fix: don't fail entire initialization when at least one device is com…

…patible Related-To: NEO-6683 Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
intel · Nov 21, 2023 · f63dd1f · f63dd1f
1 parent 490f1d9
commit f63dd1f
Show file tree

Hide file tree

Showing 5 changed files with 89 additions and 3 deletions.
diff --git a/shared/source/execution_environment/root_device_environment.h b/shared/source/execution_environment/root_device_environment.h
@@ -61,7 +61,7 @@ struct RootDeviceEnvironment : NonCopyableClass {
     bool isWddmOnLinux() const;
 
     MOCKABLE_VIRTUAL void initAubCenter(bool localMemoryEnabled, const std::string &aubFileName, CommandStreamReceiverType csrType);
-    bool initOsInterface(std::unique_ptr<HwDeviceId> &&hwDeviceId, uint32_t rootDeviceIndex);
+    MOCKABLE_VIRTUAL bool initOsInterface(std::unique_ptr<HwDeviceId> &&hwDeviceId, uint32_t rootDeviceIndex);
     void initOsTime();
     void initGmm();
     void initDebuggerL0(Device *neoDevice);

diff --git a/shared/source/os_interface/device_factory.cpp b/shared/source/os_interface/device_factory.cpp
@@ -170,12 +170,18 @@ bool DeviceFactory::prepareDeviceEnvironments(ExecutionEnvironment &executionEnv
 
     for (auto &hwDeviceId : hwDeviceIds) {
         if (initHwDeviceIdResources(executionEnvironment, std::move(hwDeviceId), rootDeviceIndex) == false) {
-            return false;
+            continue;
         }
 
         rootDeviceIndex++;
     }
 
+    executionEnvironment.rootDeviceEnvironments.resize(rootDeviceIndex);
+
+    if (rootDeviceIndex == 0) {
+        return false;
+    }
+
     executionEnvironment.setDeviceHierarchy(executionEnvironment.rootDeviceEnvironments[0]->getHelper<GfxCoreHelper>());
     executionEnvironment.sortNeoDevices();
     executionEnvironment.parseAffinityMask();

diff --git a/shared/test/common/mocks/mock_execution_environment.cpp b/shared/test/common/mocks/mock_execution_environment.cpp
@@ -10,6 +10,8 @@
 #include "shared/test/common/fixtures/mock_aub_center_fixture.h"
 #include "shared/test/common/helpers/default_hw_info.h"
 
+#include "gtest/gtest.h"
+
 namespace NEO {
 extern bool useMockGmm;
 
@@ -24,6 +26,23 @@ void MockRootDeviceEnvironment::initAubCenter(bool localMemoryEnabled, const std
     }
     RootDeviceEnvironment::initAubCenter(localMemoryEnabled, aubFileName, csrType);
 }
+bool MockRootDeviceEnvironment::initOsInterface(std::unique_ptr<HwDeviceId> &&hwDeviceId, uint32_t rootDeviceIndex) {
+    initOsInterfaceCalled++;
+    if (!initOsInterfaceResults.empty()) {
+        auto retVal = initOsInterfaceResults[initOsInterfaceCalled - 1];
+        if (retVal) {
+            RootDeviceEnvironment::initOsInterface(std::move(hwDeviceId), rootDeviceIndex);
+        }
+        return retVal;
+    }
+    return RootDeviceEnvironment::initOsInterface(std::move(hwDeviceId), rootDeviceIndex);
+}
+
+MockRootDeviceEnvironment::~MockRootDeviceEnvironment() {
+    if (initOsInterfaceExpectedCallCount) {
+        EXPECT_EQ(*initOsInterfaceExpectedCallCount, initOsInterfaceCalled);
+    }
+}
 
 MockExecutionEnvironment::MockExecutionEnvironment() : MockExecutionEnvironment(defaultHwInfo.get()) {}
 MockExecutionEnvironment::MockExecutionEnvironment(const HardwareInfo *hwInfo) : MockExecutionEnvironment(hwInfo, true, 1u) {

diff --git a/shared/test/common/mocks/mock_execution_environment.h b/shared/test/common/mocks/mock_execution_environment.h
@@ -10,16 +10,24 @@
 #include "shared/source/execution_environment/execution_environment.h"
 #include "shared/source/execution_environment/root_device_environment.h"
 
