diff --git a/opencl/test/unit_test/mocks/linux/mock_drm_command_stream_receiver.h b/opencl/test/unit_test/mocks/linux/mock_drm_command_stream_receiver.h
index f6f0517982..81df543e19 100644
--- a/opencl/test/unit_test/mocks/linux/mock_drm_command_stream_receiver.h
+++ b/opencl/test/unit_test/mocks/linux/mock_drm_command_stream_receiver.h
@@ -17,6 +17,7 @@ class TestedDrmCommandStreamReceiver : public DrmCommandStreamReceiver<GfxFamily
     using CommandStreamReceiver::globalFenceAllocation;
     using CommandStreamReceiver::makeResident;
     using DrmCommandStreamReceiver<GfxFamily>::residency;
+    using CommandStreamReceiverHw<GfxFamily>::directSubmission;
     using CommandStreamReceiverHw<GfxFamily>::CommandStreamReceiver::lastSentSliceCount;
 
     TestedDrmCommandStreamReceiver(gemCloseWorkerMode mode, ExecutionEnvironment &executionEnvironment)
diff --git a/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests.cpp b/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests.cpp
index 1e1aa70d76..4ec0f478ed 100644
--- a/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests.cpp
+++ b/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests.cpp
@@ -6,6 +6,7 @@
  */
 
 #include "shared/source/command_stream/preemption.h"
+#include "shared/source/direct_submission/linux/drm_direct_submission.h"
 #include "shared/source/gmm_helper/gmm_helper.h"
 #include "shared/source/gmm_helper/page_table_mngr.h"
 #include "shared/source/gmm_helper/resource_info.h"
@@ -1288,7 +1289,6 @@ HWTEST_TEMPLATED_F(DrmCommandStreamEnhancedTest, FlushNotAligned) {
 }
 
 struct DrmCommandStreamDirectSubmissionTest : public DrmCommandStreamEnhancedTest {
-
     template <typename GfxFamily>
     void SetUpT() {
         DebugManager.flags.EnableDirectSubmission.set(1u);
@@ -1308,6 +1308,11 @@ struct DrmCommandStreamDirectSubmissionTest : public DrmCommandStreamEnhancedTes
     DebugManagerStateRestore restorer;
 };
 
+template <typename GfxFamily>
+struct MockDrmDirectSubmission : public DrmDirectSubmission<GfxFamily, RenderDispatcher<GfxFamily>> {
+    using DrmDirectSubmission<GfxFamily, RenderDispatcher<GfxFamily>>::currentTagData;
+};
+
 HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenEnabledDirectSubmissionWhenFlushThenFlushStampIsNotUpdated) {
     auto &cs = csr->getCS();
     CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
@@ -1320,6 +1325,9 @@ HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenEnabledDirectSubmi
     csr->flush(batchBuffer, csr->getResidencyAllocations());
 
     EXPECT_EQ(csr->obtainCurrentFlushStamp(), flushStamp);
+
+    auto directSubmission = static_cast<TestedDrmCommandStreamReceiver<FamilyType> *>(csr)->directSubmission.get();
+    static_cast<MockDrmDirectSubmission<FamilyType> *>(directSubmission)->currentTagData.tagValue = 0u;
 }
 
 HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenEnabledDirectSubmissionWhenFlushThenCommandBufferAllocationIsResident) {
@@ -1334,6 +1342,9 @@ HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenEnabledDirectSubmi
 
     auto memoryOperationsInterface = executionEnvironment->rootDeviceEnvironments[rootDeviceIndex]->memoryOperationsInterface.get();
     EXPECT_EQ(memoryOperationsInterface->isResident(device.get(), *batchBuffer.commandBufferAllocation), MemoryOperationsStatus::SUCCESS);
+
+    auto directSubmission = static_cast<TestedDrmCommandStreamReceiver<FamilyType> *>(csr)->directSubmission.get();
+    static_cast<MockDrmDirectSubmission<FamilyType> *>(directSubmission)->currentTagData.tagValue = 0u;
 }
 
