fix: Check if provided CCS number is correct

Check if `ZEX_NUMBER_OF_CCS` env variable provided by the user is
correct. If it isn't then return false and print debug message.

Related-To: NEO-15230, GSD-11251
Signed-off-by: Kindracki, Jakub Tomasz <jakub.tomasz.kindracki@intel.com>

shared/source/execution_environment/execution_environment.cpp

@@ -354,8 +354,10 @@ void ExecutionEnvironment::adjustCcsCountImpl(RootDeviceEnvironment *rootDeviceE
     productHelper.adjustNumberOfCcs(*hwInfo);
 }
 
-void ExecutionEnvironment::adjustCcsCount() {
-    parseCcsCountLimitations();
+bool ExecutionEnvironment::adjustCcsCount() {
+    if (!parseCcsCountLimitations()) {
+        return false;
+    }
 
     for (auto rootDeviceIndex = 0u; rootDeviceIndex < rootDeviceEnvironments.size(); rootDeviceIndex++) {
         auto &rootDeviceEnvironment = rootDeviceEnvironments[rootDeviceIndex];
@@ -364,32 +366,42 @@ void ExecutionEnvironment::adjustCcsCount() {
             adjustCcsCountImpl(rootDeviceEnvironment.get());
         }
     }
+
+    return true;
 }
 
-void ExecutionEnvironment::adjustCcsCount(const uint32_t rootDeviceIndex) const {
+bool ExecutionEnvironment::adjustCcsCount(const uint32_t rootDeviceIndex) const {
     auto &rootDeviceEnvironment = rootDeviceEnvironments[rootDeviceIndex];
     UNRECOVERABLE_IF(!rootDeviceEnvironment);
     if (rootDeviceNumCcsMap.find(rootDeviceIndex) != rootDeviceNumCcsMap.end()) {
-        rootDeviceEnvironment->setNumberOfCcs(rootDeviceNumCcsMap.at(rootDeviceIndex));
+        if (!rootDeviceEnvironment->setNumberOfCcs(rootDeviceNumCcsMap.at(rootDeviceIndex))) {
+            return false;
+        }
     } else {
         adjustCcsCountImpl(rootDeviceEnvironment.get());
     }
+
+    return true;
 }
 
-void ExecutionEnvironment::parseCcsCountLimitations() {
+bool ExecutionEnvironment::parseCcsCountLimitations() {
     const auto &numberOfCcsString = debugManager.flags.ZEX_NUMBER_OF_CCS.get();
 
     if (numberOfCcsString.compare("default") == 0 ||
         numberOfCcsString.empty()) {
-        return;
+        return true;
     }
 
     for (auto rootDeviceIndex = 0u; rootDeviceIndex < rootDeviceEnvironments.size(); rootDeviceIndex++) {
         auto &rootDeviceEnvironment = rootDeviceEnvironments[rootDeviceIndex];
         UNRECOVERABLE_IF(!rootDeviceEnvironment);
         auto &productHelper = rootDeviceEnvironment->getHelper<ProductHelper>();
-        productHelper.parseCcsMode(numberOfCcsString, rootDeviceNumCcsMap, rootDeviceIndex, rootDeviceEnvironment.get());
+        if (!productHelper.parseCcsMode(numberOfCcsString, rootDeviceNumCcsMap, rootDeviceIndex, rootDeviceEnvironment.get())) {
+            return false;
+        }
     }
+
+    return true;
 }
 
