Skip to content

lldb: remove hardcoded instruction encodings #55

Open
@JordiChauzi

Description

@JordiChauzi

In lldb, we are using hardcoded encodings in different places:

  • lldb/source/Host/common/NativeProcessProtocol.cpp: in NativeProcessProtocol::GetSoftwareBreakpointTrapOpcode, g_dpu_opcode is the bkp encoding
  • lldb/scripts/dpu/dpu_commands.py: in dpu_attach_on_boot, we write the bkp encoding to memory
  • lldb/source/Plugins/Process/Dpu/Dpu.cpp:
    • in Dpu::Boot, breakpoint_instruction is the bkp encoding
    • in Dpu::GetThreadState, we check whether prev_instruction is the fault 1 encoding
  • lldb/source/Plugins/Process/Utility/RegisterContextDPU.cpp: in RegisterContextDPU::PCIsInstructionReturn we check whether instruction is the jump r23 encoding

This is mainly an issue when changing the ISA, which does not happen every week. However, when it happens, fixing some lldb issues may become harder than needed.

I think we should at least place all these encodings in a single place (per language), which will make it easier to change these values. Ideally we can imagine to generate these files from the metasm module.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions