ft_nm

A reimplementation of the GNU nm utility that displays symbol information from ELF binary files.

What It Does

ft_nm parses ELF (Executable and Linkable Format) binary files and displays their symbol table in a format compatible with the GNU nm tool. It supports both 32-bit and 64-bit ELF binaries and provides various filtering and sorting options.

For each file, the program:

1. Memory-maps the entire file content using mmap(2)
2. Parses the ELF header to determine file format (32-bit vs 64-bit)
3. Locates the symbol table (.symtab) section
4. Extracts all symbols with their addresses, types, and names
5. Filters symbols based on command-line options
6. Sorts symbols alphabetically (unless disabled)
7. Prints the results in nm-compatible format

Implementation Details

ftlist - The Linked List Abstraction

The implementation leverages a custom linked list library (t_list from libft) that provides significant advantages:

• Separation of Concerns: Operations like extraction, filtering, and sorting are cleanly separated using iterator functions (ft_lstiter, ft_lstiter_param)
• Reusable Operations: Generic list operations like ft_lstsort, ft_lstfilter, ft_lstadd_back allow for modular, testable code
• Custom Comparators: The sorting algorithm accepts function pointers, enabling flexible comparison logic (alphanumeric, case-insensitive, reverse)
• Memory Management: Centralized cleanup with ft_lstclear and custom destructors prevents memory leaks
• Functional Style: Operations like ft_lstmap and ft_lstfilter_param enable a functional programming approach

This design allows the main nm() function to be a simple pipeline:

extract_nm_symbols() → filter_symbols() → sort_symbols() → print_nm_symbols()

Section Name Resolution

When a symbol's st_name field is 0 (typically for STT_SECTION type symbols), the implementation retrieves the section name from the section header string table (.shstrtab) rather than the regular string table (.strtab).

This is done by:

1. Identifying the symbol as STT_SECTION type
2. Using the symbol's st_shndx to index into the section header array
3. Reading the section's sh_name field
4. Looking up that offset in .shstrtab instead of .strtab

This ensures that section symbols like .text, .data, .bss are properly named.

Address Formatting: 32-bit vs 64-bit

The symbol address printing differs based on the binary architecture:

• 64-bit binaries: Addresses are printed as 16 hexadecimal characters (e.g., 00000000004005a0)

  • Format: %016lx

• 32-bit binaries: Addresses are printed as 8 hexadecimal characters (e.g., 08048400)

  • Format: %08lx

For undefined symbols (SHN_UNDEF), no address is printed—instead, the appropriate number of spaces is displayed (17 spaces for 64-bit, 9 spaces for 32-bit) to maintain column alignment.

Supported Options

• -a, --debug-syms: Display debugger-only symbols
• -g, --extern-only: Display only external symbols
• -u, --undefined-only: Display only undefined symbols
• -r, --reverse-sort: Reverse the sorting order
• -p, --no-sort: Do not sort symbols (display in original order)
• -h, --help: Display usage information

Symbol Sorting

By default, symbols are sorted alphabetically by name using an alphanumeric comparison that:

1. Ignores non-alphanumeric characters during comparison
2. Performs case-insensitive matching
3. Falls back to strcmp for identical alphanumeric sequences

The sorting behavior for symbols with identical names is undefined (matches GNU nm behavior).

References

• System V ABI x86-64 Specification
• man elf, man 5 elf
• man nm, man readelf, man objdump