Files
Embedded-Hacking/WEEK06/WEEK06-04.md
2026-03-19 15:01:07 -04:00

194 lines
7.5 KiB
Markdown

# Embedded Systems Reverse Engineering
[Repository](https://github.com/mytechnotalent/Embedded-Hacking)
## Week 6
Static Variables in Embedded Systems: Debugging and Hacking Static Variables w/ GPIO Input Basics
### Non-Credit Practice Exercise 4: Invert the Button Logic with XOR
#### Objective
Find the `eor.w r3, r3, #1` instruction that implements the ternary operator's button inversion, patch it to `eor.w r3, r3, #0` using a hex editor to reverse the LED behavior, and verify that the LED is now ON when the button is pressed and OFF when released — the opposite of the original behavior.
#### Prerequisites
- Completed Week 6 tutorial (all GDB and hex editor sections)
- `0x0014_static-variables.bin` binary available in your build directory
- GDB (`arm-none-eabi-gdb`) and OpenOCD installed
- A hex editor (HxD, ImHex, or similar)
- Python installed (for UF2 conversion)
- Raspberry Pi Pico 2 with button on GP15 and LED on GP16
#### Task Description
The original program uses `gpio_put(LED_GPIO, !gpio_get(BUTTON_GPIO))` which the compiler implements as an XOR (`eor.w r3, r3, #1`) to invert the button state. With the pull-up resistor, button released = HIGH, so `HIGH XOR 1 = 0` (LED off). You will patch the XOR operand from `#1` to `#0`, which effectively removes the inversion: `HIGH XOR 0 = 1` (LED on when released). This exercise demonstrates how a single-byte binary patch can completely reverse hardware behavior.
#### Step-by-Step Instructions
##### Step 1: Start the Debug Session
**Terminal 1 - Start OpenOCD:**
```powershell
openocd ^
-s "C:\Users\flare-vm\.pico-sdk\openocd\0.12.0+dev\scripts" ^
-f interface/cmsis-dap.cfg ^
-f target/rp2350.cfg ^
-c "adapter speed 5000"
```
**Terminal 2 - Start GDB:**
```powershell
arm-none-eabi-gdb build\0x0014_static-variables.elf
```
**Connect to target:**
```gdb
(gdb) target remote :3333
(gdb) monitor reset halt
```
##### Step 2: Locate the GPIO Logic
From the tutorial, the GPIO input/output logic is near address `0x10000274`. Disassemble:
```gdb
(gdb) x/10i 0x10000274
```
Look for this sequence:
```
0x10000274: mov.w r1, #0xd0000000 ; SIO base address
0x10000280: ldr r3, [r1, #4] ; Read GPIO input register
0x10000282: ubfx r3, r3, #15, #1 ; Extract bit 15 (button state)
0x10000286: eor.w r3, r3, #1 ; XOR with 1 — INVERT ? OUR TARGET
0x1000028a: mcrr 0, 4, r2, r3, cr0 ; Write to GPIO output
```
The `eor.w r3, r3, #1` instruction is at address `0x10000286`.
##### Step 3: Understand the Current Logic
Trace the logic with the pull-up resistor:
| Button State | GPIO 15 Input | After UBFX | After EOR #1 | LED (GPIO 16) |
| ------------ | ------------- | ---------- | ------------ | -------------- |
| Released | 1 (HIGH) | 1 | 0 | OFF |
| Pressed | 0 (LOW) | 0 | 1 | ON |
The `eor.w #1` flips the bit, implementing the `!` (NOT) from the C code.
##### Step 4: Verify with GDB
Set a breakpoint at the `eor.w` instruction:
```gdb
(gdb) b *0x10000286
(gdb) c
```
When it hits, check what value is about to be XORed:
```gdb
(gdb) info registers r3
```
- If button is **released**: `r3 = 1` ? after EOR: `r3 = 0`
- If button is **pressed**: `r3 = 0` ? after EOR: `r3 = 1`
##### Step 5: Test the Patch in GDB
Modify the EOR operand in RAM to see the effect live:
```gdb
(gdb) set $r3 = 0
(gdb) si
(gdb) info registers r3
```
Or skip the EOR entirely by advancing the PC past it, then observe the LED behavior.