 void ExecutionEnvironment::configureNeoEnvironment() {

shared/source/execution_environment/execution_environment.h

@@ -37,8 +37,8 @@ class ExecutionEnvironment : public ReferenceTrackedObject<ExecutionEnvironment>
     virtual void prepareRootDeviceEnvironments(uint32_t numRootDevices);
     void prepareRootDeviceEnvironment(const uint32_t rootDeviceIndexForReInit);
     void parseAffinityMask();
-    void adjustCcsCount();
-    void adjustCcsCount(const uint32_t rootDeviceIndex) const;
+    bool adjustCcsCount();
+    bool adjustCcsCount(const uint32_t rootDeviceIndex) const;
     void sortNeoDevices();
     void setDeviceHierarchyMode(const GfxCoreHelper &gfxCoreHelper);
     void setDeviceHierarchyMode(const DeviceHierarchyMode deviceHierarchyMode) {
@@ -47,7 +47,7 @@ class ExecutionEnvironment : public ReferenceTrackedObject<ExecutionEnvironment>
     DeviceHierarchyMode getDeviceHierarchyMode() const { return deviceHierarchyMode; }
     void adjustRootDeviceEnvironments();
     void prepareForCleanup() const;
-    void configureCcsMode();
+    MOCKABLE_VIRTUAL void configureCcsMode();
     void setDebuggingMode(DebuggingMode debuggingMode) {
         debuggingEnabledMode = debuggingMode;
     }
@@ -97,7 +97,7 @@ class ExecutionEnvironment : public ReferenceTrackedObject<ExecutionEnvironment>
 
   protected:
     static bool comparePciIdBusNumber(std::unique_ptr<RootDeviceEnvironment> &rootDeviceEnvironment1, std::unique_ptr<RootDeviceEnvironment> &rootDeviceEnvironment2);
-    void parseCcsCountLimitations();
+    bool parseCcsCountLimitations();
     void adjustCcsCountImpl(RootDeviceEnvironment *rootDeviceEnvironment) const;
     void configureNeoEnvironment();
     void restoreCcsMode();

shared/source/execution_environment/root_device_environment.cpp

@@ -246,12 +246,19 @@ BuiltIns *RootDeviceEnvironment::getBuiltIns() {
     return this->builtins.get();
 }
 
-void RootDeviceEnvironment::setNumberOfCcs(uint32_t numberOfCcs) {
-    hwInfo->gtSystemInfo.CCSInfo.NumberOfCCSEnabled = std::min(hwInfo->gtSystemInfo.CCSInfo.NumberOfCCSEnabled, numberOfCcs);
+bool RootDeviceEnvironment::setNumberOfCcs(uint32_t numberOfCcs) {
+    if (hwInfo->gtSystemInfo.CCSInfo.NumberOfCCSEnabled < numberOfCcs || numberOfCcs == 0) {
+        NEO::printDebugString(NEO::debugManager.flags.PrintDebugMessages.get(), stderr, "Error: Invalid number of CCS: %u. Maximum available number of CCS: %u\n", numberOfCcs, hwInfo->gtSystemInfo.CCSInfo.NumberOfCCSEnabled);
+        return false;
+    }
+
+    hwInfo->gtSystemInfo.CCSInfo.NumberOfCCSEnabled = numberOfCcs;
     limitedNumberOfCcs = true;
     if (aubCenter) {
         aubCenter->getAubManager()->setCCSMode(hwInfo->gtSystemInfo.CCSInfo.NumberOfCCSEnabled);
     }
+
+    return true;
 }
 
 uint32_t RootDeviceEnvironment::getNumberOfCcs() const {

shared/source/execution_environment/root_device_environment.h

@@ -81,7 +81,7 @@ struct RootDeviceEnvironment : NonCopyableClass {
     BindlessHeapsHelper *getBindlessHeapsHelper() const;
     AssertHandler *getAssertHandler(Device *neoDevice);
     void createBindlessHeapsHelper(Device *rootDevice, bool availableDevices);
-    void setNumberOfCcs(uint32_t numberOfCcs);
+    bool setNumberOfCcs(uint32_t numberOfCcs);
     uint32_t getNumberOfCcs() const;
     bool isNumberOfCcsLimited() const;
     void setRcsExposure();

shared/source/os_interface/device_factory.cpp

@@ -164,7 +164,9 @@ bool DeviceFactory::prepareDeviceEnvironmentsForProductFamilyOverride(ExecutionE
 
