diff options
| author | Xusheng <xusheng@vector35.com> | 2024-08-02 14:04:12 +0800 |
|---|---|---|
| committer | Xusheng <xusheng@vector35.com> | 2024-08-02 16:10:31 +0800 |
| commit | a9a3f1ab55e18b345be37001c66cfa9455494b4a (patch) | |
| tree | 1554d9bc345482bb4a49b7927773c790196df252 /arch/x86 | |
| parent | 351238c300838d39ca896eeb7d180a35cc12998f (diff) | |
Properly lift push r16 instructions in x86. Fix https://github.com/Vector35/binaryninja-api/issues/4028
Diffstat (limited to 'arch/x86')
| -rw-r--r-- | arch/x86/il.cpp | 9 |
1 files changed, 7 insertions, 2 deletions
diff --git a/arch/x86/il.cpp b/arch/x86/il.cpp index 6046f274..e6d704a6 100644 --- a/arch/x86/il.cpp +++ b/arch/x86/il.cpp @@ -2861,8 +2861,13 @@ bool GetLowLevelILForInstruction(Architecture* arch, const uint64_t addr, LowLev case XED_ICLASS_PUSH: { - const unsigned int stackAdjustment = xed_decoded_inst_get_memop_address_width(xedd, 0) / 8; - if (opOneLen != stackAdjustment) // 32-bit push on 64-bit pushes a 64-bit value + // Whe the stack adjustment is different from the operand one size, zero-extend it. Note, a "push r16" in either + // x86 or x64 pushes the 2-byte register onto the stack, and there is no need to extend. See + // https://github.com/Vector35/binaryninja-api/issues/4028 and + // https://stackoverflow.com/questions/43435764/64-bit-mode-does-not-support-32-bit-push-and-pop-instructions + // for more details + const unsigned int stackAdjustment = xed_decoded_inst_get_memory_operand_length(xedd, 0); + if (opOneLen != stackAdjustment) { il.AddInstruction( il.Push(stackAdjustment, |