arch/arm64_symbolization

ARM64 architecture-specific rules for symbolization

split_load(ea:address, nextea:address, dest:address, type:symbol)

Compute an immediate load performed across two consecutive instructions

• Uses: arch.load_operation, arch.store_operation, instruction, instruction_get_dest_op, instruction_get_op, next, op_immediate, op_indirect_mapped, op_regdirect_contains_reg, symbol

• Used by: __agg_subclause8, base_addr_offset_operand_candidate, symbol_minus_symbol, symbolic_operand_attribute, symbolic_operand_mips_candidate

• Recursive: stack_def_use.live_var_at_block_end, value_reg, stack_def_use.ref_in_block, discarded_block, fallthrough_over_padding, arm_jump_table_candidate_start, no_return_call_refined, reg_def_use.live_var_at_prior_used, init_ldr_add_pc, relocation_adjustment, arm_jump_table_candidate, data_block_limit, cmp_defines, __agg_subclause6, after_end, block_last_instruction, adjusts_stack_in_block, base_relative_operation, must_fallthrough, arch.reg_relative_load, stack_def_use.block_last_def, reg_used_for, local_dynamic_tls_candidate, value_reg_unsupported, branch_to_calculated_pc_rel_addr, const_value_reg_used, symbolic_expr_from_relocation, common_tail, reg_def_use.ambiguous_last_def_in_block, reg_def_use.def_used, value_reg_edge, wis_schedule, correlated_live_reg, arm_jump_table_skip_first_entry, symbol_minus_symbol_litpool_access_pattern, wis_memo, reg_def_use.live_var_used, __agg_single6, reg_reg_arithmetic_operation_defs, call_tls_get_addr, unresolved_interval, data_in_code_propagate, reg_def_use.flow_def, block_total_points, stack_def_use.live_var_at_prior_used, block_candidate_boundaries, stack_def_use.last_def_in_block, reg_has_base_image, instruction_memory_access_size, tls_get_addr, inferred_main_dispatch, wis_prior, tls_relative_operand_mips, __agg_single10, composite_data_access_candidate, candidate_block_is_not_padding, jump_table_candidate_refined, first_block_in_byte_interval, indexed_pc_relative_load_relative, got_relative_operand, __agg_subclause3, next_type, candidate_block_is_padding, litpool_confidence, reg_def_use.block_last_def, segment_target_range, incomplete_block, reg_def_use.live_var_at_block_end, litpool_ref, jump_table_element_access, stack_def_use.used_in_block, arch.extend_load, __agg_subclause2, contains_plausible_instr_seq, arm_jump_table_cmp_limit, block_limit, nop_in_padding_candidate, reg_def_use.live_var_def, stack_base_reg_move, no_value_reg_limit, possible_target_from, relative_address, arm_jump_table_block_instruction, block_instruction_next, value_reg_limit, padding_block_limit, resolved_reaches, code_in_block_candidate_refined, __agg_single3, jump_table_max, block_next, straight_line_last_def, indefinite_litpool_ref, call_may_fallthrough_inter, invalid_jump_table_candidate, split_load_conflict, stack_def_use.def_used, block_points_proportional, data_block_candidate, arm_jump_table_data_block, compare_and_jump_register, litpool_symbolic_operand, split_load_for_symbolization, data_segment, contains_implausible_instr_seq, next_start, inferred_main_in_reg, block, split_load_point, call_tls_get_addr_mips, indexed_pc_relative_load, compare_and_jump_indirect_op_valid, jump_table_signed, split_load_operand, start_function, reg_def_use.defined_in_block, flags_and_jump_pair, reg_def_use.used_in_block, basic_target, overlap_with_litpool, __agg_subclause4, jump_table_target, next_end, jump_table_prelude, overlapping_instruction, base_relative_operand, relocation_adjustment_total, impossible_block, inter_procedural_edge, init_symbol_minus_symbol_candidate_arm, call_target_has_other_fallthrough_inter, gp_relative_operand, stack_def_use.live_var_used, jump_table_start, block_candidate_dependency_edge, unresolved_block, no_return_call_propagated, self_contained_segment, reg_def_use.ref_in_block, reg_def_use.ambiguous_block_last_def, reg_def_use.return_block_end, block_points, no_return_call, unresolved_block_overlap, def_used_for_address, invalid, cmp_reg_to_reg, unresolved_interval_order, code_in_block_candidate, next_function_entry_initial, block_boundaries, initial_function_containing_return, compare_and_jump_immediate, simple_data_access_pattern, unlikely_have_symbolic_immediate, straight_line_def_used, block_heuristic, stack_def_use.live_var_def, known_block, __agg_subclause7, base_relative_jump, code_in_block, reg_def_use.used, cinf_ldr_add_pc, plt_block, relative_address_start, likely_fallthrough, discarded_split_load, split_load_total_points, arch.simple_data_load, __agg_single2, litpool_boundaries, transition_block_limit, last_value_reg_limit, negative_block_heuristic, block_implies_block, relative_jump_table_entry_candidate, adrp_used, initialized_data_segment, is_padding, compare_and_jump_indirect, jump_table_candidate, block_overlap, split_load_candidate, wis_has_prior, tls_desc_call, reg_has_got, arm_jump_table_block_start, reg_def_use.return_val_used, reg_def_use.last_def_in_block, arm_jump_table_data_block_limit, next_block_in_byte_interval, stack_def_use.defined_in_block, may_fallthrough, data_in_code, data_access, wis_schedule_iter, stack_def_use.live_var_used_in_block, plt_entry, hi_load_prop, no_return_block, possible_target, split_load, padding_block_candidate, function_inference.function_entry_initial

