced assignment.

We've added some text files to the repository in which you will summarise which parts you've attempted and how you've gone about them. These files are in the submission directory at the top of the repository, with one text file per advanced part.

For each advanced part you have attempted, adjust that file to remove the "not attempted" message and include a short (100 - 500 word) description of your approach. If you are working with a partner, you can write these descriptions together (unlike part3.md above). If you are working with a partner, you should both submit the advanced part bundle, to avoid confusion, even though the contents will be the same.

Submit your solution by doing:

% cd ~
% give cs3231 asst2_adv asst2_adv.bundle

Copying Memory

In assignment 1, we already thought about how to move values between user programs and the kernel. However we only addressed the case where arguments and return values are passed in registers. In assignment 2, we need to move larger quantities of data in and out of user memory.

As we will soon cover in lectures, the kernel and user programs have a different view of memory. OS/161 is a copyin()/copyout() kernel. Any time that user memory is to be read, it is first copied into a kernel buffer via copyin(), and vice versa, all writes to user memory are performed via copyout().

It is important that the kernel doesn't use a user-level address as a pointer directly. For instance, the char * and void * arguments to user-level open(), write() etc should be treated as userptr_t within the OS/161 kernel.

If you haven't done it already, look for the copyin() and copyout() functions and their close relatives. There is a longer guided walkthrough available here that discusses them in detail.

The related UIO mechanism is used for copying data to or from an I/O subsystem (e.g. a file system or other device). It can copy directly to or from user memory, in which case it will perform the copyin() and copyout() calls required.

Arguments in Memory

Like in assignment 1, most system call arguments will be passed in via trapframe registers a0 through a3.

The exception is lseek(int fd, off_t pos, int whence). Its offset argument pos is 64-bit, so requires two argument registers. By convention, the register pairs available are a0/a1 and a2/a3, so pos is placed in the second of these, a1 is not used, and whence is left to be passed on the stack. There is further discussion of the implications of the calling convention in the OS/161 sources, and I recommend you look for it.

To convert the two 32-bit argument registers into a 64-bit value and vice versa, you can use the join32to64 and split64to32 helper functions.

The VOP macros

In the VFS lecture, we described how VFS vnode objects have a "method table" of function pointers. This allows a write operation on an "emufs" file and a write operation on the console device to be implemented by completed different functions.

The vnode.h header provides macros that capture the common pattern of fetching an operation from a vnode and passing the vnode back to it. These look like VOP_FOO, and, as the header explains, VOP_FOO(vnode, arg1, ...) expands to vnode->vn_ops->vop_foo(vnode, arg1, ...).

On Specification and Error Codes

When the manual page for one of these syscalls suggests possible error codes (EINVAL, EMFILE, EIO etc), what does that mean? Is it a problem if your implementation won't generate one of those errors?

For instance, does your open implementation need to be able to generate both EMFILE and ENFILE errors?

Not necessarily. Often a manual entry needs to give a specification of an interface that might have many implementations. One example is UNIX. A UNIX programmer might assume that they can read the manual entries for open read and close and create a C program that works on many UNIX implementations (Linux, MacOS X, BSD, SunOS etc). Another example is OS/161. The manual pages specify the interface. Literally hundreds of students are producing different implementations, which might have slightly different behaviour.

When the manual lists a collection of error codes, it's not a problem if your code doesn't generate all of them. It would be a problem if your code generated error codes that aren't on the list; that might be surprising to valid user programs. It's unlikely we'll test exhaustively for that kind of error though, we'd have to intentionally drive your code into rare error cases.

On the Part 3 Format

Why this strange discussion of a change? Why use markdown format?

This assignment component is written vaguely in the style of a "pull request" or "merge request" common on a system like github or gitlab. Imagine that some other author or team had written a working Part 1, and you've now branched from that to add new features. You want to convince the upstream developers to pull/merge your changes into the official mainline. This doesn't come for free, especially if your new work needs to make changes to existing code, so you need to discuss what you changed, and try to justify why those changes were needed.

On Part 3 Diffs

What is this part about git diff?

The part 3 instructions above say you should run git diff between your finished-part-1 and finished-part-2 revisions, pick some part of that diff, and quote it and describe it. The spec says so, so do it, or you'll lose easy marks.

Various people have asked what changes they could possibly discuss. If you're still wondering, we recommend you look at the diff generated as above. Everything in it is a candidate to be discussed, including diff chunks that just add new code. Some parts might be more interesting than others. If there is nothing especially interesting, that itself is worth explaining.

On Team-Work and Contributions

Most students are working in a pair, and this is the recommended approach.

Multiple students have asked how closely the pair have to work together, how much they have to contribute to each part, and to what extent it matters who commits what.

It is natural to divide up the task somewhat, especially in the early phases. The parts of the job are fairly interlinked, though, so to get the assignment working you will probably have to review your partner's work, and likely make changes or request changes. If that somehow isn't the case, you should review your partner's work anyway. Consider it your professional duty to take some responsibility for whatever you submit.

We have mentioned already that you are not allowed to simply "trade" and have one partner do all of assignment 2 and the other do all of assignment 3.

The expectation is that the partners will make roughly even contributions. If that is far from the case, you can request we intervene, in which case we will interrogate both partners and whatever documentary evidence exists. Hopefully we don't have to do that.

On Part 1 and Part 2

Multiple students have asked whether it is a problem to anticipate the needs of part 2 while implementing part 1. That is expected. It is normal to plan ahead. It is plausible that you plan ahead so well that you need to make no modifications to your part 1 plans to complete part 2. We expect it to be much more common to need to make some adjustments.

It is also normal, when implementing this kind of system, to test your work as early as possible. Until your system is tested, it's probably not working.

We would recommend you test your implementation of the part 1 features as soon as you have them available, and once they are working, checkpoint it.