[mlir][ods] Verify access to operands in inferReturnTypes

[npanchen]

The patch adds graceful handling of incorrectly constructed MLIR operation with less operands than expected.

[github-actions]

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-core

Author: Nikolay Panchenko (npanchen)

Changes

The patch adds graceful handling of incorrectly constructed MLIR operation with less operands than expected.

---

Full diff: https://github.com/llvm/llvm-project/pull/112574.diff

2 Files Affected:

• (modified) mlir/test/mlir-tblgen/op-result.td (+7)
• (modified) mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp (+16)

diff --git a/mlir/test/mlir-tblgen/op-result.td b/mlir/test/mlir-tblgen/op-result.td
index 0ca570cf8cafba..51f8b0671a328d 100644
--- a/mlir/test/mlir-tblgen/op-result.td
+++ b/mlir/test/mlir-tblgen/op-result.td
@@ -130,6 +130,8 @@ def OpL1 : NS_Op<"op_with_all_types_constraint",
 
 // CHECK-LABEL: LogicalResult OpL1::inferReturnTypes
 // CHECK-NOT: }
+// CHECK: if (operands.size() <= 0)
+// CHECK-NEXT: return ::mlir::failure();
 // CHECK: ::mlir::Type odsInferredType0 = operands[0].getType();
 // CHECK: inferredReturnTypes[0] = odsInferredType0;
 
@@ -141,6 +143,9 @@ def OpL2 : NS_Op<"op_with_all_types_constraint",
 
 // CHECK-LABEL: LogicalResult OpL2::inferReturnTypes
 // CHECK-NOT: }
+// CHECK: if (operands.size() <= 2)
+// CHECK-NEXT: return ::mlir::failure();
+// CHECK-NOT: if (operands.size() <= 0)
 // CHECK: ::mlir::Type odsInferredType0 = operands[2].getType();
 // CHECK: ::mlir::Type odsInferredType1 = operands[0].getType();
 // CHECK: inferredReturnTypes[0] = odsInferredType0;
@@ -166,6 +171,8 @@ def OpL4 : NS_Op<"two_inference_edges", [
 }
 
 // CHECK-LABEL: LogicalResult OpL4::inferReturnTypes
+// CHECK: if (operands.size() <= 0)
+// CHECK-NEXT: return ::mlir::failure();
 // CHECK: odsInferredType0 = fromInput(operands[0].getType())
 // CHECK: odsInferredType1 = infer0(odsInferredType0)
 // CHECK: odsInferredType2 = infer1(odsInferredType1)
diff --git a/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp b/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp
index ce2b6ed94c3949..c55a00cf08a7cc 100644
--- a/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp
+++ b/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp
@@ -3583,6 +3583,22 @@ void OpEmitter::genTypeInterfaceMethods() {
   fctx.addSubst("_ctxt", "context");
   body << "  ::mlir::Builder odsBuilder(context);\n";
 
+  // Preprocessing stage to verify all accesses to operands are valid.
+  int maxAccessedIndex = -1;
+  for (int i = 0, e = op.getNumResults(); i != e; ++i) {
+    const InferredResultType &infer = op.getInferredResultType(i);
+    if (!infer.isArg())
+      continue;
+    auto arg = op.getArgToOperandOrAttribute(infer.getIndex());
+    if (arg.kind() == Operator::OperandOrAttribute::Kind::Operand)
+      maxAccessedIndex =
+          std::max(maxAccessedIndex, arg.operandOrAttributeIndex());
+  }
+  if (maxAccessedIndex != -1) {
+    body << "  if (operands.size() <= " << Twine(maxAccessedIndex) << ")\n";
+    body << "    return ::mlir::failure();\n";
+  }
+
   // Process the type inference graph in topological order, starting from types
   // that are always fully-inferred: operands and results with constructible
   // types. The type inference graph here will always be a DAG, so this gives

[weiweichen]

spell auto here please

[npanchen]

Apparently, even getArgToOperandOrAttribute is declared with auto, so no one wants to expose return type.

[joker-eph]

I see it declared as OperandOrAttribute getArgToOperandOrAttribute(int index) const; ?

[npanchen]

my bad. I thought it's pure automatic type deduction, so didn't look for ->

[weiweichen]

LGTM with minor comments. Not sure if I can approve MLIR changes though.

[github-actions]

@npanchen Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!