.tdata section allocated in weird place when writing binaries

[dktapps]

Describe the bug I'm using lief to rewrite the Linux version of Minecraft Bedrock Dedicated Server, in order to convert the provided debugging symbols into linkable dynamic symbols for modding.

I use the following script to do it:

import lief

lib_symbols = lief.parse("bedrock_server_symbols.debug")
for s in lib_symbols.exported_symbols:
#    print(s)
    lib_symbols.add_dynamic_symbol(s)
lib_symbols.write("bedrock_server_symbols_test.debug")

and then strip --strip-debug bedrock_server_symbols_test.debug (otherwise it will segfault; not sure if that's my fault or not). strip, gdb and other tools report that .tdata is allocated in a weird place after the modification; I haven't noticed this causing any issues yet, but no doubt it will crop up at some point.

strip output:

strip --strip-debug bedrock_server_symbols_test.debug
strip: st5VQNfj: section `.tdata' can't be allocated in segment 1
LOAD: .init_array .fini_array .data.rel.ro .dynamic .got .got.plt .data .tdata .tbss .bss

To Reproduce Run the above Python script on the Minecraft Bedrock Dedicated Server bedrock_server_symbols.debug, and then run strip on the output (or gdb).

Expected behavior No error should be produced: all the sections should be allocated in suitable places.

Environment (please complete the following information):

• System and Version : Linux DYLANS-XPS-17 5.10.60.1-microsoft-standard-WSL2 #1 SMP Wed Aug 25 23:20:18 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
• Target format ELF
• LIEF commit version: 0.11.5-37bc2c9

[romainthomas]

Could you try with the master branch? (it fixes some ELF modification issues)

[dktapps]

Hi, with the master branch the modified binary immediately segfaults. It does this with two different binaries I tried.

https://minecraft.azureedge.net/bin-linux/bedrock-server-1.18.11.01.zip

make output:

strip --strip-debug bedrock_server_symbols.debug
python3 export-symbols.py
strip --strip-debug bedrock_server_symbols_test.debug
strip: stomyHgV: section .interp lma 0x400270 adjusted to 0x4002a8
strip: stomyHgV: section .note.ABI-tag lma 0x40028c adjusted to 0x4002c4

gdb

(gdb) run
Starting program: /home/user/bds-test/mapping/bedrock_server_symbols_test.debug

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7fee60a in ?? () from /lib64/ld-linux-x86-64.so.2
(gdb) bt
#0  0x00007ffff7fee60a in ?? () from /lib64/ld-linux-x86-64.so.2
#1  0x00007ffff7fd1ebf in ?? () from /lib64/ld-linux-x86-64.so.2
#2  0x00007ffff7febc3b in ?? () from /lib64/ld-linux-x86-64.so.2
#3  0x00007ffff7fd104c in ?? () from /lib64/ld-linux-x86-64.so.2
#4  0x00007ffff7fd0108 in ?? () from /lib64/ld-linux-x86-64.so.2
#5  0x0000000000000001 in ?? ()
#6  0x00007fffffffe4bf in ?? ()
#7  0x0000000000000000 in ?? ()
(gdb)

[romainthomas]

Ack I'll check

[romainthomas]

Ok got it. Strip does not work very well with a binary built by LIEF as strip removes the content that has been added by LIEF (at the end of the binary).

First I would suggest you to perform add_dynamic_symbol only on the debug symbols that are exported:

 lib_symbols = lief.parse("bedrock_server_symbols.debug")
for s in filter(lambda e: e.exported, lib_symbols.static_symbols):
    lib_symbols.add_dynamic_symbol(s)

As a workaround for the non-working strip, I would suggest to use the LIEF's builtin strip()[^1]

In the end, this modification should not segfault the binary:

import lief
lib_symbols = lief.parse("bedrock_server_symbols.debug")
for s in filter(lambda e: e.exported, lib_symbols.static_symbols):
    lib_symbols.add_dynamic_symbol(s)
lib_symbols.strip()
lib_symbols.write("foo.debug")

You can also observe that the binary is correctly stripped with file

file ./foo.debug
./foo.debug: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 3.2.0, stripped

[^1]: Working well on the master branch. The version 0.11.5 misses to correctly handle this modification

[dktapps]

Thanks. The reason I was using strip is because the output binary produced would segfault immediately. Stripping it made it work as expected. I'll try your suggested code.

[dktapps]

The suggested code appears to work flawlessly. Thank you!