Adding changes for handling abort signals (#979)

* Adding changes for handling abort signals

* Fix the test failure

* Fixing CmakeLists error

* Addressing review comments

* fixing warnings

* fixing execute test

* Fixing abort app test

* Address review comments

* Apply suggestions from code review

* Apply suggestions from code review

* Fixes for testing issues

* Adding kernel filtering test

* Removing text input file

* fix formatting issues

* misc fix

* Suppress signal-unsafe error in ThreadSanitizer

- rename signal handler to rocprofv3_error_signal_handler to ensure specific filtering

* Fix rocprofv3 aborted-app validation

---------

Co-authored-by: Jonathan R. Madsen <jrmadsen@users.noreply.github.com>
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>

tests/rocprofv3/CMakeLists.txt

@@ -29,3 +29,4 @@ add_subdirectory(tracing-hip-in-libraries)
 add_subdirectory(counter-collection)
 add_subdirectory(hsa-queue-dependency)
 add_subdirectory(kernel-rename)
+add_subdirectory(aborted-app)

tests/rocprofv3/aborted-app/CMakeLists.txt

@@ -0,0 +1,44 @@
+#
+# rocprofv3 tool test
+#
+cmake_minimum_required(VERSION 3.21.0 FATAL_ERROR)
+
+project(
+    rocprofiler-tests-aborted-app
+    LANGUAGES CXX
+    VERSION 0.0.0)
+
+find_package(rocprofiler-sdk REQUIRED)
+
+rocprofiler_configure_pytest_files(CONFIG pytest.ini COPY validate.py conftest.py
+                                                          input.json)
+
+string(REPLACE "LD_PRELOAD=" "ROCPROF_PRELOAD=" PRELOAD_ENV
+               "${ROCPROFILER_MEMCHECK_PRELOAD_ENV}")
+
+set(aborted-app-env "${PRELOAD_ENV}" ROCPROF_TESTING_RAISE_SIGNAL=1
+                    ROCPROF_INTERNAL_TEST_SIGNAL_HANDLER_VIA_EXIT=1)
+
+# app-abort
+add_test(
+    NAME rocprofv3-test-execute-app-abort
+    COMMAND
+        $<TARGET_FILE:rocprofiler-sdk::rocprofv3> -i
+        ${CMAKE_CURRENT_BINARY_DIR}/input.json -d
+        ${CMAKE_CURRENT_BINARY_DIR}/out-aborted-app -- $<TARGET_FILE:vector-ops> 1 1)
+
+set_tests_properties(
+    rocprofv3-test-execute-app-abort
+    PROPERTIES TIMEOUT 45 LABELS "integration-tests" ENVIRONMENT "${aborted-app-env}"
+               WILL_FAIL TRUE)
+
+add_test(
+    NAME rocprofv3-test-validate-app-abort
+    COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_BINARY_DIR}/validate.py --json-input
+            ${CMAKE_CURRENT_BINARY_DIR}/out-aborted-app/pass_1/out_results.json)
+
+set_tests_properties(
+    rocprofv3-test-validate-app-abort
+    PROPERTIES TIMEOUT 45 LABELS "integration-tests" DEPENDS
+               "rocprofv3-test-execute-app-abort" FAIL_REGULAR_EXPRESSION
+               "${ROCPROFILER_DEFAULT_FAIL_REGEX}")

tests/rocprofv3/aborted-app/conftest.py

@@ -0,0 +1,23 @@
+#!/usr/bin/env python3
+
+import json
+import pytest
+import pandas as pd
+
+from rocprofiler_sdk.pytest_utils.dotdict import dotdict
+from rocprofiler_sdk.pytest_utils import collapse_dict_list
+
+
+def pytest_addoption(parser):
+    parser.addoption(
+        "--json-input",
+        action="store",
+        help="Path to JSON file.",
+    )
+
+
+@pytest.fixture
+def json_data(request):
+    filename = request.config.getoption("--json-input")
+    with open(filename, "r") as inp:
+        return dotdict(collapse_dict_list(json.load(inp)))

tests/rocprofv3/aborted-app/input.json

@@ -0,0 +1,16 @@
+{
+    "jobs": [
+        {
+            "pmc":["SQ_WAVES"],
+            "kernel_include_regex": "addition",
+            "kernel_exclude_regex": "subtract",
+            "hip_runtime_trace": true,
+            "kernel_trace": true,
+            "output_file": "out",
+            "output_format": [
+                "json"
+            ],
+            "truncate_kernels": true
+        }
+    ]
+}

