Difference between .dynamic .dynsym and .dynstr in an ELF executable

问题

My preliminary knowledge is that:

• .dynamic contains libraries that the executable needs to load
• .dynsym contains external symbols such as setsockopt@GLIBC_2.0
• .dynstr contains strings of function requirements

Overall, I am a bit confused as to how these sections work together to create a binary - specifically .dynsym and .dynstr. So my question is two-fold. Are my statements above correct? If so, how do these three sections work together to create a binary?

回答1:

Are my statements above correct?

The .dynsym section contains a set of fixed-length records of type Elf32_Sym or Elf64_Sym.

Since these are fixed length entries, they can not by themselves describe arbitrary length symbols (strings) that the binary exports or imports.

Therefore these entries don't contain the strings. Instead, they contain an offset into .dynstr (in the .st_name field), and the symbol name is found at this offset.

So it's not true that ".dynsym contains setsockopt@GLIBC_2.0" and that ".dynstr contains strings of function requirements" (whatever that last statement means).

The .dynsym contains an Elf32_Sym or an Elf64_sym describing an imported symbol setsockopt, and referencing the offset of "setsockopt" string in the .dynstr section.

Likewise, ".dynamic contains libraries that the executable needs to load" is false -- the section doesn't contain any libraries.

It contains fixed length entries of Elf64_Dyn or Elf32_Dyn, some of which (such as ones with .d_tag == DT_NEEDED or DT_RPATH) may reference strings from .dynstr via their offsets. The dynamic loader interprets these entries in certain way -- for DT_NEEDED as "must load this other library", for DT_RPATH as "must search these colon-separated paths", etc.

来源：https://stackoverflow.com/questions/53156275/difference-between-dynamic-dynsym-and-dynstr-in-an-elf-executable