{simple,shoddy,smart} minsize-oriented linker
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
PoroCYon 429ba14e8c I'm an idiot 2 years ago
ld more size crunching, but not completely perfect 3 years ago
rt fix dnload version 3 years ago
src I'm an idiot 2 years ago
test fixes. dnload mode should now be usableish 3 years ago
.gitignore src->rt; smol*.py -> src; add optional 16-bit hash version 3 years ago
LICENSE add LICENSE 3 years ago
Makefile 32-bit version working (not yet the dnload version) 3 years ago
README.md fix importing weak symbols 3 years ago
rmtrailzero.py get rid of a few more bytes 3 years ago
smoldd.py fix crash when a hash's LSbyte is 0 3 years ago



Shoddy minsize-oriented linker

PoC by Shiz, bugfixing and 64-bit version by PoroCYon.


./smol.py -lfoo -lbar input.o... smol-output.asm
nasm -I src/ [-Doption ...] -o nasm-output.o smol-output.asm
ld -T ld/link.ld --oformat=binary -o output.elf nasm-output.o input.o...
# or cc -T ld/link.ld -Wl,--oformat=binary -o output.elf nasm-output.o input.o...
  • USE_INTERP: Include an interp segment in the output ELF file. If not, the dynamic linker must be invoked explicitely! (You probably want to enable this.) Costs the size of a phdr plus the size of the interp string.
  • ALIGN_STACK: 64-bit only: realign the stack so that SSE instructions won't segfault. Costs 1 byte.
  • USE_NX: Don't use RWE segments at all. Not very well tested. Costs the size of 1 phdr, plus some extra stuff on i386. Don't forget to pass -n to smol.py as well.
  • USE_DL_FINI: keep track of the _dl_fini function and pass it to your _start. Costs 2 bytes, plus maybe a few more depending on how it's passed to __libc_start_main.
  • USE_DT_DEBUG: retrieve the struct link_map from the r_debug linker data (which is placed at DT_DEBUG at startup) instead of exploiting data leakage from _dt_start_user. Might be more compatible and compressable, but strictly worse size-wise by 10 (i386) or 3 (x86_64) bytes.
  • SKIP_ENTRIES: skip the first two entries of the struct link_map, which represent the main binary and the vDSO. Costs around 5 bytes.
  • USE_DNLOAD_LOADER: use the symbol loading mechanism as used in dnload (i.e. traverse the symtab of the imported libraries). Slightly larger, but probably better compressable and more compatible with other libcs and future versions of glibc.
  • NO_START_ARG: don't pass the stack pointer to _start as the first arg. Will make it unable to read argc/argv/environ, but gives you 3 bytes.
  • SKIP_ZERO_VALUE: skip a Sym with a st_value field of 0. If this isn't enabled, weak symbols etc. might be imported instead of the real ones, causing breakage. Many libraries don't have weak symbols at all, though. Costs 4 (i386) or 5 (x86_64) bytes.
usage: smol.py [-h] [-m TARGET] [-l LIB] [-L DIR] [--nasm NASM] [--cc CC]
               [--scanelf SCANELF] [--readelf READELF]
               input [input ...] output

positional arguments:
  input                 input object file
  output                output nasm file

optional arguments:
  -h, --help            show this help message and exit
  -m TARGET, --target TARGET
                        architecture to generate asm code for (default: auto)
  -l LIB, --library LIB
                        libraries to link against
  -L DIR, --libdir DIR  directories to search libraries in
  --nasm NASM           which nasm binary to use
  --cc CC               which cc binary to use
  --scanelf SCANELF     which scanelf binary to use
  --readelf READELF     which readelf binary to use
  -n, --nx              Use NX (i.e. don't use RWE pages). Costs the size of
                        one phdr, plus some extra bytes on i386. Don't forget
                        to pass -DUSE_NX to the assembly loader as well!

A minimal crt (and _start funcion) are provided in case you want to use main.


smoldd.py is a script that tries to resolve all symbols from the hashes when imported by a smol-ified binary. This can thus be used to detect user mistakes during dynamic linking. (Think of it as an equivalent of ldd, except that it also checks whether the imported functions are present as well.)

NOTE: smoldd.py currently doesn't support 64-bit binaries anymore, as there's currently no (good) way of retrieving the symbol hash table anymore.

Internal workings

smol.py inspects the input object files for needed library files and symbols. It then outputs the list of needed libraries, hashes of the needed symbols and provides stubs for the external functions. This is then combined with a custom-made, small ELF header and 'runtime linker' which resolves the symbols (from the hashes) so that the function stubs are usable.

The runtime linker uses an unorthodox way of resolving the symbols (which only works for glibc): on both i386 and x86_64, the linker startup code (_dl_start_user) leaks the global struct link_map to the user code: on i386, a pointer to it is passed directly through eax:

# (eax, edx, ecx, esi) = (_dl_loaded, argc, argv, envp)
movl _rtld_local@GOTOFF(%ebx), %eax
## [ boring stuff... ]
pushl %eax
# Call the function to run the initializers.
call _dl_init
## eax still lives thanks to the ABI and calling convention
## [ boring stuff... ]
# Jump to the user's entry point.
jmp *%edi
## eax contains the pointer to the link_map!

On x86_64, it's a bit more convoluted: the contents of _rtld_local is loaded into rsi, but because of the x86_64 ABI, the caller isn't required to restore that register. However, due to the call instruction, a pointer to the instruction after the call will be placed on the stack. And thus, at _start, that pointer will be available at rsp - 8. Then, the offset to the "load from _rtld_local"-instruction can be calculated, and the part of the instruction which contains the offset to _rtld_local, from the instruction after the load (of which the address is now also known), can be read, and thus the location and contents of that global variable are available as well.

When using DT_DEBUG, a different mechanism is used to take hold of the struct link_map: on program startup, ld.so will place a pointer to its debug data in the value of the DT_DEBUG key-value-pair. In glibc, this is the r_debug datatype. The second field of that type, is a pointer to the root struct link_map.

Now the code continues with walking the "import tables" for the needed libraries (which already have been automatically parsed by ld.so), looks though their hash tables for the hashes of the imported symbols, gets their addresses, and replaces the hashes in the table with the function addresses.

However, because the struct link_map can change between glibc versions, especially the size of the l_info field (a fixed-size array, the DT_*NUM constants tend to change every few versions). To remediate this, one can note that the l_entry field comes a few bytes after l_info, that the root struct link_map is the one of the main executable, and that the contents of the l_entry field is known at compile-time. Thus, the loader scans the struct for the entry point address, and uses that as an offset for the 'far fields' of the struct link_map. ('Near' fields like l_name and l_addr are resp. 8 and 0, and will thus pretty much never change.)


auld alrj blackle breadbox faemiyah gib3&tix0 las leblane parcelshit unlord