contrib: add symbol-check test for non-existence of vmova instructions in Windows build #28413

1. 
   fanquake commented at 12:35 pm on September 5, 2023: member

   Given we are patching the Windows GCC used for releases, to avoid the use of certain aligned assembly instructions, we should probably add a symbol-check test to ensure the patches are working as intended.

   See #24726, #24736 and #28151 for the background here.

2. fanquake added the label Windows on Sep 5, 2023
3. 
   laanwj commented at 10:32 am on April 11, 2024: member

   Wouldn’t checking the instructions be more of a security-check instead of a symbol-check?

   Assigning myself as I’m interested on working on this.

4. laanwj assigned laanwj on Apr 11, 2024
5. 
   laanwj commented at 8:53 pm on April 14, 2024: member

   Is introducing a dependency on the python capstone binding acceptable here? It’s not possible to do this check without a disassembler of some kind, we don’t want to rely on calling out to objdump, and i’m not sure we want our own x86 mini-disassembler (hard to review, slow in pure python). Capstone is great for this and might help in future instruction security checks as well. But it’s another build-time dep.
6. 
   laanwj commented at 3:51 am on April 15, 2024: member

   For reference, with capstone this check is as simple as:

    0# Intel® 64 and IA-32 Architectures Software Developer’s Manual:
    1#   chapter 14.9, table 14-22. Instructions Requiring Explicitly Aligned Memory
    2#   chapter 15.7, Table 15-6. SIMD Instructions Requiring Explicitly Aligned Memory
    3#
    4# This amounts to the following instructions:
    5#
    6# instruction                  chapter 4.3 section
    7# ---------------------------  ---------------------------------
    8# (V)MOVDQA xmm, mBBB          MOVDQA,VMOVDQA32/64—Move Aligned Packed Integer Values
    9# (V)MOVDQA mBBB, xmm          MOVDQA,VMOVDQA32/64—Move Aligned Packed Integer Values
   10# (V)MOVAPS xmm, mBBB          MOVAPS—Move Aligned Packed Single Precision Floating-Point Values
   11# (V)MOVAPS mBBB, xmm          MOVAPS—Move Aligned Packed Single Precision Floating-Point Values
   12# (V)MOVAPD xmm, mBBB          MOVAPD—Move Aligned Packed Double Precision Floating-Point Values
   13# (V)MOVAPD mBBB, xmm          MOVAPD—Move Aligned Packed Double Precision Floating-Point Values
   14# (V)MOVNTPS mBBB, xmm         MOVNTPS—Store Packed Single Precision Floating-Point Values Using Non-Temporal Hint
   15# (V)MOVNTPD mBBB, xmm         MOVNTPD—Store Packed Double Precision Floating-Point Values Using Non-Temporal Hint
   16# (V)MOVNTDQ mBBB, xmm         MOVNTDQ—Store Packed Integers Using Non-Temporal Hint
   17# (V)MOVNTDQA xmm, mBBB        MOVNTDQA—Load Double Quadword Non-Temporal Aligned Hint
   18#
   19# BBB is the bit size, which can be 128, 256 or 512.
   20#
   21FORBIDDEN_VMOVA = {
   22    capstone.x86.X86_INS_MOVDQA,    capstone.x86.X86_INS_VMOVDQA,   capstone.x86.X86_INS_VMOVDQA32, capstone.x86.X86_INS_VMOVDQA64,
   23    capstone.x86.X86_INS_MOVAPS,    capstone.x86.X86_INS_VMOVAPS,
   24    capstone.x86.X86_INS_MOVAPD,    capstone.x86.X86_INS_VMOVAPD,
   25    capstone.x86.X86_INS_MOVNTPS,   capstone.x86.X86_INS_VMOVNTPS,
   26    capstone.x86.X86_INS_MOVNTPD,   capstone.x86.X86_INS_VMOVNTPD,
   27    capstone.x86.X86_INS_MOVNTDQ,   capstone.x86.X86_INS_VMOVNTDQ,
   28    capstone.x86.X86_INS_MOVNTDQA,  capstone.x86.X86_INS_VMOVNTDQA,
   29}
   30
   31def check_ELF_no_vmova(binary) -> bool:
   32    '''
   33    Check for vmov instructions that require alignment.
   34    These are a potential problem due to a stack alignment bug in GCC on Windows.
   35    See [#28413](/bitcoin-bitcoin/28413/) for specifics.
   36    '''
   37    cs = capstone.Cs(capstone.CS_ARCH_X86, capstone.CS_MODE_64)
   38
   39    found_forbidden = False
   40    for segment in binary.segments:
   41        # Find loaded, executable segments
   42        if segment.type == lief.ELF.SEGMENT_TYPES.LOAD and (segment.flags & lief.ELF.SEGMENT_FLAGS.X) != 0:
   43            # disassemble segment, check every instruction
   44            for i in cs.disasm(segment.content, segment.virtual_address): # -> CsInsn
   45                if i.id in FORBIDDEN_VMOVA:
   46                    found_forbidden = True
   47
   48    return not found_forbidden
   

   Still need to port this to PE.

   And figure out what the exceptions are in the current release, and how to either get rid of them, or identify them.

Contributors
fanquake laanwj

Labels
Windows