diff --git a/opencl/test/unit_test/test_files/igdrcl.config b/opencl/test/unit_test/test_files/igdrcl.config
index 8ba78ef90f..be423f1996 100644
--- a/opencl/test/unit_test/test_files/igdrcl.config
+++ b/opencl/test/unit_test/test_files/igdrcl.config
@@ -399,3 +399,4 @@ UseTileMemoryBankInVirtualMemoryCreation = -1
 DisableScratchPages = 0
 ForceAllResourcesUncached = 0
 ForcePreParserEnabledForMiArbCheck = -1
+BatchBufferStartPrepatchingWaEnabled = -1
\ No newline at end of file
diff --git a/shared/source/command_stream/command_stream_receiver_hw_base.inl b/shared/source/command_stream/command_stream_receiver_hw_base.inl
index a1ef972d69..92ffbcc365 100644
--- a/shared/source/command_stream/command_stream_receiver_hw_base.inl
+++ b/shared/source/command_stream/command_stream_receiver_hw_base.inl
@@ -94,13 +94,21 @@ inline void CommandStreamReceiverHw<GfxFamily>::addBatchBufferEnd(LinearStream &
 template <typename GfxFamily>
 inline void CommandStreamReceiverHw<GfxFamily>::programEndingCmd(LinearStream &commandStream, Device &device, void **patchLocation, bool directSubmissionEnabled) {
     if (directSubmissionEnabled) {
+        uint64_t startAddress = 0;
+        if (DebugManager.flags.BatchBufferStartPrepatchingWaEnabled.get() == 1) {
+            startAddress = commandStream.getGraphicsAllocation()->getGpuAddress() + commandStream.getUsed();
+        }
+
         *patchLocation = commandStream.getSpace(sizeof(MI_BATCH_BUFFER_START));
         auto bbStart = reinterpret_cast<MI_BATCH_BUFFER_START *>(*patchLocation);
         MI_BATCH_BUFFER_START cmd = {};
-        addBatchBufferStart(&cmd, 0ull, false);
+
+        addBatchBufferStart(&cmd, startAddress, false);
         *bbStart = cmd;
     } else {
-        PreemptionHelper::programStateSipEndWa<GfxFamily>(commandStream, device);
+        if (!EngineHelpers::isBcs(osContext->getEngineType())) {
+            PreemptionHelper::programStateSipEndWa<GfxFamily>(commandStream, device);
+        }
         this->addBatchBufferEnd(commandStream, patchLocation);
     }
 }
@@ -1091,16 +1099,7 @@ uint32_t CommandStreamReceiverHw<GfxFamily>::flushBcsTask(const BlitPropertiesCo
     }
 
     void *endingCmdPtr = nullptr;
-    if (blitterDirectSubmission) {
-        endingCmdPtr = commandStream.getSpace(0);
-        EncodeBatchBufferStartOrEnd<GfxFamily>::programBatchBufferStart(&commandStream,
-                                                                        0ull,
-                                                                        false);
-    } else {
-        auto batchBufferEnd = reinterpret_cast<MI_BATCH_BUFFER_END *>(
-            commandStream.getSpace(sizeof(MI_BATCH_BUFFER_END)));
-        *batchBufferEnd = GfxFamily::cmdInitBatchBufferEnd;
-    }
+    programEndingCmd(commandStream, device, &endingCmdPtr, blitterDirectSubmission);
 
     EncodeNoop<GfxFamily>::alignToCacheLine(commandStream);
 