     executionEnvironment.setDeviceHierarchyMode(executionEnvironment.rootDeviceEnvironments[0]->getHelper<GfxCoreHelper>());
     executionEnvironment.parseAffinityMask();
-    executionEnvironment.adjustCcsCount();
+    if (!executionEnvironment.adjustCcsCount()) {
+        return false;
+    }
     executionEnvironment.calculateMaxOsContextCount();
     return true;
 }
@@ -272,7 +274,9 @@ bool DeviceFactory::prepareDeviceEnvironments(ExecutionEnvironment &executionEnv
     executionEnvironment.sortNeoDevices();
     executionEnvironment.parseAffinityMask();
     executionEnvironment.adjustRootDeviceEnvironments();
-    executionEnvironment.adjustCcsCount();
+    if (!executionEnvironment.adjustCcsCount()) {
+        return false;
+    }
     executionEnvironment.calculateMaxOsContextCount();
 
     return true;
@@ -294,7 +298,9 @@ bool DeviceFactory::prepareDeviceEnvironment(ExecutionEnvironment &executionEnvi
         return false;
     }
 
-    executionEnvironment.adjustCcsCount(rootDeviceIndex);
+    if (!executionEnvironment.adjustCcsCount(rootDeviceIndex)) {
+        return false;
+    }
     return true;
 }
 

shared/source/os_interface/product_helper.h

@@ -227,7 +227,7 @@ class ProductHelper {
     virtual void fillPipelineSelectPropertiesSupportStructure(PipelineSelectPropertiesSupport &propertiesSupport, const HardwareInfo &hwInfo) const = 0;
     virtual void fillStateBaseAddressPropertiesSupportStructure(StateBaseAddressPropertiesSupport &propertiesSupport) const = 0;
 
-    virtual void parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const = 0;
+    virtual bool parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const = 0;
 
     virtual bool isFusedEuDisabledForDpas(bool kernelHasDpasInstructions, const uint32_t *lws, const uint32_t *groupCount, const HardwareInfo &hwInfo) const = 0;
     virtual bool isCalculationForDisablingEuFusionWithDpasNeeded(const HardwareInfo &hwInfo) const = 0;

shared/source/os_interface/product_helper.inl

@@ -715,12 +715,16 @@ void ProductHelperHw<gfxProduct>::fillStateBaseAddressPropertiesSupportStructure
 }
 
 template <PRODUCT_FAMILY gfxProduct>
-void ProductHelperHw<gfxProduct>::parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const {
+bool ProductHelperHw<gfxProduct>::parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const {
 
     auto ccsCount = StringHelpers::toUint32t(ccsModeString);
 
     rootDeviceNumCcsMap.insert({rootDeviceIndex, ccsCount});
-    rootDeviceEnvironment->setNumberOfCcs(ccsCount);
+    if (!rootDeviceEnvironment->setNumberOfCcs(ccsCount)) {
+        return false;
+    }
+
+    return true;
 }
 
 template <PRODUCT_FAMILY gfxProduct>

shared/source/os_interface/product_helper_hw.h

@@ -165,7 +165,7 @@ class ProductHelperHw : public ProductHelper {
     void fillFrontEndPropertiesSupportStructure(FrontEndPropertiesSupport &propertiesSupport, const HardwareInfo &hwInfo) const override;
     void fillPipelineSelectPropertiesSupportStructure(PipelineSelectPropertiesSupport &propertiesSupport, const HardwareInfo &hwInfo) const override;
     void fillStateBaseAddressPropertiesSupportStructure(StateBaseAddressPropertiesSupport &propertiesSupport) const override;
-    void parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const override;
+    bool parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const override;
 
     bool isFusedEuDisabledForDpas(bool kernelHasDpasInstructions, const uint32_t *lws, const uint32_t *groupCount, const HardwareInfo &hwInfo) const override;
     bool isCalculationForDisablingEuFusionWithDpasNeeded(const HardwareInfo &hwInfo) const override;