+#include <optional>
+
 namespace NEO {
 
 struct MockRootDeviceEnvironment : public RootDeviceEnvironment {
+    using BaseClass = RootDeviceEnvironment;
     using RootDeviceEnvironment::hwInfo;
     using RootDeviceEnvironment::isDummyAllocationInitialized;
     using RootDeviceEnvironment::isWddmOnLinuxEnable;
     using RootDeviceEnvironment::RootDeviceEnvironment;
-    ~MockRootDeviceEnvironment() override = default;
+    ~MockRootDeviceEnvironment() override;
 
     void initAubCenter(bool localMemoryEnabled, const std::string &aubFileName, CommandStreamReceiverType csrType) override;
+    bool initOsInterface(std::unique_ptr<HwDeviceId> &&hwDeviceId, uint32_t rootDeviceIndex) override;
+
+    std::vector<bool> initOsInterfaceResults;
+    uint32_t initOsInterfaceCalled = 0u;
+    std::optional<uint32_t> initOsInterfaceExpectedCallCount;
     bool initAubCenterCalled = false;
     bool localMemoryEnabledReceived = false;
     std::string aubFileNameReceived = "";

diff --git a/shared/test/unit_test/os_interface/device_factory_tests.cpp b/shared/test/unit_test/os_interface/device_factory_tests.cpp
@@ -121,3 +121,56 @@ TEST_F(DeviceFactoryTests, givenDisabledRcsWhenPrepareDeviceEnvironmentsCalledTh
         EXPECT_NE(executionEnvironment.rootDeviceEnvironments[0]->getHardwareInfo()->featureTable.flags.ftrRcsNode, releaseHelper->isRcsExposureDisabled());
     }
 }
+
+TEST_F(DeviceFactoryTests, givenMultipleDevicesWhenInitializeResourcesSucceedsForAtLeastOneDeviceThenSuccessIsReturned) {
+    DebugManagerStateRestore restorer;
+    DebugManager.flags.CreateMultipleRootDevices.set(3);
+    MockExecutionEnvironment executionEnvironment(defaultHwInfo.get(), true, 3u);
+
+    EXPECT_EQ(3u, executionEnvironment.rootDeviceEnvironments.size());
+    auto rootDeviceEnvironment0 = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[0].get());
+    auto rootDeviceEnvironment1 = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[1].get());
+    auto rootDeviceEnvironment2 = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[2].get());
+
+    rootDeviceEnvironment0->initOsInterfaceResults.push_back(true);
+    rootDeviceEnvironment0->initOsInterfaceExpectedCallCount = 1u;
+
+    // making rootDeviceEnvironment1 returning false on first call and true on second call
+    rootDeviceEnvironment1->initOsInterfaceResults.push_back(false);
+    rootDeviceEnvironment1->initOsInterfaceResults.push_back(true);
+    rootDeviceEnvironment1->initOsInterfaceExpectedCallCount = 2u;
+
+    rootDeviceEnvironment2->initOsInterfaceExpectedCallCount = 0u;
+
+    bool success = DeviceFactory::prepareDeviceEnvironments(executionEnvironment);
+    ASSERT_TRUE(success);
+
+    EXPECT_EQ(2u, executionEnvironment.rootDeviceEnvironments.size());
+
+    EXPECT_EQ(1u, rootDeviceEnvironment0->initOsInterfaceCalled);
+    EXPECT_EQ(2u, rootDeviceEnvironment1->initOsInterfaceCalled);
+}
+
+TEST_F(DeviceFactoryTests, givenMultipleDevicesWhenInitializeResourcesFailsForAllDevicesThenFailureIsReturned) {
+    DebugManagerStateRestore restorer;
+    DebugManager.flags.CreateMultipleRootDevices.set(3);
+    MockExecutionEnvironment executionEnvironment(defaultHwInfo.get(), true, 3u);
+
+    EXPECT_EQ(3u, executionEnvironment.rootDeviceEnvironments.size());
+    auto rootDeviceEnvironment0 = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[0].get());
+    auto rootDeviceEnvironment1 = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[1].get());
+    auto rootDeviceEnvironment2 = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[2].get());
+
+    // making root device environment failing three times (once per device)
+    rootDeviceEnvironment0->initOsInterfaceResults.push_back(false);
+    rootDeviceEnvironment0->initOsInterfaceResults.push_back(false);
+    rootDeviceEnvironment0->initOsInterfaceResults.push_back(false);
+    rootDeviceEnvironment0->initOsInterfaceExpectedCallCount = 3u;
+    rootDeviceEnvironment1->initOsInterfaceExpectedCallCount = 0u;
+    rootDeviceEnvironment2->initOsInterfaceExpectedCallCount = 0u;
+
+    bool success = DeviceFactory::prepareDeviceEnvironments(executionEnvironment);
+    EXPECT_FALSE(success);
+
+    EXPECT_EQ(0u, executionEnvironment.rootDeviceEnvironments.size());
+}