Output and Input

Let's learn to write text!

Unsurprisingly, your program writes text to the screen by invoking a system call. Specifically, this is the write system call, and its syscall number is 1. However, the write system call also needs to specify, via its parameters, what data to write and where to write it to.

You may remember, from the Practicing Piping module of the Linux Luminarium dojo, the concept of File Descriptors (FDs). As a reminder, each process starts out with three FDs:

• FD 0: Standard Input is the channel through which the process takes input. For example, your shell uses Standard Input to read the commands that you input.
• FD 1: Standard Output is the channel through which processes output normal data, such as the flag when it is printed to you in previous challenges or the output of utilities such as ls.
• FD 2: Standard Error is the channel through which processes output error details. For example, if you mistype a command, the shell will output, over standard error, that this command does not exist.

It turns out that, in your write system call, this is how you specify where to write the data to! The first (and only) parameter to your exit system call was your exit code (mov rdi, 42), and the first (but, in this case, not only!) parameter to write is the file descriptor. If you want to write to standard output, you would set rdi to 1. If you want to write to standard error, you would set rdi to 2. Super simple!

This leaves us with what to write. Now, you could imagine a world where we specify what to write through yet another register parameter to the write system call. But these registers don't fit a ton of data, and to write out a long story like this challenge description, you'd need to invoke the write system call multiple times. Relatively speaking, this has a lot of performance cost --- the CPU needs to switch from executing the instructions of your program to executing the instructions of Linux itself, do a bunch of housekeeping computation, interact with your hardware to get the actual pixels to show up on your screen, and then switch back. This is slow, and so we try to minimize the number of times we invoke system calls.

Of course, the solution to this is to write multiple characters at the same time. The write system call does this by taking two parameters for the "what": a where (in memory) to start writing from and a how many characters to write. These parameters are passed as the second and third parameters to write. In the kinda-C syntax that we learned from strace, this would be:

write(file_descriptor, memory_address, number_of_characters_to_write)

For a more concrete example, if you wanted to write 10 characters starting from some memory address to standard output (file descriptor 1), this would be:

write(1, memory_address, 10);

Wow, that's simple! Now, how do we actually specify these parameters?

1. We'll pass the first parameter of a system call, as we reviewed above, in the rdi register.
2. We'll pass the second parameter via the rsi register. The agreed-upon convention in Linux is that rsi is used as the second parameter to system calls.
3. We'll pass the third parameter via the rdx register. This is the most confusing part of this entire module: rdi (the register holding the first parameter) has such a similar name to rdx that it's really easy to mix up and, unfortunately, the naming is this way for historic reasons and is here to stay. Oh well... It's just something we have to be careful about. Maybe a mnemonic like "rdi is the initial parameter while rdx is the xtra parameter"? Or just think of it as having to keep track of different friends with similar names, and you'll be fine.

And, of course, the write syscall index into rax itself: 1. Other than the rdi vs rdx confusion, this is really easy!

Now, you know how to set the system call number and how to set the rest of the registers. But where in memory is the data you need to write?

In this challenge, your program is invoked with a command-line argument, something like:

/tmp/your-program H

Recall that when a program is run with arguments, the stack stores pointers to each argument. These are addresses stored in memory: [rsp+16] doesn't contain the argument text directly --- it contains the address where that text lives.

So, to get the memory address of the first argument, you simply load the pointer from the stack, as you've done before!

mov rsi, [rsp+16]

This puts the memory address of the first argument's text into rsi --- exactly what write needs as its second parameter!

Your program will be invoked with a single character as its first argument. Call write to write that single character (for now! We'll do multiple-character writes later) to standard output, and we'll give you the flag!

Okay, our previous solution wrote output but then crashed. In this level, you will write output, and then not crash!

We'll do this by invoking the write system call, and then invoking the exit system call to cleanly exit the program. How do we invoke two system calls? Just like you invoke two instructions! First, you set up the necessary registers and invoke write, then you set up the necessary registers and invoke exit!

Your previous solution had 5 instructions (loading the first argument's address from the stack, setting rdi, setting rdx, setting rax, and syscall). This one should have those 5, plus three more for exit (setting rdi to the exit code, setting rax to syscall index 60, and syscall). For this level, let's exit with exit code 42!

Okay, we have one thing left for this run of challenges. You've written out a single byte, and now we'll practice writing out multiple bytes. In this level, the flag itself is passed as the first argument to your program! Can you write all 64 characters of it to stdout?

Hint: The only thing you should have to change compared to your previous solution is the value in rdx!

You now know how to output data to stdout using write. But how does your program receive input data? It reads it from stdin!