shared/source/xe_hpc_core/pvc/os_agnostic_product_helper_pvc.inl

@@ -153,21 +153,28 @@ bool ProductHelperHw<gfxProduct>::isBlitCopyRequiredForLocalMemory(const RootDev
 }
 
 template <>
-void ProductHelperHw<gfxProduct>::parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const {
+bool ProductHelperHw<gfxProduct>::parseCcsMode(std::string ccsModeString, std::unordered_map<uint32_t, uint32_t> &rootDeviceNumCcsMap, uint32_t rootDeviceIndex, RootDeviceEnvironment *rootDeviceEnvironment) const {
     auto numberOfCcsEntries = StringHelpers::split(ccsModeString, ",");
 
     for (const auto &entry : numberOfCcsEntries) {
         auto subEntries = StringHelpers::split(entry, ":");
+        if (subEntries.size() < 2) {
+            NEO::printDebugString(NEO::debugManager.flags.PrintDebugMessages.get(), stderr, "Error: Invalid ZEX_NUMBER_OF_CCS format '%s'\n", ccsModeString.c_str());
+            return false;
+        }
+
         uint32_t rootDeviceIndexParsed = StringHelpers::toUint32t(subEntries[0]);
 
         if (rootDeviceIndexParsed == rootDeviceIndex) {
-            if (subEntries.size() > 1) {
-                uint32_t maxCcsCount = StringHelpers::toUint32t(subEntries[1]);
-                rootDeviceNumCcsMap.insert({rootDeviceIndex, maxCcsCount});
-                rootDeviceEnvironment->setNumberOfCcs(maxCcsCount);
+            uint32_t maxCcsCount = StringHelpers::toUint32t(subEntries[1]);
+            rootDeviceNumCcsMap.insert({rootDeviceIndex, maxCcsCount});
+            if (!rootDeviceEnvironment->setNumberOfCcs(maxCcsCount)) {
+                return false;
             }
         }
     }
+
+    return true;
 }
 
 template <>

shared/test/common/mocks/mock_execution_environment.cpp

@@ -91,4 +91,8 @@ void MockExecutionEnvironment::initGmm() {
     }
 }
 
+void MockExecutionEnvironment::addToRootDeviceNumCcsMap(uint32_t rootDeviceIndex, uint32_t numCcs) {
+    this->rootDeviceNumCcsMap.insert({rootDeviceIndex, numCcs});
+}
+
 } // namespace NEO

shared/test/common/mocks/mock_execution_environment.h

@@ -52,6 +52,7 @@ struct MockExecutionEnvironment : ExecutionEnvironment {
     MockExecutionEnvironment(const HardwareInfo *hwInfo);
     MockExecutionEnvironment(const HardwareInfo *hwInfo, bool useMockAubCenter, uint32_t numRootDevices);
     void initGmm();
+    void addToRootDeviceNumCcsMap(uint32_t rootDeviceIndex, uint32_t numCcs);
 };
 
 } // namespace NEO

shared/test/unit_test/execution_environment/execution_environment_tests.cpp

@@ -780,6 +780,49 @@ TEST(ExecutionEnvironment, givenExecutionEnvironmentWhenSetDevicePermissionError
     EXPECT_FALSE(executionEnvironment.isDevicePermissionError());
 }
 