adrp_used(EA:address, Reg:register, UsedEA:address, Dest:address)

A reference to a page-aligned value loaded by an adr or adrp instruction.

EA: Address of the adr/adrp instruction Reg: The register that holds the used value UsedEA: Address where the value is used Dest: The destination address

We care primarily about adrp, but sometimes if the target is close enough, the assembler will substitute an adr to the page-aligned address instead.

Reg is not necessarily the register loaded by adrp! It may be the result of:

adrp x0, label mov x1, x0

UsedEA:

# x1 is used

and Reg would be x1

• Uses: arch.adr_dest, instruction

• Recursive: stack_def_use.live_var_at_block_end, value_reg, stack_def_use.ref_in_block, discarded_block, fallthrough_over_padding, arm_jump_table_candidate_start, no_return_call_refined, reg_def_use.live_var_at_prior_used, init_ldr_add_pc, relocation_adjustment, arm_jump_table_candidate, data_block_limit, cmp_defines, __agg_subclause6, after_end, block_last_instruction, adjusts_stack_in_block, base_relative_operation, must_fallthrough, arch.reg_relative_load, stack_def_use.block_last_def, reg_used_for, local_dynamic_tls_candidate, value_reg_unsupported, branch_to_calculated_pc_rel_addr, const_value_reg_used, symbolic_expr_from_relocation, common_tail, reg_def_use.ambiguous_last_def_in_block, reg_def_use.def_used, value_reg_edge, wis_schedule, correlated_live_reg, arm_jump_table_skip_first_entry, symbol_minus_symbol_litpool_access_pattern, wis_memo, reg_def_use.live_var_used, __agg_single6, reg_reg_arithmetic_operation_defs, call_tls_get_addr, unresolved_interval, data_in_code_propagate, reg_def_use.flow_def, block_total_points, stack_def_use.live_var_at_prior_used, block_candidate_boundaries, stack_def_use.last_def_in_block, reg_has_base_image, instruction_memory_access_size, tls_get_addr, inferred_main_dispatch, wis_prior, tls_relative_operand_mips, __agg_single10, composite_data_access_candidate, candidate_block_is_not_padding, jump_table_candidate_refined, first_block_in_byte_interval, indexed_pc_relative_load_relative, got_relative_operand, __agg_subclause3, next_type, candidate_block_is_padding, litpool_confidence, reg_def_use.block_last_def, segment_target_range, incomplete_block, reg_def_use.live_var_at_block_end, litpool_ref, jump_table_element_access, stack_def_use.used_in_block, arch.extend_load, __agg_subclause2, contains_plausible_instr_seq, arm_jump_table_cmp_limit, block_limit, nop_in_padding_candidate, reg_def_use.live_var_def, stack_base_reg_move, no_value_reg_limit, possible_target_from, relative_address, arm_jump_table_block_instruction, block_instruction_next, value_reg_limit, padding_block_limit, resolved_reaches, code_in_block_candidate_refined, __agg_single3, jump_table_max, block_next, straight_line_last_def, indefinite_litpool_ref, call_may_fallthrough_inter, invalid_jump_table_candidate, split_load_conflict, stack_def_use.def_used, block_points_proportional, data_block_candidate, arm_jump_table_data_block, compare_and_jump_register, litpool_symbolic_operand, split_load_for_symbolization, data_segment, contains_implausible_instr_seq, next_start, inferred_main_in_reg, block, split_load_point, call_tls_get_addr_mips, indexed_pc_relative_load, compare_and_jump_indirect_op_valid, jump_table_signed, split_load_operand, start_function, reg_def_use.defined_in_block, flags_and_jump_pair, reg_def_use.used_in_block, basic_target, overlap_with_litpool, __agg_subclause4, jump_table_target, next_end, jump_table_prelude, overlapping_instruction, base_relative_operand, relocation_adjustment_total, impossible_block, inter_procedural_edge, init_symbol_minus_symbol_candidate_arm, call_target_has_other_fallthrough_inter, gp_relative_operand, stack_def_use.live_var_used, jump_table_start, block_candidate_dependency_edge, unresolved_block, no_return_call_propagated, self_contained_segment, reg_def_use.ref_in_block, reg_def_use.ambiguous_block_last_def, reg_def_use.return_block_end, block_points, no_return_call, unresolved_block_overlap, def_used_for_address, invalid, cmp_reg_to_reg, unresolved_interval_order, code_in_block_candidate, next_function_entry_initial, block_boundaries, initial_function_containing_return, compare_and_jump_immediate, simple_data_access_pattern, unlikely_have_symbolic_immediate, straight_line_def_used, block_heuristic, stack_def_use.live_var_def, known_block, __agg_subclause7, base_relative_jump, code_in_block, reg_def_use.used, cinf_ldr_add_pc, plt_block, relative_address_start, likely_fallthrough, discarded_split_load, split_load_total_points, arch.simple_data_load, __agg_single2, litpool_boundaries, transition_block_limit, last_value_reg_limit, negative_block_heuristic, block_implies_block, relative_jump_table_entry_candidate, adrp_used, initialized_data_segment, is_padding, compare_and_jump_indirect, jump_table_candidate, block_overlap, split_load_candidate, wis_has_prior, tls_desc_call, reg_has_got, arm_jump_table_block_start, reg_def_use.return_val_used, reg_def_use.last_def_in_block, arm_jump_table_data_block_limit, next_block_in_byte_interval, stack_def_use.defined_in_block, may_fallthrough, data_in_code, data_access, wis_schedule_iter, stack_def_use.live_var_used_in_block, plt_entry, hi_load_prop, no_return_block, possible_target, split_load, padding_block_candidate, function_inference.function_entry_initial