Like write, read is a system call that shunts data around between file descriptors and memory, and its syscall number is 0. In read's case, it reads some amount of bytes from the provided file descriptor and stores them in memory. The C-style syntax is the same as write:

read(0, some_address, 5);

This will read 5 bytes from file descriptor 0 (stdin) into memory starting from some_address. So, if you type in (or pipe in) HELLO HACKERS into stdin, the above read call would result in the following memory configuration:

     Address     │ Contents
+───────────────────────────+
│ some_address   │ 48       │
│ some_address+1 │ 45       │
│ some_address+2 │ 4c       │
│ some_address+3 │ 4c       │
│ some_address+4 │ 4f       │
+───────────────────────────+

What are those numbers?? They are hexadecimal representations of ASCII-encoded letters. If those words don't make sense, please run through the first half or so of the Dealing with Data module and then come back here!

In this level, we will combine read with our previous write abilities. The flag will be piped into your program's stdin --- 64 bytes of it. Your program should:

1. first read 64 bytes from stdin to your program's memory
2. write those 64 bytes from that memory location to stdout
3. finally, exit with the exit code 42.

But what address should you use? You need somewhere that's valid and writable, and you already know about one such place: the stack! The rsp register points to the top of the stack, and there's plenty of writable space there. So you can just use rsp as your memory address: mov rsi, rsp.

DEBUGGING: Having trouble? Recall the Introspection module! Build your program and run it with strace to see what's happening at the system call level, or run it in gdb to inspect the values of registers and memory to see what's unexpected.

REMEMBER: You've basically already written steps 2 and 3 (though in the previous challenges, you loaded rsi from [rsp+16] --- here, you'll set it to rsp directly with mov rsi, rsp!). All you have to do is add step 1!

So far, your program has only interacted with stdin and stdout, but what about files on disk? To access a file, you first need to open it using the open system call.

The open system call (syscall number 2) takes a pointer to a filename string and returns a brand-new file descriptor referring to that file:

open("/flag", 0);

The second argument specifies additional modes and permissions for the file, but 0 requests the default: read-only.

The registers for open follow the same convention:

When open returns, rax contains the new file descriptor (fd) number. Recall that file descriptor 0 is stdin, file descriptor 1 is stdout, and file descriptor 2 is stderr. Other files that are open are just represented by other file descriptors, incrementing from 3 onwards! You'll use this fd as the first argument to read, just like you did for stdin earlier, but this time read will read from your file.

How to load the filename into memory? In this level, the path to the flag (/flag) will be passed as the first argument to your program. You already know how to load that: mov rdi, [rsp+16].

Your program should:

1. Load a pointer to the filename (stored at [rsp+16], the first argument) into rdi
2. Specify the default of read access for the second argument (set rsi to 0).
3. open it (syscall 2)
4. read 64 bytes from the returned fd into memory. The returned fd will be stored in rax; you'll need to move that to rdi for read's first argument. Make sure to do this before you set the syscall number for write!
5. write those 64 bytes to stdout
6. exit with code 42 (syscall 60)

DEBUGGING: Having trouble? Use strace to see your system calls in action --- it will show you exactly what arguments each syscall receives and what it returns. If open is returning -1, double-check your filename pointer. If read returns 0, the file descriptor from open might be wrong.

In the previous level, the filename was passed as an argument to your program. But what if you need to open a file whose path you already know? You can hardcode the filename string directly into your program by writing it onto the stack, byte by byte!

The open syscall needs a pointer to the filename, so you need the bytes / f l a g stored somewhere in memory. You already know a writable memory address: rsp (the stack). You can write each character one byte at a time:

mov BYTE PTR [rsp], '/'
mov BYTE PTR [rsp+1], 'f'
mov BYTE PTR [rsp+2], 'l'
mov BYTE PTR [rsp+3], 'a'
mov BYTE PTR [rsp+4], 'g'
mov BYTE PTR [rsp+5], 0

A few things to note here:

• BYTE PTR: When you write to a memory address like [rsp] using an immediate value (a number or character), the CPU doesn't know how many bytes you intend to write --- one? two? eight? BYTE PTR is a size directive that tells the assembler "I mean exactly one byte." Without it, the assembler won't know what you want and will refuse to assemble the instruction.

• Single quotes: In assembly, a single-quoted character like 'f' represents that character's one-byte ASCII value. So 'f' is just a convenient way of writing 0x66, and '/' is 0x2f.

• The null byte: The last byte we write is 0 --- a special null byte. This is how Linux knows where a string ends: it reads bytes starting from the pointer you give it and stops when it hits a 0 byte. Without it, open would keep reading past "flag" into whatever else is on the stack, and you'd be trying to open a file with a nonsense name!

After writing these bytes, rsp points to the null-terminated string "/flag", ready to pass to open.

Your turn! This time, no arguments are passed to your program. You must construct the filename yourself.

Your program should:

1. Write "/flag\0" onto the stack byte by byte using mov BYTE PTR [rsp+N], ...
2. open it (syscall 2): rdi = rsp (the string you just wrote), rsi = 0
3. read 64 bytes from the returned fd into memory (syscall 0)
4. write those 64 bytes to stdout (syscall 1)
5. exit with code 42 (syscall 60)

DEBUGGING: Having trouble? Use strace to trace your syscalls. If open returns -1, your string pointer or encoding might be off. Try x/s $rsp in gdb to see what string is actually on the stack.