[AArch64][Windows] Add llvm-readobj support for ec_context unwind opcode

[bylaws]

ARM64EC uses the same CONTEXT structure as x86_64 as opposed to the regular ARM64 context, a new unwind opcode (MSFT_OP_EC_CONTEXT) is added to handle this.

[llvmbot]

@llvm/pr-subscribers-llvm-binary-utilities

Author: Billy Laws (bylaws)

Changes

ARM64EC uses the same CONTEXT structure as x86_64 as opposed to the regular ARM64 context, a new unwind opcode (MSFT_OP_EC_CONTEXT) is added to handle this.

---

Full diff: https://github.com/llvm/llvm-project/pull/69515.diff

2 Files Affected:

• (modified) llvm/tools/llvm-readobj/ARMWinEHPrinter.cpp (+8)
• (modified) llvm/tools/llvm-readobj/ARMWinEHPrinter.h (+2)

diff --git a/llvm/tools/llvm-readobj/ARMWinEHPrinter.cpp b/llvm/tools/llvm-readobj/ARMWinEHPrinter.cpp
index ef77d4b2fd224bd..cf5c77cf107c36f 100644
--- a/llvm/tools/llvm-readobj/ARMWinEHPrinter.cpp
+++ b/llvm/tools/llvm-readobj/ARMWinEHPrinter.cpp
@@ -171,6 +171,7 @@ const Decoder::RingEntry Decoder::Ring64[] = {
     {0xff, 0xe8, 1, &Decoder::opcode_trap_frame},
     {0xff, 0xe9, 1, &Decoder::opcode_machine_frame},
     {0xff, 0xea, 1, &Decoder::opcode_context},
+    {0xff, 0xeb, 1, &Decoder::opcode_ec_context},
     {0xff, 0xec, 1, &Decoder::opcode_clear_unwound_to_call},
     {0xff, 0xfc, 1, &Decoder::opcode_pac_sign_lr},
 };
@@ -969,6 +970,13 @@ bool Decoder::opcode_context(const uint8_t *OC, unsigned &Offset,
   return false;
 }
 
+bool Decoder::opcode_ec_context(const uint8_t *OC, unsigned &Offset,
+                                unsigned Length, bool Prologue) {
+  SW.startLine() << format("0x%02x                ; EC context\n", OC[Offset]);
+  ++Offset;
+  return false;
+}
+
 bool Decoder::opcode_clear_unwound_to_call(const uint8_t *OC, unsigned &Offset,
                                            unsigned Length, bool Prologue) {
   SW.startLine() << format("0x%02x                ; clear unwound to call\n",
diff --git a/llvm/tools/llvm-readobj/ARMWinEHPrinter.h b/llvm/tools/llvm-readobj/ARMWinEHPrinter.h
index 0ffebe5834c480b..92b3a78e4330098 100644
--- a/llvm/tools/llvm-readobj/ARMWinEHPrinter.h
+++ b/llvm/tools/llvm-readobj/ARMWinEHPrinter.h
@@ -129,6 +129,8 @@ class Decoder {
                             unsigned Length, bool Prologue);
   bool opcode_context(const uint8_t *Opcodes, unsigned &Offset, unsigned Length,
                       bool Prologue);
+  bool opcode_ec_context(const uint8_t *Opcodes, unsigned &Offset, unsigned Length,
+                         bool Prologue);
   bool opcode_clear_unwound_to_call(const uint8_t *Opcodes, unsigned &Offset,
                                     unsigned Length, bool Prologue);
   bool opcode_pac_sign_lr(const uint8_t *Opcodes, unsigned &Offset,

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[bylaws]

cc: @mstorsjo @cjacek

[mstorsjo]

The change itself looks good, thanks!

This is lacking a test, but OTOH I don't see a standalone test for disassembling all the other corresponding opcodes either.

Many of them were added in llvm-readobj in the same commit as the MC layer code for assembling them, in f5e2ea9, so the readobj changes are covered by those MC tests at least.

So as long as this is merged close together with #69520 which does provide some test coverage for this, I think it's fine.