split_load_tail(EA:address, Offset:number, Reg:register, Type:symbol)

An instruction which could be the second half of a split load.

• Uses: arch.memory_access, arch.reg_arithmetic_operation, instruction, op_immediate

• Used by: split_load_for_symbolization

split_load_for_symbolization(ea:address, nextea:address, dest:address, type:symbol)

This version of split_load uses def_used, so this should not be used for the code inference step due to cyclic negation issue. Instead, this is for the symbolization step.

• Uses: split_load_tail

• Used by: symbolic_operand_attribute

• Recursive: stack_def_use.live_var_at_block_end, value_reg, stack_def_use.ref_in_block, discarded_block, fallthrough_over_padding, arm_jump_table_candidate_start, no_return_call_refined, reg_def_use.live_var_at_prior_used, init_ldr_add_pc, relocation_adjustment, arm_jump_table_candidate, data_block_limit, cmp_defines, __agg_subclause6, after_end, block_last_instruction, adjusts_stack_in_block, base_relative_operation, must_fallthrough, arch.reg_relative_load, stack_def_use.block_last_def, reg_used_for, local_dynamic_tls_candidate, value_reg_unsupported, branch_to_calculated_pc_rel_addr, const_value_reg_used, symbolic_expr_from_relocation, common_tail, reg_def_use.ambiguous_last_def_in_block, reg_def_use.def_used, value_reg_edge, wis_schedule, correlated_live_reg, arm_jump_table_skip_first_entry, symbol_minus_symbol_litpool_access_pattern, wis_memo, reg_def_use.live_var_used, __agg_single6, reg_reg_arithmetic_operation_defs, call_tls_get_addr, unresolved_interval, data_in_code_propagate, reg_def_use.flow_def, block_total_points, stack_def_use.live_var_at_prior_used, block_candidate_boundaries, stack_def_use.last_def_in_block, reg_has_base_image, instruction_memory_access_size, tls_get_addr, inferred_main_dispatch, wis_prior, tls_relative_operand_mips, __agg_single10, composite_data_access_candidate, candidate_block_is_not_padding, jump_table_candidate_refined, first_block_in_byte_interval, indexed_pc_relative_load_relative, got_relative_operand, __agg_subclause3, next_type, candidate_block_is_padding, litpool_confidence, reg_def_use.block_last_def, segment_target_range, incomplete_block, reg_def_use.live_var_at_block_end, litpool_ref, jump_table_element_access, stack_def_use.used_in_block, arch.extend_load, __agg_subclause2, contains_plausible_instr_seq, arm_jump_table_cmp_limit, block_limit, nop_in_padding_candidate, reg_def_use.live_var_def, stack_base_reg_move, no_value_reg_limit, possible_target_from, relative_address, arm_jump_table_block_instruction, block_instruction_next, value_reg_limit, padding_block_limit, resolved_reaches, code_in_block_candidate_refined, __agg_single3, jump_table_max, block_next, straight_line_last_def, indefinite_litpool_ref, call_may_fallthrough_inter, invalid_jump_table_candidate, split_load_conflict, stack_def_use.def_used, block_points_proportional, data_block_candidate, arm_jump_table_data_block, compare_and_jump_register, litpool_symbolic_operand, split_load_for_symbolization, data_segment, contains_implausible_instr_seq, next_start, inferred_main_in_reg, block, split_load_point, call_tls_get_addr_mips, indexed_pc_relative_load, compare_and_jump_indirect_op_valid, jump_table_signed, split_load_operand, start_function, reg_def_use.defined_in_block, flags_and_jump_pair, reg_def_use.used_in_block, basic_target, overlap_with_litpool, __agg_subclause4, jump_table_target, next_end, jump_table_prelude, overlapping_instruction, base_relative_operand, relocation_adjustment_total, impossible_block, inter_procedural_edge, init_symbol