##### Step 6: Examine the Instruction Encoding
Look at the raw bytes:
```gdb
(gdb) x/4bx 0x10000286
0x10000286 <main+82>: 0x83 0xf0 0x01 0x03
```
The `eor.w` instruction is a 32-bit Thumb-2 encoding. The 4 bytes break down as:
- `0x83 0xF0` — opcode + source register (r3)
- `0x01`**the immediate value (`#1`)** ? this is what we change
- `0x03` — destination register (r3)
##### Step 7: Patch with the Hex Editor
1. In HxD, open `C:\Users\flare-vm\Desktop\Embedded-Hacking-main\0x0014_static-variables\build\0x0014_static-variables.bin`
2. The instruction starts at file offset: `0x10000286 - 0x10000000 = 0x286`
3. The immediate byte is the 3rd byte: offset `0x286 + 2 = 0x288`
4. Press **Ctrl+G** and enter offset: `288`
5. You should see `01` — change it to `00`
6. Verify the surrounding bytes (`83 F0` before and `03` after) are unchanged
7. Click **File** ? **Save As** ? `0x0014_static-variables-h.bin` (in the same `build` directory)
> ?? **Why offset `0x288`?** The 4-byte instruction starts at `0x286`, but the immediate value `#1` is in the **third byte** (index 2), so it's at `0x286 + 2 = 0x288`.
##### Step 8: Predict the New Behavior
After patching, the logic changes:
| Button State | GPIO 15 Input | After UBFX | After EOR #0 | LED (GPIO 16) |
| ------------ | ------------- | ---------- | ------------ | -------------- |
| Released | 1 (HIGH) | 1 | 1 | **ON** |
| Pressed | 0 (LOW) | 0 | 0 | **OFF** |
The LED behavior is now **inverted** from the original!
##### Step 9: Convert to UF2 and Flash
```powershell
cd C:\Users\flare-vm\Desktop\Embedded-Hacking-main\0x0014_static-variables
python ..\uf2conv.py build\0x0014_static-variables-h.bin --base 0x10000000 --family 0xe48bff59 --output build\hacked.uf2
```
1. Hold BOOTSEL and plug in your Pico 2
2. Drag and drop `hacked.uf2` onto the RPI-RP2 drive
##### Step 10: Verify the Hack
Test the button:
- **Button NOT pressed**: LED should now be **ON** (was OFF before patching)
- **Button PRESSED**: LED should now be **OFF** (was ON before patching)
The LED behavior is completely reversed by changing a single byte!
#### Expected Output
After completing this exercise, you should be able to:
- Locate XOR / EOR instructions in disassembled GPIO logic
- Understand how XOR implements logical NOT for single-bit values
- Patch a Thumb-2 encoded immediate operand
- Predict hardware behavior changes from binary patches
#### Questions for Reflection
###### Question 1: Why does XOR with 1 act as a NOT for single-bit values? Write out the truth table for `x XOR 1` and `x XOR 0` where x is 0 or 1.
###### Question 2: Instead of changing `eor.w r3, r3, #1` to `eor.w r3, r3, #0`, could you achieve the same result by NOPing (removing) the instruction entirely? What bytes encode a NOP in Thumb?
###### Question 3: The pull-up resistor means "pressed = LOW." If you removed the pull-up (changed `gpio_pull_up` to no pull), would the button still work? Why or why not?
###### Question 4: The `ubfx r3, r3, #0xf, #0x1` instruction extracts bit 15. If you changed `#0xf` to `#0x10` (bit 16), what GPIO pin would you be reading? What value would you get if nothing is connected to that pin?
#### Tips and Hints
- `eor.w r3, r3, #1` is a 32-bit Thumb-2 instruction (4 bytes), not a 16-bit Thumb instruction
- A Thumb NOP is `00 bf` (2 bytes) — you would need two NOPs to replace a 4-byte instruction
- Use GDB `x/1tw` to view a word in binary format, making bit manipulation easier to see
- The SIO base address `0xd0000000` provides single-cycle access to GPIO — it's separate from the IO_BANK0 registers at `0x40028000`
#### Next Steps
- Review all four exercises and verify you can patch any part of the binary: data values, arithmetic operations, and logic operations
- Try combining multiple hacks in a single binary: change the initial value, speed up the overflow, AND invert the button logic
- Compare your patched binary with the original using `fc /b original.bin patched.bin` in the command prompt to see all changed bytes