Home

Awesome

insert_dylib

This is a fork of the command line utility for inserting a dylib load command into a Mach-O (https://github.com/Tyilo/insert_dylib). By making use of the https://github.com/LeanVel/cctools project, this tool has being ported to LINUX.

It does the following (to each arch if the binary is fat):

Installation

$ git clone https://github.com/LeanVel/insert_dylib
$ cd insert_dylib
$ sudo ./Install.sh

Usage

Usage: insert_dylib dylib_path binary_path [new_binary_path]
Option flags: --inplace --weak --overwrite --strip-codesig --no-strip-codesig --all-yes

insert_dylib inserts a load command to load the dylib_path in binary_path.

Unless --inplace option is specified, insert_dylib will produce a new binary at new_binary_path.
If neither --inplace nor new_binary_path is specified, the output binary will be located at the same location as the input binary with _patched prepended to the name.

If the --weak option is specified, insert_dylib will insert a LC_LOAD_WEAK_DYLIB load command instead of LC_LOAD_DYLIB.

Example

$ cat > test.c
int main(void) {
	printf("Testing\n");
	return 0;
}
^D
$ clang test.c -o test &> /dev/null
$ insert_dylib /usr/lib/libfoo.dylib test
The provided dylib path doesn't exist. Continue anyway? [y/n] y
Added LC_LOAD_DYLIB to test_patched
$ ./test
Testing
$ ./test_patched
dyld: Library not loaded: /usr/lib/libfoo.dylib
  Referenced from: /Users/Tyilo/./test_patched
  Reason: image not found
Trace/BPT trap: 5

otool diff between original and patched binary

$ diff -u <(otool -hl test) <(otool -hl test_patched)
--- /dev/fd/63	2014-07-30 04:08:40.000000000 +0200
+++ /dev/fd/62	2014-07-30 04:08:40.000000000 +0200
@@ -1,7 +1,7 @@
-test:
+test_patched:
 Mach header
       magic cputype cpusubtype  caps    filetype ncmds sizeofcmds      flags
- 0xfeedfacf 16777223          3  0x80          2    16       1296 0x00200085
+ 0xfeedfacf 16777223          3  0x80          2    17       1344 0x00200085
 Load command 0
       cmd LC_SEGMENT_64
   cmdsize 72
@@ -231,3 +231,10 @@
   cmdsize 16
   dataoff 8296
  datasize 64
+Load command 16
+          cmd LC_LOAD_DYLIB
+      cmdsize 48
+         name /usr/lib/libfoo.dylib (offset 24)
+   time stamp 0 Thu Jan  1 01:00:00 1970
+      current version 0.0.0
+compatibility version 0.0.0

--weak option

$ insert_dylib --weak /usr/lib/libfoo.dylib test test_patched2
The provided dylib path doesn't exist. Continue anyway? [y/n] y
Added LC_LOAD_WEAK_DYLIB to test_patched2
$ ./test_patched2
Testing

Removing code signature

To remove the code signature it is enough to delete the LC_CODE_SIGNATURE load command and fixup the mach header's ncmds and sizeofcmds, assuming it is the last load command.

However if you just do this codesign_allocate (used by codesign and ldid) will fail with the error:

.../codesign_allocate: file not in an order that can be processed (link edit information does not fill the __LINKEDIT segment):

To fix this insert_dylib assumes that the code signature that LC_CODE_SIGNATURE is in the end of the __LINKEDIT segment and the that the segment is in the end of the architectures slice.

It then truncate that slice to remove the code signature part of the __LINKEDIT segment. It also updates the LC_SEGMENT (or LC_SEGMENT64) load command for the __LINKEDIT segment from the new file size. If the binary is fat we also update the size and we might also move the slice and so the offset should also be updated.

After removing the code signature from the __LINKEDIT segment, the last thing in that segment is typically the string table. As the code signature seems to be aligned by 0x10, and so after removing the code signature, nothing points to the padding at the end of the segment, which codesign_allocate doesn't like either. To fix this we just increase the size of the string table in the LC_SYMTAB command so it also includes the padding.

Todo