_minus_symbol_candidate_arm, call_target_has_other_fallthrough_inter, gp_relative_operand, stack_def_use.live_var_used, jump_table_start, block_candidate_dependency_edge, unresolved_block, no_return_call_propagated, self_contained_segment, reg_def_use.ref_in_block, reg_def_use.ambiguous_block_last_def, reg_def_use.return_block_end, block_points, no_return_call, unresolved_block_overlap, def_used_for_address, invalid, cmp_reg_to_reg, unresolved_interval_order, code_in_block_candidate, next_function_entry_initial, block_boundaries, initial_function_containing_return, compare_and_jump_immediate, simple_data_access_pattern, unlikely_have_symbolic_immediate, straight_line_def_used, block_heuristic, stack_def_use.live_var_def, known_block, __agg_subclause7, base_relative_jump, code_in_block, reg_def_use.used, cinf_ldr_add_pc, plt_block, relative_address_start, likely_fallthrough, discarded_split_load, split_load_total_points, arch.simple_data_load, __agg_single2, litpool_boundaries, transition_block_limit, last_value_reg_limit, negative_block_heuristic, block_implies_block, relative_jump_table_entry_candidate, adrp_used, initialized_data_segment, is_padding, compare_and_jump_indirect, jump_table_candidate, block_overlap, split_load_candidate, wis_has_prior, tls_desc_call, reg_has_got, arm_jump_table_block_start, reg_def_use.return_val_used, reg_def_use.last_def_in_block, arm_jump_table_data_block_limit, next_block_in_byte_interval, stack_def_use.defined_in_block, may_fallthrough, data_in_code, data_access, wis_schedule_iter, stack_def_use.live_var_used_in_block, plt_entry, hi_load_prop, no_return_block, possible_target, split_load, padding_block_candidate, function_inference.function_entry_initial

lo_reloc_index(RelocType:symbol, OpIndex:unsigned)

OpIndex: operand index for the corresponding relocation

• Used by: symbolic_expr_from_relocation, symbolic_operand_attribute

movz_movk_insn(EA:address, Reg:register, Val:number, ShiftMask:unsigned, Operation:symbol)

Individual MOVZ or MOVK instruction with its 16-bit immediate and shift.

• Uses: arch.move_reg_imm, instruction, op_shifted

• Used by: movz_movk_chain, symbolic_operand_attribute, symbolic_operand_candidate

movz_movk_chain(EA_first:address, EA_last:address, Reg:register, Value:number, UsedShiftMask:unsigned, Count:unsigned)

A chain of MOVZ followed by one or more MOVKs to the same register. EA_first is always the MOVZ instruction. EA_last is the final instruction in the chain. Value is the accumulated 64-bit constant constructed from all instructions. UsedShiftMask tracks which 16-bit lanes have been written. Count is the number of instructions in the chain (1-4).

• Uses: movz_movk_insn, next

• Used by: movz_movk_complete, movz_movk_member

• Recursive: movz_movk_chain

movz_movk_complete(EA_first:address, EA_last:address, Value:number)

The longest complete MOVZ+MOVK chain: a chain that cannot be extended further. This is the one we use for symbolization.

• Uses: movz_movk_chain

• Used by: may_have_symbolic_immediate, movz_movk_member, symbolic_operand_attribute, symbolic_operand_candidate

movz_movk_member(EA:address, EA_first:address, EA_last:address)

Helper: is EA part of a complete MOVZ+MOVK chain?

• Uses: movz_movk_chain, movz_movk_complete

• Used by: may_have_symbolic_immediate, symbolic_operand_attribute, symbolic_operand_candidate, unlikely_have_symbolic_immediate

movz_movk_shift_group(ShiftMask:unsigned, Group:symbol)

Helper: map a shift amount to the corresponding group attribute name.

• Used by: symbolic_operand_attribute, symbolic_operand_candidate