When / How does Linux load shared libraries into address space?
Solution 1
Libraries are loaded by ld.so
(dynamic linker or run-time linker aka rtld, ld-linux.so.2
or ld-linux.so.*
in case of Linux; part of glibc). It is declared as "interpreter" (INTERP; .interp
section) of all dynamic linked ELF binaries. So, when you start program, Linux will start an ld.so
(load into memory and jump to its entry point), then ld.so
will load your program into memory, prepare it and then run it. You can also start dynamic program with
/lib/ld-linux.so.2 ./your_program your_prog_params
ld.so
does an actual open
and mmap
of all needed ELF files, both ELF file of your program and ELF files of all neeeded libraries. Also, it fills GOT and PLT tables and does relocations resolving (it writes addresses of functions from libraries to call sites, in many cases with indirect calls).
The typical load address of some library you can get with ldd
utility. It is actually a bash script, which sets a debug environment variable of ld.so (actually LD_TRACE_LOADED_OBJECTS=1
in case of glibc's rtld) and starts a program. You even can also do it yourself without needs of the script, e.g. with using bash easy changing of environment variables for single run:
LD_TRACE_LOADED_OBJECTS=1 /bin/echo
The ld.so
will see this variable and will resolve all needed libraries and print load addresses of them. But with this variable set, ld.so
will not actually start a program (not sure about static constructors of program or libraries). If the ASLR feature is disabled, load address will be the same most times. Modern Linuxes often has ASLR enabled, so to disable it, use echo 0 | sudo tee /proc/sys/kernel/randomize_va_space
.
You can find offset of system
function inside the libc.so
with nm
utility from binutils. I think, you should use nm -D /lib/libc.so
or objdump -T /lib/libc.so
and grep output.
Solution 2
"Go right to the source and ask the horse..."
Drepper - How To Write Shared Libraries
Must-read documentation for Linux library writers. Explains the mechanics of loading in some detail.
Solution 3
The nm
command, used on libc.so
, will show you the location of the system
symbol in libc.so
. However, if ASLR is enabled, the address libc.so
is loaded at, and thus the final address of system
will vary randomly each time your program is run. Even without ASLR, you'll need to determine the address libc.so
gets loaded at and offset the address of system
by that amount.
Solution 4
If you just want the address of a function while not hardcoding the name, you could dlopen()
the main program:
void *self = dlopen(NULL, RTLD_NOW);
dlsym(self, "system"); // returns the pointer to the system() function
If you just want the address of a function of which you know the name at compile-time, simply use void *addr = &system;
Related videos on Youtube
Ryan
Updated on January 25, 2020Comments
-
Ryan over 4 years
My question is the following:
When is the address of shared objects specified in programs? During linking? Loading? If I wanted to find the memory address of the
system
command inside oflibc
inside of my program I could find it easily ingdb
, but what if I don't want to bring the program into a debugger?Could this address change from run to run? Are there any other static analysis tool that will allow be to view where libraries or functions will be loaded into this program's memory space when run?
EDIT: I want this information outside of the program (ie. using utilities like
objdump
to gather information)-
Ben Voigt about 13 yearsand then there's
prelink
, which changes the order considerably.
-
-
Ryan about 13 yearsplease see edit in OP, but definitely leave this answer as it is helpful for another variation of the potentially vague title. (how you answered it)
-
R.. GitHub STOP HELPING ICE about 13 yearsIt's possible that neither will do what OP wants, since
system
is probably resolved to a PLT entry in the main program image that performs the actual jump into the shared library. -
Ryan about 13 yearsThis is pretty much exactly what I wanted - I'm assuming the best way to determine the offset of system within libc.so is to use nm again with the debugging symbols installed? Or is there an easier / more robust manner to do this.
-
osgx about 13 years@Ryan, nm needs no debug symbols, it can read a symbol table directly (which is used by
ld.so
). -
Ryan about 13 yearswonderful information, thank you. Do you know of any good articles that explain how this process works (generating GOT / PLT tables), or would googling yield sufficient results?
-
Ryan about 13 years@osgx so then would my original question be correct, subtracting the debugging symbols?
-
osgx about 13 yearsIf you want to call a
system
with absolute address, you can do it without using a GOT and PLT tables. In my point, the best googling forld.so
is codesearch: google.com/… -
Ryan about 13 yearsYes, I know you can do it without using GOT and PLT, it was mere curiosity on my part! :)
-
osgx about 13 yearsmay be,... the book on linkers will help? iecc.com/linker/linker10.html Also there is a blog of author of
gold
linker and elf part of binutils ld linker www.airs.com/blog/archives/41 -
osgx about 13 yearspdf of some blogposts from airs.com www.cs.pitt.edu/~abraham/CS0449/slides/Linking.pdf
-
Matt Joiner about 13 yearsThis was very informative! I didn't know that much about ld.so and ldd!
-
Employed Russian almost 7 yearsThis answer is incorrect in that
ld.so
does not load the main program, the kernel does. The kernel also doesn't look at any sections (which could be stripped completely), it finds the interpreter inPT_INTERP
segment. -
osgx almost 7 years@EmployedRussian, Can you post more correct answer? I'm not expert in this and you is.