fix: Fail device init if kernel debugging is misconfigured

Also print error to stderr

Related-to: GSD-10780

Signed-off-by: Brandon Yates <brandon.yates@intel.com>

shared/source/device/device.cpp

@@ -141,7 +141,9 @@ bool Device::createDeviceImpl() {
         }
 
         // initialize common resources once
-        initializeCommonResources();
+        if (!initializeCommonResources()) {
+            return false;
+        }
     }
 
     // create engines
@@ -175,15 +177,14 @@ bool Device::initDeviceWithEngines() {
     return createEngines();
 }
 
-void Device::initializeCommonResources() {
+bool Device::initializeCommonResources() {
     if (getExecutionEnvironment()->isDebuggingEnabled()) {
         const auto rootDeviceIndex = getRootDeviceIndex();
         auto rootDeviceEnvironment = getExecutionEnvironment()->rootDeviceEnvironments[rootDeviceIndex].get();
         rootDeviceEnvironment->initDebuggerL0(this);
         if (rootDeviceEnvironment->debugger == nullptr) {
-            NEO::printDebugString(NEO::debugManager.flags.PrintDebugMessages.get(), stderr,
-                                  "Debug mode is not enabled in the system.\n");
-            getExecutionEnvironment()->setDebuggingMode(DebuggingMode::disabled);
+            NEO::printDebugString(NEO::debugManager.flags.PrintDebugMessages.get(), stderr, "Debug mode is not enabled in the system.\n");
+            return false;
         }
     }
 
@@ -211,6 +212,7 @@ void Device::initializeCommonResources() {
         deviceBitfields.emplace(getRootDeviceIndex(), getDeviceBitfield());
         deviceUsmMemAllocPoolsManager.reset(new UsmMemAllocPoolsManager(getMemoryManager(), rootDeviceIndices, deviceBitfields, this, InternalMemoryType::deviceUnifiedMemory));
     }
+    return true;
 }
 
 void Device::initUsmReuseLimits() {

shared/source/device/device.h

@@ -270,7 +270,7 @@ class Device : public ReferenceTrackedObject<Device>, NEO::NonCopyableAndNonMova
 
     MOCKABLE_VIRTUAL bool createDeviceImpl();
     bool initDeviceWithEngines();
-    void initializeCommonResources();
+    bool initializeCommonResources();
     bool initDeviceFully();
     void initUsmReuseLimits();
     virtual bool createEngines();

shared/source/dll/linux/debugger_l0_dll_linux.cpp

@@ -19,7 +19,12 @@ std::unique_ptr<NEO::Debugger> DebuggerL0::create(NEO::Device *device) {
         return nullptr;
     }
     auto osInterface = device->getRootDeviceEnvironment().osInterface.get();
-    if (!osInterface || !osInterface->isDebugAttachAvailable()) {
+    if (!osInterface) {
+        return nullptr;
+    }
+    if (!osInterface->isDebugAttachAvailable()) {
+        auto cardName = osInterface->getDriverModel()->as<Drm>()->getSysFsPciPathBaseName();
+        IoFunctions::fprintf(stderr, "Kernel debug mode is not enabled for %s. Device is not available for use\n", cardName.c_str());
         return nullptr;
     }
 

shared/source/os_interface/linux/drm_neo.cpp

@@ -179,6 +179,15 @@ int Drm::getEnabledPooledEu(int &enabled) {
     return getParamIoctl(DrmParam::paramHasPooledEu, &enabled);
 }
 