diff --git a/shared/source/debug_settings/debug_variables_base.inl b/shared/source/debug_settings/debug_variables_base.inl
index b38a7c0646..0d0a3eb4f2 100644
--- a/shared/source/debug_settings/debug_variables_base.inl
+++ b/shared/source/debug_settings/debug_variables_base.inl
@@ -195,6 +195,7 @@ DECLARE_DEBUG_VARIABLE(int32_t, OverridePatIndex, -1, "-1: default, >=0: PatInde
 DECLARE_DEBUG_VARIABLE(int32_t, UseTileMemoryBankInVirtualMemoryCreation, -1, "-1: default - on, 0: do not assign tile memory bank to virtual memory space, 1: assign tile memory bank to virtual memory space")
 DECLARE_DEBUG_VARIABLE(int32_t, OverrideTimestampEvents, -1, "-1: default (based on user settings), 0: Force disable timestamp events (no timestamps will be reported), 1: Force enable timestamp events")
 DECLARE_DEBUG_VARIABLE(int32_t, ForcePreParserEnabledForMiArbCheck, -1, "-1: default , 0: PreParser disabled, 1: PreParser enabled")
+DECLARE_DEBUG_VARIABLE(int32_t, BatchBufferStartPrepatchingWaEnabled, -1, "-1: default , 0: disabled, 1: enabled. WA applies valid VA pointing to 'self' instead of 0x0. This mitigates incorrect VA preparsing.")
 DECLARE_DEBUG_VARIABLE(bool, DisableScratchPages, false, "Disable scratch pages during VM creations")
 /*LOGGING FLAGS*/
 DECLARE_DEBUG_VARIABLE(int32_t, PrintDriverDiagnostics, -1, "prints driver diagnostics messages to standard output, value corresponds to hint level")
diff --git a/shared/test/unit_test/direct_submission/direct_submission_tests_1.cpp b/shared/test/unit_test/direct_submission/direct_submission_tests_1.cpp
index 37ad85918c..34832bb5c7 100644
--- a/shared/test/unit_test/direct_submission/direct_submission_tests_1.cpp
+++ b/shared/test/unit_test/direct_submission/direct_submission_tests_1.cpp
@@ -617,6 +617,46 @@ HWTEST_F(DirectSubmissionTest, givenDirectSubmissionAvailableWhenProgrammingEndi
     EXPECT_EQ(expectedSize, mockCsr->getCmdSizeForEpilogue(dispatchFlags));
 }
 
+HWTEST_F(DirectSubmissionTest, givenDebugFlagSetWhenProgrammingEndingCommandThenUseNonZeroBatchBufferStart) {
+    using MI_BATCH_BUFFER_START = typename FamilyType::MI_BATCH_BUFFER_START;
+
+    DebugManagerStateRestore restorer;
+
+    DebugManager.flags.BatchBufferStartPrepatchingWaEnabled.set(1);
+
+    MockGraphicsAllocation mockAllocation;
+
+    int32_t executionStamp = 0;
+    std::unique_ptr<MockCsr<FamilyType>> mockCsr = std::make_unique<MockCsr<FamilyType>>(executionStamp, *pDevice->executionEnvironment,
+                                                                                         pDevice->getRootDeviceIndex(), pDevice->getDeviceBitfield());
+
+    mockCsr->setupContext(*osContext);
+    mockCsr->directSubmissionAvailable = true;
+    bool ret = mockCsr->isDirectSubmissionEnabled();
+    EXPECT_TRUE(ret);
+
+    void *location = nullptr;
+    uint8_t buffer[128];
+    mockCsr->commandStream.replaceBuffer(&buffer[0], 128u);
+    mockCsr->commandStream.replaceGraphicsAllocation(&mockAllocation);
+    auto &device = *pDevice;
+    mockCsr->programEndingCmd(mockCsr->commandStream, device, &location, ret);
+    EXPECT_EQ(sizeof(MI_BATCH_BUFFER_START), mockCsr->commandStream.getUsed());
+
+    DispatchFlags dispatchFlags = DispatchFlagsHelper::createDefaultDispatchFlags();
+    dispatchFlags.epilogueRequired = true;
+    size_t expectedSize = sizeof(MI_BATCH_BUFFER_START) +
+                          mockCsr->getCmdSizeForEpilogueCommands(dispatchFlags);
+    expectedSize = alignUp(expectedSize, MemoryConstants::cacheLineSize);
+    EXPECT_EQ(expectedSize, mockCsr->getCmdSizeForEpilogue(dispatchFlags));
+
+    auto bbStartCmd = reinterpret_cast<MI_BATCH_BUFFER_START *>(mockCsr->commandStream.getCpuBase());
+
+    EXPECT_EQ(mockCsr->commandStream.getGraphicsAllocation()->getGpuAddress(), bbStartCmd->getBatchBufferStartAddress());
+
+    mockCsr->commandStream.replaceGraphicsAllocation(nullptr);
+}
+
 HWTEST_F(DirectSubmissionTest,
          givenDirectSubmissionDiagnosticNotAvailableWhenDiagnosticRegistryIsUsedThenDoNotEnableDiagnostic) {
     using Dispatcher = RenderDispatcher<FamilyType>;