 HWTEST_TEMPLATED_F(DrmCommandStreamEnhancedTest, CheckDrmFree) {
diff --git a/shared/source/direct_submission/direct_submission_hw.h b/shared/source/direct_submission/direct_submission_hw.h
index d796b03704..0ec3bc18dc 100644
--- a/shared/source/direct_submission/direct_submission_hw.h
+++ b/shared/source/direct_submission/direct_submission_hw.h
@@ -19,7 +19,7 @@ namespace NEO {
 struct RingSemaphoreData {
     uint32_t QueueWorkCount;
     uint8_t ReservedCacheline[60];
-    uint32_t Reserved1Uint32;
+    uint32_t tagAllocation;
     uint32_t Reserved2Uint32;
     uint32_t Reserved3Uint32;
     uint32_t Reserved4Uint32;
diff --git a/shared/source/direct_submission/direct_submission_hw.inl b/shared/source/direct_submission/direct_submission_hw.inl
index f26fe659c7..11fbdc9d80 100644
--- a/shared/source/direct_submission/direct_submission_hw.inl
+++ b/shared/source/direct_submission/direct_submission_hw.inl
@@ -86,7 +86,7 @@ bool DirectSubmissionHw<GfxFamily, Dispatcher>::allocateResources() {
     memset(semaphorePtr, 0, sizeof(RingSemaphoreData));
     semaphoreData->QueueWorkCount = 0;
     cpuCachelineFlush(semaphorePtr, MemoryConstants::cacheLineSize);
-    workloadModeOneStoreAddress = static_cast<volatile void *>(&semaphoreData->Reserved1Uint32);
+    workloadModeOneStoreAddress = static_cast<volatile void *>(&semaphoreData->Reserved4Uint32);
     *static_cast<volatile uint32_t *>(workloadModeOneStoreAddress) = 0u;
 
     auto ret = makeResourcesResident(allocations);
diff --git a/shared/source/direct_submission/linux/drm_direct_submission.h b/shared/source/direct_submission/linux/drm_direct_submission.h
index 137d938001..a57cbdf87d 100644
--- a/shared/source/direct_submission/linux/drm_direct_submission.h
+++ b/shared/source/direct_submission/linux/drm_direct_submission.h
@@ -29,5 +29,10 @@ class DrmDirectSubmission : public DirectSubmissionHw<GfxFamily, Dispatcher> {
     void handleSwitchRingBuffers() override;
     uint64_t updateTagValue() override;
     void getTagAddressValue(TagData &tagData) override;
+
+    void wait(uint32_t taskCountToWait);
+
+    TagData currentTagData;
+    volatile uint32_t *tagAddress;
 };
 } // namespace NEO
diff --git a/shared/source/direct_submission/linux/drm_direct_submission.inl b/shared/source/direct_submission/linux/drm_direct_submission.inl
index 3b3bfbabe6..c1a64e927b 100644
--- a/shared/source/direct_submission/linux/drm_direct_submission.inl
+++ b/shared/source/direct_submission/linux/drm_direct_submission.inl
@@ -24,15 +24,14 @@ inline std::unique_ptr<DirectSubmissionHw<GfxFamily, Dispatcher>> DirectSubmissi
 template <typename GfxFamily, typename Dispatcher>
 DrmDirectSubmission<GfxFamily, Dispatcher>::DrmDirectSubmission(Device &device,
                                                                 OsContext &osContext)
-    : DirectSubmissionHw<GfxFamily, Dispatcher>(device, osContext) {
-    this->disableMonitorFence = true;
-}
+    : DirectSubmissionHw<GfxFamily, Dispatcher>(device, osContext){};
 
 template <typename GfxFamily, typename Dispatcher>
 inline DrmDirectSubmission<GfxFamily, Dispatcher>::~DrmDirectSubmission() {
     if (this->ringStart) {
+        this->wait(static_cast<uint32_t>(this->currentTagData.tagValue));
         this->stopRingBuffer();
-        auto bb = static_cast<DrmAllocation *>(this->ringCommandStream.getGraphicsAllocation())->getBO();
+        auto bb = static_cast<DrmAllocation *>(this->ringBuffer)->getBO();
         bb->wait(-1);
     }
     this->deallocateResources();
@@ -40,6 +39,9 @@ inline DrmDirectSubmission<GfxFamily, Dispatcher>::~DrmDirectSubmission() {
 
 template <typename GfxFamily, typename Dispatcher>
 bool DrmDirectSubmission<GfxFamily, Dispatcher>::allocateOsResources() {
+    this->currentTagData.tagAddress = this->semaphoreGpuVa + MemoryConstants::cacheLineSize;
+    this->currentTagData.tagValue = 0u;
+    this->tagAddress = reinterpret_cast<volatile uint32_t *>(reinterpret_cast<uint8_t *>(this->semaphorePtr) + MemoryConstants::cacheLineSize);
     return true;
 }
 
