Finding Information About Libraries

Okay, so now that you know about libraries, the question is, how do you find out what libraries you have on your system and what they do? Well, let's skip that question for a minute and ask another question: How do programmers describe functions to each other in their documentation? Let's take a look at the function printf. Its calling interface (usually referred to as a prototype) looks like this:

 int printf(char *string, ...); 

In Linux, functions are described in the C programming language. In fact, most Linux programs are written in C. That is why most documentation and binary compatibility is defined using the C language. The interface to the printf function above is described using the C programming language.

This definition means that there is a function printf. The things inside the parenthesis are the function's parameters or arguments. The first parameter here is char *string. This means there is a parameter named string (the name isn't important, except to use for talking about it), which has a type char *.char means that it wants a single-byte character. The * after it means that it doesn't actually want a character as an argument, but instead it wants the address of a character or sequence of characters. If you look back at our helloworld program, you will notice that the function call looked like this:

  pushl $hello  call  printf 

So, we pushed the address of the hello string, rather than the actual characters. You might notice that we didn't push the length of the string. The way that printf found the end of the string was because we ended it with a null character (\0). Many functions work that way, especially C language functions. The int before the function definition tell what type of value the function will return in %eax when it returns. printf will return an int when it's through. Now, after the char *string, we have a series of periods, .... This means that it can take an indefinite number of additional arguments after the string. Most functions can only take a specified number of arguments. printf, however, can take many. It will look into the string parameter, and everywhere it sees the characters %s, it will look for another string from the stack to insert, and everywhere it sees %d it will look for a number from the stack to insert. This is best described using an example:

 #PURPOSE:  This program is to demonstrate how to call printf #  .section .data #This string is called the format string.  It's the first #parameter, and printf uses it to find out how many parameters #it was given, and what kind they are. firststring:  .ascii "Hello! %s is a %s who loves the number %d\n\0" name:  .ascii "Jonathan\0" personstring:  .ascii "person\0" #This could also have been an .equ, but we decided to give it #a real memory location just for kicks numberloved:  .long 3  .section .text  .globl _start _start:  #note that the parameters are passed in the  #reverse order that they are listed in the  #function's prototype.  pushl numberloved    #This is the %d  pushl $personstring  #This is the second %s  pushl $name          #This is the first %s  pushl $firststring   #This is the format string                       #in the prototype  call printf  pushl $0  call  exit 

Type it in with the filename printf-example.s, and then do the following commands:

 as printf-example.s -o printf-example.o ld printf-example.o -o printf-example -lc \    -dynamic-linker /lib/ld-linux.so.2 

Then run the program with ./printf-example, and it should say this:

 Hello! Jonathan is a person who loves the number 3 

Now, if you look at the code, you'll see that we actually push the format string last, even though it's the first parameter listed. You always push a functions parameters in reverse order.^[2] You may be wondering how the printf function knows how many parameters there are. Well, it searches through your string, and counts how many %ds and %ss it finds, and then grabs that number of parameters from the stack. If the parameter matches a %d, it treats it as a number, and if it matches a %s, it treats it as a pointer to a null-terminated string. printf has many more features than this, but these are the most-used ones. So, as you can see, printf can make output a lot easier, but it also has a lot of overhead, because it has to count the number of characters in the string, look through it for all of the control characters it needs to replace, pull them off the stack, convert them to a suitable representation (numbers have to be converted to strings, etc), and stick them all together appropriately.

We've seen how to use the C programming language prototypes to call library functions. To use them effectively, however, you need to know several more of the possible data types for reading functions. Here are the main ones:

int

• An int is an integer number (4 bytes on x86 processor).

long

• A long is also an integer number (4 bytes on an x86 processor).

long long

• A long long is an integer number that's larger than a long (8 bytes on an x86 processor).

short

• A short is an integer number that's shorter than an int (2 bytes on an x86 processor).

char

• A char is a single-byte integer number. This is mostly used for storing character data, since ASCII strings usually are represented with one byte per character.

float

• A float is a floating-point number (4 bytes on an x86 processor). Floating-point numbers will be explained in more depth in the Section called Floating-point Numbers in Chapter 10.

.double

• A double is a floating-point number that is larger than a float (8 bytes on an x86 processor).

unsigned

• unsigned is a modifier used for any of the above types which keeps them from being used as signed quantities. The difference between signed and unsigned numbers will be discussed in Chapter 10.

*

• An asterisk (*) is used to denote that the data isn't an actual value, but instead is a pointer to a location holding the given value (4 bytes on an x86 processor). So, let's say in memory location my_location you have the number 20 stored. If the prototype said to pass an int, you would use direct addressing mode and do pushl my_location. However, if the prototype said to pass an int *, you would do pushl $my_location - an immediate mode push of the address that the value resides in. In addition to indicating the address of a single value, pointers can also be used to pass a sequence of consecutive locations, starting with the one pointed to by the given value. This is called an array.

struct

• A struct is a set of data items that have been put together under a name. For example you could declare:

   struct teststruct {  int a;  char *b; }; 

  and any time you ran into struct teststruct you would know that it is actually two words right next to each other, the first being an integer, and the second a pointer to a character or group of characters. You never see structs passed as arguments to functions. Instead, you usually see pointers to structs passed as arguments. This is because passing structs to functions is fairly complicated, since they can take up so many storage locations.

typedef

• A typedef basically allows you to rename a type. For example, I can do typedef int myowntype; in a C program, and any time I typed myowntype, it would be just as if I typed int. This can get kind of annoying, because you have to look up what all of the typedefs and structs in a function prototype really mean. However, typedefs are useful for giving types more meaningful and descriptive names.

  Compatibility Note: The listed sizes are for intel-compatible (x86) machines. Other machines will have different sizes. Also, even when parameters shorter than a word are passed to functions, they are passed as longs on the stack.

That's how to read function documentation. Now, let's get back to the question of how to find out about libraries. Most of your system libraries are in /usr/lib or /lib. If you want to just see what symbols they define, just run objdump -R FILENAME where FILENAME is the full path to the library. The output of that isn't too helpful, though, for finding an interface that you might need. Usually, you have to know what library you want at the beginning, and then just read the documentation. Most libraries have manuals or man pages for their functions. The web is the best source of documentation for libraries. Most libraries from the GNU project also have info pages on them, which are a little more thorough than man pages.

^[2]The reason that parameters are pushed in the reverse order is because of functions which take a variable number of parameters like printf. The parameters pushed in last will be in a known position relative to the top of the stack. The program can then use these parameters to determine where on the stack the additional arguments are, and what type they are. For example, printf uses the format string to determine how many other parameters are being sent. If we pushed the known arguments first, you wouldn't be able to tell where they were on the stack.