Google Summer of Code 2021 Summary
GSoC 2021 is officially finished and we are happy to congratulate all 3 participants for passing the program and completing the most important parts of their tasks. It brought us some long-needed code cleanup and user-visible changes in the analysis and binary/heap parsing. See what students wrote themselves:
08A: Refactoring ELF binaries loading
This summer I have been doing the GSoC for Rizin. The subject of the GSoC was to refactor and improve how elf binaries are loaded by Rizin.
I have added support for the elf hash table and gnu hash table. Those 2 data structures are used to deduct the number of dynamic symbols in the file, which replaced the old way of doing it (assuming that the data is a symbol until there is an error).
Moreover, I have changed the source of trust used to load symbols’ versions (from sections information to dynamic section’s information). So Rizin is now able to read symbols’ versions even if there is no section.
> rz-bin -V bins/elf/analysis/clark
WARNING: Invalid section header (check array failed).
Version symbols has 9 entries:
Addr: 0x080482c2 Offset: 0x000002c2
0x00000000: 0 (*local*)
0x00000001: 2 (GLIBC_2.0)
0x00000002: 2 (GLIBC_2.0)
0x00000003: 0 (*local*)
0x00000004: 2 (GLIBC_2.0)
0x00000005: 2 (GLIBC_2.0)
0x00000006: 2 (GLIBC_2.0)
0x00000007: 2 (GLIBC_2.0)
0x00000008: 1 (*global*)
Version need has 1 entries:
Addr: 0x080482d4 Offset: 0x000002d4
0x000002d4: Version: 1 File: libc.so.6 Cnt: 1
0x000002e4: Name: GLIBC_2.0 Flags: none Version: 2
There was a hard-coded maximum length for all string found in any elf string table. This limitation was removed and some small check of the string table integrity were added.
> rizin bins/elf/long-symbol.elf
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
-- Add custom Have you setup your ~/.rizinrc today?
[0x00001040]> is~AAA
28 0x00001139 0x00001139 GLOBAL FUNC 15 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
The main problem with how symbols and imports were loaded, was their mutual dependency during the loading phase. So both processes were split and heavily refactored. As a side effect, an old bug in the symbols loading was fixed.
The call to the function system is correctly identified:
> rizin bins/elf/analysis/phdr-override
WARNING: The segment 3 at 0x774 seems to be invalid.
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
-- Change your fortune types with 'e cfg.fortunes.file=fun,tips' in your ~/.rizinrc
[0x004003f0]> s main
[0x004004e6]> af
[0x004004e6]> pdf
┌ int main (int argc, char **argv, char **envp);
│ ; var int64_t var_10h @ rbp-0x10
│ ; var int64_t var_4h @ rbp-0x4
│ ; arg int argc @ rdi
│ ; arg char **argv @ rsi
│ 0x004004e6 push rbp
│ 0x004004e7 mov rbp, rsp
│ 0x004004ea sub rsp, 0x10
│ 0x004004ee mov dword [var_4h], edi ; argc
│ 0x004004f1 mov qword [var_10h], rsi ; argv
│ 0x004004f5 mov rax, qword [var_10h]
│ 0x004004f9 add rax, 8
│ 0x004004fd mov rax, qword [rax]
│ 0x00400500 mov rdi, rax
│ 0x00400503 mov eax, 0
│ 0x00400508 call sym.imp.system ; int system(const char *string)
│ 0x0040050d mov eax, 0
│ 0x00400512 leave
└ 0x00400513 ret
During the loading phase, sections and segment information checks have been added to verify the integrity of the data. Those checks are stricter than the elf loader. So 3 configurations variable were implemented to allow the user to customize how segments and sections are loaded.
> rizin bins/elf/analysis/phdr-override
WARNING: The segment 3 at 0x774 seems to be invalid.
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
-- Press 'C' in visual mode to toggle colors
[0x004003f0]>
> rizin -e elf.checks.segments=false bins/elf/analysis/phdr-override
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
WARNING: Neither hash nor gnu_hash exist. Falling back to heuristics for deducing the number of dynamic symbols...
-- Add colors to your screen with 'e scr.color=X' where 1 is 16 colors, 2 is 256 colors and 3 is 16M colors
[0x004003f0]>
There is still a lot of work to do, specially on the elf plugin interface. If you want to follow the update on this, you can use this link: Refactoring the elf plugin interface
In conclusion, the GSoC was an incredible source of motivation to contribute to the Open Source community. And it helped me improve my knowledge of elf internals. I would like to thank my mentors Anton Kochkov and Florian Märkl for their help during the GSoC.
Pulak: Heap viewer for Cutter
Hi, I am Pulak Malhotra. Over the past few months, I participated in GSoC with the Rizin organization. My main contributions revolve around the heap parsing code for Rizin and the GUI implementation of heap viewer for Cutter.