tests/rocprofv3/aborted-app/pytest.ini

@@ -0,0 +1,5 @@
+
+[pytest]
+addopts = --durations=20 -rA -s -vv
+testpaths = validate.py
+pythonpath = @ROCPROFILER_SDK_TESTS_BINARY_DIR@/pytest-packages

tests/rocprofv3/aborted-app/validate.py

@@ -0,0 +1,108 @@
+#!/usr/bin/env python3
+
+import sys
+import pytest
+import re
+
+kernel_trace_list = sorted(["addition_kernel", "subtract_kernel"])
+kernel_counter_list = ["addition_kernel"]
+
+
+def unique(lst):
+    return list(set(lst))
+
+
+def test_counter_collection_json_data(json_data):
+    data = json_data["rocprofiler-sdk-tool"]
+
+    def get_kernel_name(kernel_id):
+        return data["kernel_symbols"][kernel_id]["formatted_kernel_name"]
+
+    counter_collection_data = data["callback_records"]["counter_collection"]
+
+    for counter in counter_collection_data:
+        kernel_name = get_kernel_name(counter.dispatch_data.dispatch_info.kernel_id)
+        assert kernel_name in kernel_counter_list
+
+
+def test_kernel_trace_json(json_data):
+    data = json_data["rocprofiler-sdk-tool"]
+
+    def get_kernel_name(kernel_id):
+        return data["kernel_symbols"][kernel_id]["formatted_kernel_name"]
+
+    def get_kind_name(kind_id):
+        return data["strings"]["buffer_records"][kind_id]["kind"]
+
+    kernel_dispatch_data = data["buffer_records"]["kernel_dispatch"]
+    kernels = []
+    assert len(kernel_dispatch_data) == 2
+    for dispatch in kernel_dispatch_data:
+        dispatch_info = dispatch["dispatch_info"]
+        kernel_name = get_kernel_name(dispatch_info["kernel_id"])
+
+        assert get_kind_name(dispatch["kind"]) == "KERNEL_DISPATCH"
+        assert dispatch["correlation_id"]["internal"] > 0
+        assert dispatch_info["agent_id"]["handle"] > 0
+        assert dispatch_info["queue_id"]["handle"] > 0
+        assert dispatch_info["kernel_id"] > 0
+        if not re.search(r"__amd_rocclr_.*", kernel_name):
+            kernels.append(kernel_name)
+
+        assert dispatch_info["workgroup_size"]["x"] == 64
+        assert dispatch_info["workgroup_size"]["y"] == 1
+        assert dispatch_info["workgroup_size"]["z"] == 1
+        assert dispatch_info["grid_size"]["x"] == 1024
+        assert dispatch_info["grid_size"]["y"] == 1024
+        assert dispatch_info["grid_size"]["z"] == 1
+        assert dispatch["end_timestamp"] >= dispatch["start_timestamp"]
+
+    assert kernels == kernel_trace_list
+
+
+def test_hip_api_trace_json(json_data):
+    data = json_data["rocprofiler-sdk-tool"]
+
+    def get_operation_name(kind_id, op_id):
+        return data["strings"]["buffer_records"][kind_id]["operations"][op_id]
+
+    def get_kind_name(kind_id):
+        return data["strings"]["buffer_records"][kind_id]["kind"]
+
+    valid_domain_names = ("HIP_RUNTIME_API",)
+
+    hip_api_data = data["buffer_records"]["hip_api"]
+
+    functions = []
+    for api in hip_api_data:
+        kind = get_kind_name(api["kind"])
+        assert kind in valid_domain_names
+        assert api["end_timestamp"] >= api["start_timestamp"]
+        functions.append(get_operation_name(api["kind"], api["operation"]))
+
+    expected_functions = (
+        [
+            "hipGetDeviceCount",
+            "hipSetDevice",
+            "hipDeviceSynchronize",
+            "hipStreamCreateWithFlags",
+        ]
+        + (["hipHostMalloc"] * 3)
+        + (["hipMallocAsync"] * 3)
+        + (["hipMemcpyAsync"] * 2)
+        + [
+            "hipStreamSynchronize",
+            "hipDeviceSynchronize",
+            "hipLaunchKernel",
+            "hipGetLastError",
+            "hipLaunchKernel",
+            "hipGetLastError",
+        ]
+    )
+
+    assert functions == expected_functions
+
+
+if __name__ == "__main__":
+    exit_code = pytest.main(["-x", __file__] + sys.argv[1:])
+    sys.exit(exit_code)