From a7382f9efe1d6dbd81aad4f7c1eb6aa0f301b2b4 Mon Sep 17 00:00:00 2001
From: York Jasper Niebuhr <yjnworkstation@gmail.com>
Date: Mon, 27 Oct 2025 20:40:12 +0100
Subject: [PATCH] Finalizer association of symbols with pinpoint data

---
 finalize/accumulation.h             | 16 ++++++++
 finalize/finalize.cpp               | 17 +++++++--
 finalize/symbol_collection.cpp      | 59 +++++++++++++++++++++++------
 finalize/symbol_collection.h        |  1 +
 pinpoint/stage1/asm_offset_pass.cpp |  8 ++++
 5 files changed, 86 insertions(+), 15 deletions(-)

diff --git a/finalize/accumulation.h b/finalize/accumulation.h
index f9a858e..4a347d3 100644
--- a/finalize/accumulation.h
+++ b/finalize/accumulation.h
@@ -4,14 +4,28 @@
 #include <unordered_map>
 #include <map>
 #include <list>
+#include <optional>
+#include <cstdint>
 
 struct IPIN {
+	struct HIT {
+		uint64_t vaddr;
+		uint32_t imm_offset;
+		uint32_t imm_size;
+	};
+
 	std::string symbol;
 	std::size_t local_target;
 	std::size_t field_offset;
+
+	std::optional<HIT> hit;
 };
 
 struct DPIN {
+	struct HIT {
+		uint64_t vaddr;
+	};
+
 	struct COMPONENT {
 		std::size_t offset;
 		std::size_t level;
@@ -20,6 +34,8 @@ struct DPIN {
 
 	std::string symbol;
 	std::list<COMPONENT> components;
+
+	std::optional<HIT> hit;
 };
 
 struct FIELD {
diff --git a/finalize/finalize.cpp b/finalize/finalize.cpp
index 9636031..8274950 100644
--- a/finalize/finalize.cpp
+++ b/finalize/finalize.cpp
@@ -89,12 +89,21 @@ int main(int argc, char** argv) {
 		return 1;
 	}
 
+	if (!associate_symbols()) {
+		std::cerr << "Failed to associate symbols with accumulated data!" << std::endl;
+		return 1;
+	}
+
 	// TODO
 	/*
-	1. Find virtual address and file address for all pins
-	2. Emit patcher program into final executable and set __spslr_program
-		-> make all patcher addresses relative to &__spslr_program
-		-> dpatch entries ordered by level (high levels/nest depths first)
+	1. For each ipin, disassemble instruction and find immediate offset
+	2. Find __spslr_program
+	3. Serialize entire patcher program
+		-> all pin addresses relative to &__spslr_program
+		-> dpatch entry components ordered by level
+	4. Add new section .spslr with patcher program
+	5. Set __spslr_program to (&.spslr - &__spslr_program)
+	6. Output final program
 	*/
 }
 
diff --git a/finalize/symbol_collection.cpp b/finalize/symbol_collection.cpp
index 025b43d..f30a709 100644
--- a/finalize/symbol_collection.cpp
+++ b/finalize/symbol_collection.cpp
@@ -1,20 +1,10 @@
 #include "symbol_collection.h"
+#include "accumulation.h"
 
 #include <iostream>
 
 #define SPSLR_PINPOINT_CU_UID_LABEL "__spslr_cu_" /* suffixed with "<uid>" */
 
-/*
-struct GlobalSymtab {
-    std::unordered_map<std::string, VADDR> symbols;
-};
-
-struct LocalSymtab {
-    std::string cu;
-    std::unordered_map<std::string, VADDR> symbols;
-};
-*/
-
 using namespace LIEF::ELF;
 
 GlobalSymtab global_syms;
@@ -91,3 +81,50 @@ bool collect_symbols(const std::unique_ptr<LIEF::ELF::Binary>& bin) {
 	return true;
 }
 
+bool associate_symbols() {
+	// Fetch symbols for each accumulated CU
+	for (auto& [cu_uid, cu] : units) {
+		if (!local_syms.contains(cu_uid)) {
+			std::cerr << "Could not find symbols for unit UID " << cu_uid << std::endl;
+			return false;
+		}
+
+		const LocalSymtab& lsyms = local_syms.at(cu_uid);
+
+		for (auto& [ipin_sym, ipin] : cu.ipins) {
+			if (!lsyms.symbols.contains(ipin_sym)) {
+				std::cerr << "Unable to locate symbol \"" << ipin_sym << "\" in unit " << cu_uid << std::endl;
+				return false;
+			}
+
+			ipin.hit.emplace();
+			ipin.hit->vaddr = lsyms.symbols.at(ipin_sym);
+			ipin.hit->imm_offset = 0;
+			ipin.hit->imm_size = 0;
+		}
+
+		for (auto& [dpin_sym, dpin] : cu.dpins) {
+			if (!lsyms.symbols.contains(dpin_sym)) {
+				std::cerr << "Unable to locate symbol \"" << dpin_sym << "\" in unit " << cu_uid << std::endl;
+				return false;
+			}
+
+			dpin.hit.emplace();
+			dpin.hit->vaddr = lsyms.symbols.at(dpin_sym);
+		}
+	}
+
+	// Fetch global symbols for global dpins
+	for (auto& [dpin_sym, dpin] : global_dpins) {
+		if (!global_syms.symbols.contains(dpin_sym)) {
+			std::cerr << "Unable to locate global symbol \"" << dpin_sym << "\"" << std::endl;
+			return false;
+		}
+
+		dpin.hit.emplace();
+		dpin.hit->vaddr = global_syms.symbols.at(dpin_sym);
+	}
+
+	return true;
+}
+
diff --git a/finalize/symbol_collection.h b/finalize/symbol_collection.h
index 4f73882..9c37703 100644
--- a/finalize/symbol_collection.h
+++ b/finalize/symbol_collection.h
@@ -19,4 +19,5 @@ extern GlobalSymtab global_syms;
 extern std::unordered_map<std::string, LocalSymtab> local_syms;
 
 bool collect_symbols(const std::unique_ptr<LIEF::ELF::Binary>& bin);
+bool associate_symbols();
 
diff --git a/pinpoint/stage1/asm_offset_pass.cpp b/pinpoint/stage1/asm_offset_pass.cpp
index 3b529b4..5e97088 100644
--- a/pinpoint/stage1/asm_offset_pass.cpp
+++ b/pinpoint/stage1/asm_offset_pass.cpp
@@ -76,6 +76,14 @@ static gimple* make_stage1_pin(tree lhs, UID target, std::size_t offset) {
 	if (!new_gasm)
 		return nullptr;
 
+	/*
+	TODO
+	These asm statements should not actually be volatile.
+	Instead, a final pass has to detect which have been removed.
+	Only data about still present ipins should be given to finalizer!
+	*/
+	gimple_asm_set_volatile(new_gasm, true); 
+
 	pins.emplace(uid, pin);
 	return new_gasm;
 }