+TEST(ExecutionEnvironment, givenExecutionEnvironmentWhenCcsNumberIsInvalidThenAdjustCcsCountReturnsFalse) {
+    {
+        DebugManagerStateRestore restorer;
+        MockExecutionEnvironment executionEnvironment;
+        debugManager.flags.ZEX_NUMBER_OF_CCS.set("0");
+
+        EXPECT_FALSE(executionEnvironment.adjustCcsCount());
+    }
+    {
+        DebugManagerStateRestore restorer;
+        MockExecutionEnvironment executionEnvironment;
+        debugManager.flags.ZEX_NUMBER_OF_CCS.set("0:0");
+
+        EXPECT_FALSE(executionEnvironment.adjustCcsCount());
+    }
+    {
+        DebugManagerStateRestore restorer;
+        MockExecutionEnvironment executionEnvironment;
+        debugManager.flags.ZEX_NUMBER_OF_CCS.set("100");
+
+        EXPECT_FALSE(executionEnvironment.adjustCcsCount());
+    }
+    {
+        DebugManagerStateRestore restorer;
+        MockExecutionEnvironment executionEnvironment;
+        debugManager.flags.ZEX_NUMBER_OF_CCS.set("0:100");
+
+        EXPECT_FALSE(executionEnvironment.adjustCcsCount());
+    }
+    {
+        MockExecutionEnvironment executionEnvironment;
+        executionEnvironment.addToRootDeviceNumCcsMap(0, 0);
+
+        EXPECT_FALSE(executionEnvironment.adjustCcsCount(0));
+    }
+    {
+        MockExecutionEnvironment executionEnvironment;
+        executionEnvironment.addToRootDeviceNumCcsMap(0, 100);
+
+        EXPECT_FALSE(executionEnvironment.adjustCcsCount(0));
+    }
+}
+
 void ExecutionEnvironmentSortTests::SetUp() {
     executionEnvironment.prepareRootDeviceEnvironments(numRootDevices);
     for (uint32_t rootDeviceIndex = 0; rootDeviceIndex < numRootDevices; rootDeviceIndex++) {

shared/test/unit_test/os_interface/device_factory_tests.cpp

@@ -8,12 +8,14 @@
 #include "shared/source/helpers/constants.h"
 #include "shared/source/helpers/product_config_helper.h"
 #include "shared/source/os_interface/device_factory.h"
+#include "shared/source/os_interface/os_interface.h"
 #include "shared/source/release_helper/release_helper.h"
 #include "shared/test/common/helpers/debug_manager_state_restore.h"
 #include "shared/test/common/helpers/default_hw_info.h"
 #include "shared/test/common/helpers/device_caps_reader_test_helper.h"
 #include "shared/test/common/helpers/gtest_helpers.h"
 #include "shared/test/common/helpers/stream_capture.h"
+#include "shared/test/common/mocks/mock_driver_model.h"
 #include "shared/test/common/mocks/mock_execution_environment.h"
 #include "shared/test/common/mocks/mock_product_helper.h"
 #include "shared/test/common/test_macros/hw_test.h"
@@ -54,6 +56,30 @@ TEST_F(DeviceFactoryTests, givenHwIpVersionOverrideWhenPrepareDeviceEnvironments
     EXPECT_NE(0u, executionEnvironment.rootDeviceEnvironments[0]->getHardwareInfo()->platform.usDeviceID);
 }
 
+TEST_F(DeviceFactoryTests, givenHwIpVersionOverrideWhenPrepareDeviceEnvironmentsForProductFamilyOverrideIsCalledWithNumberOfCcsSetToZeroThenFalseIsReturned) {
+    ExecutionEnvironment executionEnvironment{};
+    auto config = defaultHwInfo.get()->ipVersion.value;
+    debugManager.flags.OverrideHwIpVersion.set(config);
+    debugManager.flags.ZEX_NUMBER_OF_CCS.set("0");
+
+    bool success = DeviceFactory::prepareDeviceEnvironmentsForProductFamilyOverride(executionEnvironment);
+    EXPECT_FALSE(success);
+    EXPECT_EQ(config, executionEnvironment.rootDeviceEnvironments[0]->getHardwareInfo()->ipVersion.value);
+    EXPECT_NE(0u, executionEnvironment.rootDeviceEnvironments[0]->getHardwareInfo()->platform.usDeviceID);
+}
+
+TEST_F(DeviceFactoryTests, givenHwIpVersionOverrideWhenPrepareDeviceEnvironmentsForProductFamilyOverrideIsCalledWithNumberOfCcsSetToZeroColonZeroThenFalseIsReturned) {
+    ExecutionEnvironment executionEnvironment{};
+    auto config = defaultHwInfo.get()->ipVersion.value;
+    debugManager.flags.OverrideHwIpVersion.set(config);
+    debugManager.flags.ZEX_NUMBER_OF_CCS.set("0:0");
+
+    bool success = DeviceFactory::prepareDeviceEnvironmentsForProductFamilyOverride(executionEnvironment);
+    EXPECT_FALSE(success);
+    EXPECT_EQ(config, executionEnvironment.rootDeviceEnvironments[0]->getHardwareInfo()->ipVersion.value);
+    EXPECT_NE(0u, executionEnvironment.rootDeviceEnvironments[0]->getHardwareInfo()->platform.usDeviceID);
+}
+
 TEST_F(DeviceFactoryTests, givenHwIpVersionOverrideWhenPrepareDeviceEnvironmentsForProductFamilyOverrideIsCalledThenReleaseHelperContainsCorrectIpVersion) {
     ExecutionEnvironment executionEnvironment{};
     auto config = defaultHwInfo.get()->ipVersion.value;
@@ -156,6 +182,44 @@ TEST_F(DeviceFactoryTests, givenMultipleDevicesWhenInitializeResourcesSucceedsFo
     EXPECT_EQ(2u, rootDeviceEnvironment1->initOsInterfaceCalled);
 }
 