The initial work started with improving the output format of the dmh
family of commands. I made them much more readable, taking inspiration from gdb gef
. I added a new command, dmhd
, which prints concise information about different bins of a given arena. I also refactored and rewrote a significant part of the Glibc heap codebase, making it more modular and maintainable, including porting it to the new shell.
I added new Rizin API calls and used them in Cutter to implement the GUI version of the heap viewer. Heap viewer in Cutter has many features, like getting a list of heap chunks in an arena, editing the heap chunks, getting information about bins in the arena, and visualizations for linked lists of the bins. I encourage everyone to give it a try in their next heap exploitation hack!
After Glibc heap, I made some contributions towards the windows heap and windows heap widget. Some of the changes have been merged, like the Rizin API and the new shell port. I’ll try my best to ship the other modifications to production soon.
GSoC was one of my first experiences working on a real-world project, and I learned and grew a lot. I want to give special thanks to my mentors Yossizap and Megabeets, and the Rizin community members XVilka, Ret2libc, Deroad, Gustavolcr, and Thestr4ng3r, who were always there to answer my questions and review and give feedback for my PRs.
Aswin: Support for CPU and platform profiles
Hello everybody!
I’m Aswin and this is a brief summary about the work I did on the summer of 2021 with Rizin on adding support for CPU and platform profiles. Rizin previously relied upon manually writing code for adding a new CPU or an IO port and it was a bit tedious to handle them all and it was not user friendly. Providing a level of abstraction in handling this entropy in embedded systems by adding support for editable CPU and platform profiles was the goal of this project.
After getting accepted, the first thing I did was to remove the existing implementation of
RzSyscallPorts
- the module which took care of the architecture and CPU specific system
registers. Here, I made two new modules: RzSysregsDB
and RzSysregItem
to make this happen.
RzSysregsDB
just housed a hashtable which paired the address of the port and an RzSysregItem
which contained the comment, type and all the other information related it.
Then, I started working on CPU profiles. The whole idea of CPU profiles is to store all the CPU specifics in one file, parse it and use it
at places like analysis, emulation and wherever it’s needed. Inside CPU profiles,
we store information like size of the ROM, size of the RAM CPU and other things and they
are parsed and stored into various data structures inside RzArchProfile
, where RzArchTarget
houses the name of the CPU and architecture and a pointer to RzArchProfile
. Information about
the CPU IO registers and Extended IO registers can also be added in CPU profiles. During the
analysis loop, they are added as flags (labels) at their corresponding offsets. A feature to map the ROM as sections (iS
) were also added with it.
This is how the IO and extended IO registers are defined in the SDB files:
SPH=reg
SPH.address=0x3e
SPH.comment=Stack higher bits SP8 SP10
After that, I added support for platform profiles. Platform profiles were introduced to handle the platform specific differences. These files contains the name, offset and a short description of each port or register, which are parsed and added as flags and comments. Support for one platforms like BCM2835, which one of the Raspberry Pi runs on, BCM2711 and OMAP 3430 were added along with the x86 IO ports were added subsequently.
A new configuration variable asm.platform
was also added
to choose the platform profile. This will let the user choose the name of the profile they want to load and Rizin will load the
profile based upon the CPU and the architecture that the user have previously set. For that, I added a new variable platforms
to RzAsmPlugin
which will hold the list of all supported platforms of that architecture.
Platform Profiles also follow a format similar to the CPU profiles that you saw earlier. Here’s an excerpt BCM 2835’s platform profile:
AUX_MU_IER_REG=name
AUX_MU_IER_REG.address=0x7e215044
AUX_MU_IER_REG.comment=Mini UART Interrupt Enable
AUX_MU_IIR_REG=name
AUX_MU_IIR_REG.address=0x7e215048
AUX_MU_IIR_REG.comment=Mini UART Interrupt Identify
Then, I worked on porting uefi_r2 - a tool used to analyze UEFI modules to Rizin.
This tool works by analyzing the firmware using Rizin’s RzAnalysis
utilities and inspecting its functions,
strings and other particulars - for example, while searching for the UEFI GUIDs inside the analyzed strings.
Here, the tool is a Python package and all the interaction with rizin
is done through rz-pipe
’s Python module.
Overall, this was not particularly challenging but it was indeed very informative. UEFI is insanely complex!
Later, I continued to work on improving the SVD parser plugin I had started making during the microtask. SVD files are files containing information about a device’s peripherals, MMIO registers and other particulars. They are usually made by the manufacturer. This plugin would load the data from SVD file to Rizin mainly the registers’ name, size, base address and its offset and adds them as flags and comments.
I would like to thank my mentors xvilka and deroad for their guidance. I was regularly in touch with them and they were constantly trying make sure that everything was going smooth.
Also kudos to all the folks at #Rizin-dev
, #gsoc-2021
and the other channels where my questions were answered.