Emit hidden hostcall argument for sanitized kernels.

[pvellien]

Summary

this patch - https://reviews.llvm.org/D110337 changes the way how hostcall hidden argument is emitted for printf, but the sanitized kernels also use hostcall buffer to report a error for invalid memory access, which is not handled by the above patch and it leads to vdi runtime error.

:0:rocdevice.cpp :2581: 2190944203723 us: 520847: [tid:0x7fee1d7d1700] Device::callbackQueue aborting with error : HSA_STATUS_ERROR_MEMORY_FAULT: Agent attempted to access an inaccessible address. code: 0x2b
Aborted (core dumped)

[arsenm]

llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp

825

Space after //

llvm/test/CodeGen/AMDGPU/hsa-metadate-hostcall-present-v3-asan.ll

1–2 ↗

(On Diff #383349)

You can do this with a single llc invocation.

Also typo "metadate" in the filename

[pvellien]

Addressed review comments.

[pvellien]

rebased

[yaxunl]

You can add

getModule().addModuleFlag(llvm::Module::Override, "amdgpu_hostcall", 1);

at https://github.com/llvm/llvm-project/blob/b2a2c38349a18b89b04d342632d5ea02f86dfdd6/clang/lib/CodeGen/CodeGenModule.cpp#L542

It will be a more consistent and efficient way to represent the need for hostcall.

[pvellien]

Addressed comments.

[yaxunl]

clang/lib/CodeGen/CodeGenModule.cpp

547 ↗

(On Diff #384163)

need to check this in clang/test/CodeGenCUDA/amdgpu-asan.cu

[pvellien]

Added a check in amdgpu-asan.cu file

[yaxunl]

LGTM. Thanks.

[pvellien]

Thanks for accepting. I don't have commit access to llvm trunk yet. Could you please commit this patch? thanks.

llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp

auto Int8PtrTy =
       Type::getInt8PtrTy(Func.getContext(), AMDGPUAS::GLOBAL_ADDRESS);
 
   // Emit "printf buffer" argument if printf is used, emit "hostcall buffer"
   // if "hostcall" module flag is set, otherwise emit dummy "none" argument.
   if (HiddenArgNumBytes >= 32) {
     if (M->getNamedMetadata("llvm.printf.fmts"))
       emitKernelArg(DL, Int8PtrTy, Align(8), "hidden_printf_buffer", Offset,
                     Args);
-    else if (M->getModuleFlag("amdgpu_hostcall")) {
+    else if (M->getModuleFlag("amdgpu_hostcall") ||
+             // Emit hidden_hostcall_buffer argument for address sanitized
+             // functions since they use hostcall buffer to report invalid
+             // memory access.
+             Func.hasFnAttribute(llvm::Attribute::SanitizeAddress)) {
       // The printf runtime binding pass should have ensured that hostcall and
       // printf are not used in the same module.
       assert(!M->getNamedMetadata("llvm.printf.fmts"));
       emitKernelArg(DL, Int8PtrTy, Align(8), "hidden_hostcall_buffer", Offset,
                     Args);
     } else
       emitKernelArg(DL, Int8PtrTy, Align(8), "hidden_none", Offset, Args);
   }

llvm/test/CodeGen/AMDGPU/hsa-metadata-hostcall-present-v3-asan.ll

+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 --amdhsa-code-object-version=3 -amdgpu-dump-hsa-metadata -amdgpu-verify-hsa-metadata -filetype=obj -o - < %s 2>&1 | FileCheck  %s
+
+; CHECK:              ---
+; CHECK:      amdhsa.kernels:
+; CHECK:        - .args:
+; CHECK-NEXT:       - .name:           a
+; CHECK-NEXT:         .offset:         0
+; CHECK-NEXT:         .size:           1
+; CHECK-NEXT:         .type_name:      char
+; CHECK-NEXT:         .value_kind:     by_value
+; CHECK-NEXT:       - .offset:         8
+; CHECK-NEXT:         .size:           8
+; CHECK-NEXT:         .value_kind:     hidden_global_offset_x
+; CHECK-NEXT:       - .offset:         16
+; CHECK-NEXT:         .size:           8
+; CHECK-NEXT:         .value_kind:     hidden_global_offset_y
+; CHECK-NEXT:       - .offset:         24
+; CHECK-NEXT:         .size:           8
+; CHECK-NEXT:         .value_kind:     hidden_global_offset_z
+; CHECK-NEXT:       - .address_space:  global
+; CHECK-NEXT:         .offset:         32
+; CHECK-NEXT:         .size:           8
+; CHECK-NEXT:         .value_kind:     hidden_hostcall_buffer
+; CHECK:          .language:       OpenCL C
+; CHECK-NEXT:     .language_version:
+; CHECK-NEXT:       - 2
+; CHECK-NEXT:       - 0
+; CHECK:          .name:           test_kernel
+; CHECK:          .symbol:         test_kernel.kd
+
+define amdgpu_kernel void @test_kernel(i8 %a) #0
+    !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !3
+    !kernel_arg_base_type !3 !kernel_arg_type_qual !4 {
+  ret void
+}
+
+; CHECK:  amdhsa.version:
+; CHECK-NEXT: - 1
+; CHECK-NEXT: - 0
+
+attributes #0 = { sanitize_address "amdgpu-implicitarg-num-bytes"="48" }
+
+!1 = !{i32 0}
+!2 = !{!"none"}
+!3 = !{!"char"}
+!4 = !{!""}
+
+!opencl.ocl.version = !{!90}
+!90 = !{i32 2, i32 0}
+
+; CHECK: AMDGPU HSA Metadata Parser Test: PASS