Lecture 013

Reason for Linking

1. Modularity / Design (using common library)
2. Efficiency

Static Linking

Dynamic Linking

Linking Steps

Symbol Resolution

Symbol Table: store symbol definition

• content

  • name
  • size
  • location of symbol

• one symbol can only have one definition

• types of symbol

  • global variable name
  • function name (include main and library header)
  • (maybe custom type and struct?)

• does not contain

  • built-in type

Executable and Link-able Format (ELF)

Different Types of ELF

Relocatable Object File (.o)

• code and data to be combined with other relocatable file

• each .o is produced from exactly one .c file

Executable Object File (a.out)

• executable, copied directly into memory

Shared Object File (.so)

• relocatable object file that can be loaded into memory and linked dynamically at load time or run-time. (called Dynamic Link Libraries DLL by Windows)

ELF Object Storage

Symbols

Global Symbols

• defined in a module that can be referenced by other modules

• non-static function, non-static global variable

External Symbols

• global symbols that are referenced but not defined by current module

• like external function printf

Local Symbols

• symbols defined and only referenced in current module

• static function, global variables, static variables either in function or global

Not in symbol table:

• local variable in function

• function arguments

• string in function may be in symbol table, depending on compiler

static variable: exist in life time of program, but accessible only in function's scope

Step 1: Symbol Resolution

Strong Symbol: procedures and initialized globals Weak Symbol: uninitialized globals, or declared with extern.

• if it is declared extern, symbol has to be resolve by a strong symbol for linking.

• extern can check type mismatch

Linking Rules

Linking Rules

1. Multiple strong symbols are not allowed.
   • otherwise produce linker error
2. Given a strong symbol and multiple weak symbols, the symbol is defined by strong symbol.
   • reference to weak symbol resolved to strong symbol
3. If there are multiple weak symbols, pick an arbitrary one
4. Linker does not do type check (including function argument type)

Good Practice With Global Variable

• use static if can

• initialize it

• use extern if you need to reference external variable

Header File

Header file: contain weak symbol that can be shared by multiple other files

• files that include weak symbol of header file must reference with the same type (type checking using include)

• if there are strong symbols in header file, since header file replace everything in normal file, there will be strong symbol conflict in multiple files

Step 2: Relocation

Before relocation, global shared variables are not placed in byte code

• 

After relocation, zero values are replaced

• 

Libraries

Static Library

Static Libraries (.a archive files)

• concatenating all related .o files into a single file with an index (archive)

• resolve reference by looking for the symbols in one or more archives

• 

• 

Linking with Static Library

• library provide .h file so that we don't have to write weak symbols. (libc.a is compiled with gcc automatically)

• 

• linking process in command line by order

  • scan .o and .a files for unresolved reference, building unresolved list
  • try to resolve each one in the list
  • error if unresolved reference at the end of scan
  • order matter, so you should put .o file in front and library file back

  • 

Disadvantage

• duplicate stored executable

• duplicate when load into memory

• Minor bug fix require rebuild and relink

Shared Library

Dynamic Linking:

• link when executable first loaded and run (load-time linking)

  • in linux, ld-linux.so handles dynamic linking
  • libc.so usually dynamically linked

• link after program has begun running

  • call to dlopen()
  • in distributing software, web servers, plugins

Region in file specify dynamic libraries

• .interp section: which dynamic linker to use (ld-linux.so)

• .dynamic section: name of dynamic libraries

• ldd a.out command see dynamically resolved libraries