+class MockExecutionEnvironmentConfigureCssMode : public MockExecutionEnvironment {
+  public:
+    using MockExecutionEnvironment::MockExecutionEnvironment;
+    using MockExecutionEnvironment::rootDeviceEnvironments;
+
+    void configureCcsMode() override {
+        return;
+    }
+};
+
+TEST_F(DeviceFactoryTests, givenDeviceWhenInitializeResourcesSucceedsButCcsNumberIsZeroThenFalseIsReturned) {
+    debugManager.flags.CreateMultipleRootDevices.set(1);
+    debugManager.flags.ZEX_NUMBER_OF_CCS.set("0");
+    MockExecutionEnvironmentConfigureCssMode executionEnvironment(defaultHwInfo.get(), true, 1u);
+
+    EXPECT_EQ(1u, executionEnvironment.rootDeviceEnvironments.size());
+    auto rootDeviceEnvironment = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[0].get());
+
+    rootDeviceEnvironment->initOsInterfaceResults.push_back(true);
+
+    bool success = DeviceFactory::prepareDeviceEnvironments(executionEnvironment);
+    EXPECT_FALSE(success);
+}
+
+TEST_F(DeviceFactoryTests, givenDeviceWhenInitializeResourcesSucceedsButCcsNumberIsZeroColonZeroThenFalseIsReturned) {
+    debugManager.flags.CreateMultipleRootDevices.set(1);
+    debugManager.flags.ZEX_NUMBER_OF_CCS.set("0:0");
+    MockExecutionEnvironmentConfigureCssMode executionEnvironment(defaultHwInfo.get(), true, 1u);
+
+    EXPECT_EQ(1u, executionEnvironment.rootDeviceEnvironments.size());
+    auto rootDeviceEnvironment = static_cast<MockRootDeviceEnvironment *>(executionEnvironment.rootDeviceEnvironments[0].get());
+
+    rootDeviceEnvironment->initOsInterfaceResults.push_back(true);
+
+    bool success = DeviceFactory::prepareDeviceEnvironments(executionEnvironment);
+    EXPECT_FALSE(success);
+}
+
 TEST_F(DeviceFactoryTests, givenMultipleDevicesWhenInitializeResourcesFailsForAllDevicesThenFailureIsReturned) {
     DebugManagerStateRestore restorer;
     debugManager.flags.CreateMultipleRootDevices.set(3);
@@ -386,4 +450,4 @@ HWTEST_F(DeviceFactoryOverrideTest, GivenAubModeWhenValidateDeviceFlagsThenIsPro
         EXPECT_FALSE(hasSubstr(capturedStderr, expectedMissingProductFamilyStderrSubstr));
         EXPECT_FALSE(hasSubstr(capturedStderr, expectedMissingHardwareInfoStderrSubstr));
     }
-}
+}