[WebAssembly] Error out on invalid personality functions

[aheejin]

Summary

Without explicitly checking and erroring out, an invalid personality
function, which is not __gxx_wasm_personality_v0, caused
a segmentation fault down the line because WasmEHFuncInfo was not created.
This explicitly checks the validity of personality functions in
functions with EH pads and errors out explicitly with a helpful error
message. This also adds some more assertions to ensure WasmEHFuncInfo
is correctly created and non-null.

Invalid personality functions wouldn't be generated by our Clang, but
can be present in handwritten ll files, and more often, in files
transformed by passes like metarenamer, which is often used with
bugpoint to simplify names in bugpoint-reduced files.

llvm/lib/CodeGen/WasmEHPrepare.cpp

 // to the wrapper function _Unwind_CallPersonality in libunwind, and we retrieve
 // the selector after it returns.
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/CodeGen/WasmEHFuncInfo.h"
+#include "llvm/IR/EHPersonalities.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/IntrinsicsWebAssembly.h"
 #include "llvm/InitializePasses.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 
 using namespace llvm;
 
 #define DEBUG_TYPE "wasmehprepare"
for (BasicBlock &BB : F) {
     if (isa<CatchPadInst>(Pad))
       CatchPads.push_back(&BB);
     else if (isa<CleanupPadInst>(Pad))
       CleanupPads.push_back(&BB);
   }
   if (CatchPads.empty() && CleanupPads.empty())
     return false;
 
+  if (!F.hasPersonalityFn() ||
+      !isScopedEHPersonality(classifyEHPersonality(F.getPersonalityFn()))) {
+    report_fatal_error("Function '" + F.getName() +
+                       "' does not have a correct Wasm personality function "
+                       "'__gxx_wasm_personality_v0'");
+  }
   assert(F.hasPersonalityFn() && "Personality function not found");
 
   // __wasm_lpad_context global variable.
   // This variable should be thread local. If the target does not support TLS,
   // we depend on CoalesceFeaturesAndStripAtomics to downgrade it to
   // non-thread-local ones, in which case we don't allow this object to be
   // linked with other objects using shared memory.
   LPadContextGV = cast<GlobalVariable>(

llvm/lib/Target/WebAssembly/WebAssemblyCFGStackify.cpp

bool WebAssemblyCFGStackify::fixCatchUnwindMismatches(MachineFunction &MF) {
   //       ...
   //     end_try
   //   delegate 1 (caller)     ;; (new)
   // catch_all                 ;; ehpad B
   //   ...
   // end_try
 
   const auto *EHInfo = MF.getWasmEHFuncInfo();
+  assert(EHInfo);
   SmallVector<const MachineBasicBlock *, 8> EHPadStack;
   // For EH pads that have catch unwind mismatches, a map of <EH pad, its
   // correct unwind destination>.
   DenseMap<MachineBasicBlock *, MachineBasicBlock *> EHPadToUnwindDest;
 
   for (auto &MBB : reverse(MF)) {
     for (auto &MI : reverse(MBB)) {
       if (MI.getOpcode() == WebAssembly::TRY)

llvm/lib/Target/WebAssembly/WebAssemblyExceptionInfo.cpp

void WebAssemblyExceptionInfo::recalculate(
   // children of the parent exception, so we have to traverse the parent chain.
   //
   // We should traverse BBs in the preorder of the dominator tree, because
   // otherwise the result can be incorrect. For example, when there are three
   // exceptions A, B, and C and A > B > C (> is subexception relationship here),
   // and A's unwind destination is B and B's is C. When we visit B before A, we
   // end up extracting C only out of B but not out of A.
   const auto *EHInfo = MF.getWasmEHFuncInfo();
+  assert(EHInfo);
   SmallVector<std::pair<WebAssemblyException *, WebAssemblyException *>>
       UnwindWEVec;
   for (auto *DomNode : depth_first(&MDT)) {
     MachineBasicBlock *EHPad = DomNode->getBlock();
     if (!EHPad->isEHPad())
       continue;
     if (!EHInfo->hasUnwindDest(EHPad))
       continue;

llvm/test/CodeGen/WebAssembly/wasm-eh-invalid-personality.ll

+; RUN: not --crash llc < %s -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling 2>&1 | FileCheck %s
+
+target triple = "wasm32-unknown-unknown"
+
+; Tests if the compiler correctly errors out when a function having EH pads does
+; not have a correct Wasm personality function.
+
+define void @test() personality ptr @invalid_personality {
+; CHECK: LLVM ERROR: Function 'test' does not have a correct Wasm personality function '__gxx_wasm_personality_v0'
+entry:
+  invoke void @foo()
+          to label %try.cont unwind label %catch.dispatch
+
+catch.dispatch:                                   ; preds = %entry
+  %0 = catchswitch within none [label %catch.start] unwind to caller
+
+catch.start:                                      ; preds = %catch.dispatch
+  %1 = catchpad within %0 [ptr null]
+  catchret from %1 to label %try.cont
+
+try.cont:                                         ; preds = %catch, %entry
+  ret void
+}
+
+declare void @foo()
+declare i32 @invalid_personality(...)