[llvm] [BOLT] Use ULEB128 encoding for PIE/DSO exception tables (PR #116911)

[Maksim Panchenko via llvm-commits]

https://github.com/maksfb updated https://github.com/llvm/llvm-project/pull/116911

>From 0c3cda3f7cd03e5770ea1f17933481690eae4551 Mon Sep 17 00:00:00 2001
From: Maksim Panchenko <maks at fb.com>
Date: Tue, 19 Nov 2024 19:15:18 -0800
Subject: [PATCH 1/2] [BOLT] Use ULEB128 encoding for PIE/DSO exception tables

Use ULEB128 encoding for call sites in PIE/DSO binaries. The encoding
reduces the size of the tables compared to sdata4 and is the default
format used by Clang.

Note that for fixed-address executables we still use absolute
addressing to cover cases where landing pads can reside in different
function fragments.
---
 bolt/lib/Core/BinaryEmitter.cpp | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/bolt/lib/Core/BinaryEmitter.cpp b/bolt/lib/Core/BinaryEmitter.cpp
index 408663180935c2..7c4ff960a439d1 100644
--- a/bolt/lib/Core/BinaryEmitter.cpp
+++ b/bolt/lib/Core/BinaryEmitter.cpp
@@ -959,9 +959,9 @@ void BinaryEmitter::emitLSDA(BinaryFunction &BF, const FunctionFragment &FF) {
         if (NeedsLPAdjustment)
           LPOffsetExpr = MCBinaryExpr::createAdd(
               LPOffsetExpr, MCConstantExpr::create(1, *BC.Ctx), *BC.Ctx);
-        Streamer.emitValue(LPOffsetExpr, 4);
+        Streamer.emitULEB128Value(LPOffsetExpr);
       } else {
-        Streamer.emitIntValue(0, 4);
+        Streamer.emitULEB128IntValue(0);
       }
     };
   }
@@ -979,7 +979,10 @@ void BinaryEmitter::emitLSDA(BinaryFunction &BF, const FunctionFragment &FF) {
   // Emit the landing pad call site table. We use signed data4 since we can emit
   // a landing pad in a different part of the split function that could appear
   // earlier in the address space than LPStart.
-  Streamer.emitIntValue(dwarf::DW_EH_PE_sdata4, 1);
+  if (BC.HasFixedLoadAddress)
+    Streamer.emitIntValue(dwarf::DW_EH_PE_sdata4, 1);
+  else
+    Streamer.emitIntValue(dwarf::DW_EH_PE_uleb128, 1);
 
   MCSymbol *CSTStartLabel = BC.Ctx->createTempSymbol("CSTStart");
   MCSymbol *CSTEndLabel = BC.Ctx->createTempSymbol("CSTEnd");
@@ -996,8 +999,13 @@ void BinaryEmitter::emitLSDA(BinaryFunction &BF, const FunctionFragment &FF) {
 
     // Start of the range is emitted relative to the start of current
     // function split part.
-    Streamer.emitAbsoluteSymbolDiff(BeginLabel, StartSymbol, 4);
-    Streamer.emitAbsoluteSymbolDiff(EndLabel, BeginLabel, 4);
+    if (BC.HasFixedLoadAddress) {
+      Streamer.emitAbsoluteSymbolDiff(BeginLabel, StartSymbol, 4);
+      Streamer.emitAbsoluteSymbolDiff(EndLabel, BeginLabel, 4);
+    } else {
+      Streamer.emitAbsoluteSymbolDiffAsULEB128(BeginLabel, StartSymbol);
+      Streamer.emitAbsoluteSymbolDiffAsULEB128(EndLabel, BeginLabel);
+    }
     emitLandingPad(CallSite.LP);
     Streamer.emitULEB128IntValue(CallSite.Action);
   }

>From 04fb4524e9f8a6c48993e57c820bfd8600249706 Mon Sep 17 00:00:00 2001
From: Maksim Panchenko <maks at fb.com>
Date: Wed, 20 Nov 2024 00:13:46 -0800
Subject: [PATCH 2/2] fixup! [BOLT] Use ULEB128 encoding for PIE/DSO exception
 tables

---
 bolt/lib/Core/BinaryEmitter.cpp | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/bolt/lib/Core/BinaryEmitter.cpp b/bolt/lib/Core/BinaryEmitter.cpp
index 7c4ff960a439d1..4b5d8154728ccd 100644
--- a/bolt/lib/Core/BinaryEmitter.cpp
+++ b/bolt/lib/Core/BinaryEmitter.cpp
@@ -976,9 +976,8 @@ void BinaryEmitter::emitLSDA(BinaryFunction &BF, const FunctionFragment &FF) {
     Streamer.emitLabel(TTBaseRefLabel);
   }
 
-  // Emit the landing pad call site table. We use signed data4 since we can emit
-  // a landing pad in a different part of the split function that could appear
-  // earlier in the address space than LPStart.
+  // Emit encoding of entries in the call site table. The format is used for the
+  // call site start, length, and corresponding landing pad.
   if (BC.HasFixedLoadAddress)
     Streamer.emitIntValue(dwarf::DW_EH_PE_sdata4, 1);
   else