+std::string Drm::getSysFsPciPathBaseName() {
+    auto fullPath = getSysFsPciPath();
+    size_t pos = fullPath.rfind("/");
+    if (std::string::npos == pos) {
+        return fullPath;
+    }
+    return fullPath.substr(pos + 1, std::string::npos);
+}
+
 std::string Drm::getSysFsPciPath() {
     std::string path = std::string(Os::sysFsPciPathPrefix) + hwDeviceId->getPciPath() + "/drm";
     std::string expectedFilePrefix = path + "/card";

shared/source/os_interface/linux/drm_neo.h

@@ -264,6 +264,7 @@ class Drm : public DriverModel {
     void cleanup() override;
     bool readSysFsAsString(const std::string &relativeFilePath, std::string &readString);
     MOCKABLE_VIRTUAL std::string getSysFsPciPath();
+    MOCKABLE_VIRTUAL std::string getSysFsPciPathBaseName();
     std::unique_ptr<HwDeviceIdDrm> &getHwDeviceId() { return hwDeviceId; }
 
     template <typename DataType>

shared/test/unit_test/device/neo_device_tests.cpp

@@ -2096,7 +2096,7 @@ TEST_F(DeviceTests, GivenDebuggingEnabledWhenDeviceIsInitializedThenL0DebuggerIs
     EXPECT_NE(nullptr, device->getL0Debugger());
 }
 
-TEST_F(DeviceTests, givenDebuggerRequestedByUserAndNotAvailableWhenDeviceIsInitializedThenErrorIsPrintedButNotReturned) {
+TEST_F(DeviceTests, givenDebuggerRequestedByUserAndNotAvailableWhenDeviceIsInitializedThenDeviceIsNullAndErrorIsPrinted) {
     extern bool forceCreateNullptrDebugger;
 
     VariableBackup backupForceCreateNullptrDebugger{&forceCreateNullptrDebugger, true};
@@ -2111,7 +2111,7 @@ TEST_F(DeviceTests, givenDebuggerRequestedByUserAndNotAvailableWhenDeviceIsIniti
     auto output = testing::internal::GetCapturedStderr();
 
     EXPECT_EQ(std::string("Debug mode is not enabled in the system.\n"), output);
-    EXPECT_EQ(nullptr, device->getL0Debugger());
+    EXPECT_EQ(nullptr, device);
 }
 
 TEST_F(DeviceTests, givenDebuggerRequestedByUserWhenDeviceWithSubDevicesCreatedThenInitializeDebuggerOncePerRootDevice) {

shared/test/unit_test/os_interface/linux/drm_tests.cpp

@@ -2269,6 +2269,23 @@ TEST(DrmTest, GivenProductSpecificIoctlHelperAvailableAndDebugFlagToIgnoreIsSetW
     EXPECT_EQ(0u, customFuncCalled);
 }
 
+TEST(DrmTest, GivenSysFsPciPathWhenCallinggetSysFsPciPathBaseNameThenResultIsCorrect) {
+    auto executionEnvironment = std::make_unique<MockExecutionEnvironment>();
+
+    class DrmMockPciPath : public DrmMock {
+      public:
+        DrmMockPciPath(RootDeviceEnvironment &rootDeviceEnvironment) : DrmMock(rootDeviceEnvironment) {}
+        std::string mockSysFsPciPath = "/sys/devices/pci0000:00/0000:00:02.0/drm/card0";
+        std::string getSysFsPciPath() override { return mockSysFsPciPath; }
+    };
+    DrmMockPciPath drm{*executionEnvironment->rootDeviceEnvironments[0]};
+    EXPECT_STREQ("card0", drm.getSysFsPciPathBaseName().c_str());
+    drm.mockSysFsPciPath = "/sys/devices/pci0000:00/0000:00:02.0/drm/card7";
+    EXPECT_STREQ("card7", drm.getSysFsPciPathBaseName().c_str());
+    drm.mockSysFsPciPath = "card8";
+    EXPECT_STREQ("card8", drm.getSysFsPciPathBaseName().c_str());
+}
+
 using DrmHwTest = ::testing::Test;
 HWTEST_F(DrmHwTest, GivenDrmWhenSetupHardwareInfoCalledThenGfxCoreHelperIsInitializedFromProductHelper) {
     DebugManagerStateRestore restore;
@@ -2291,4 +2308,4 @@ HWTEST_F(DrmHwTest, GivenDrmWhenSetupHardwareInfoCalledThenGfxCoreHelperIsInitia
     drm.setupHardwareInfo(&device, false);
 
     EXPECT_TRUE(raii.mockGfxCoreHelper->initFromProductHelperCalled);
-}
+}