@@ -76,16 +78,30 @@ bool DrmDirectSubmission<GfxFamily, Dispatcher>::handleResidency() {
 
 template <typename GfxFamily, typename Dispatcher>
 void DrmDirectSubmission<GfxFamily, Dispatcher>::handleSwitchRingBuffers() {
+    if (this->ringStart) {
+        if (this->completionRingBuffers[this->currentRingBuffer] != 0) {
+            this->wait(static_cast<uint32_t>(this->completionRingBuffers[this->currentRingBuffer]));
+        }
+    }
 }
 
 template <typename GfxFamily, typename Dispatcher>
 uint64_t DrmDirectSubmission<GfxFamily, Dispatcher>::updateTagValue() {
+    this->currentTagData.tagValue++;
+    this->completionRingBuffers[this->currentRingBuffer] = this->currentTagData.tagValue;
     return 0ull;
 }
 
 template <typename GfxFamily, typename Dispatcher>
 void DrmDirectSubmission<GfxFamily, Dispatcher>::getTagAddressValue(TagData &tagData) {
-    tagData.tagAddress = 0ull;
-    tagData.tagValue = 0ull;
+    tagData.tagAddress = this->currentTagData.tagAddress;
+    tagData.tagValue = this->currentTagData.tagValue + 1;
 }
+
+template <typename GfxFamily, typename Dispatcher>
+void DrmDirectSubmission<GfxFamily, Dispatcher>::wait(uint32_t taskCountToWait) {
+    while (taskCountToWait > *this->tagAddress) {
+    }
+}
+
 } // namespace NEO
diff --git a/shared/test/unit_test/direct_submission/direct_submission_tests.cpp b/shared/test/unit_test/direct_submission/direct_submission_tests.cpp
index 600d910148..8ce1e3b3ef 100644
--- a/shared/test/unit_test/direct_submission/direct_submission_tests.cpp
+++ b/shared/test/unit_test/direct_submission/direct_submission_tests.cpp
@@ -461,7 +461,7 @@ HWTEST_F(DirectSubmissionDispatchBufferTest,
     EXPECT_EQ(0x40u + 1u, storeData->getDataDword0());
     uint64_t expectedGpuVa = directSubmission.semaphoreGpuVa;
     auto semaphore = static_cast<RingSemaphoreData *>(directSubmission.semaphorePtr);
-    expectedGpuVa += ptrDiff(&semaphore->Reserved1Uint32, directSubmission.semaphorePtr);
+    expectedGpuVa += ptrDiff(&semaphore->Reserved4Uint32, directSubmission.semaphorePtr);
     EXPECT_EQ(expectedGpuVa, storeData->getAddress());
 }
 
diff --git a/shared/test/unit_test/direct_submission/linux/drm_direct_submission_tests.cpp b/shared/test/unit_test/direct_submission/linux/drm_direct_submission_tests.cpp
index 6f3fb62cb7..c28156eab0 100644
--- a/shared/test/unit_test/direct_submission/linux/drm_direct_submission_tests.cpp
+++ b/shared/test/unit_test/direct_submission/linux/drm_direct_submission_tests.cpp
@@ -46,11 +46,13 @@ template <typename GfxFamily, typename Dispatcher>
 struct MockDrmDirectSubmission : public DrmDirectSubmission<GfxFamily, Dispatcher> {
     using BaseClass = DrmDirectSubmission<GfxFamily, Dispatcher>;
     using BaseClass::allocateResources;
+    using BaseClass::currentTagData;
     using BaseClass::DrmDirectSubmission;
     using BaseClass::getTagAddressValue;
     using BaseClass::handleResidency;
     using BaseClass::submit;
     using BaseClass::switchRingBuffers;
+    using BaseClass::tagAddress;
     using BaseClass::updateTagValue;
 };
 
@@ -74,8 +76,8 @@ HWTEST_F(DrmDirectSubmissionTest, givenDrmDirectSubmissionWhenCallingLinuxImplem
 
     TagData tagData = {1ull, 1ull};
     drmDirectSubmission.getTagAddressValue(tagData);
-    EXPECT_EQ(0ull, tagData.tagAddress);
-    EXPECT_EQ(0ull, tagData.tagValue);
+    EXPECT_EQ(drmDirectSubmission.currentTagData.tagAddress, tagData.tagAddress);
+    EXPECT_EQ(drmDirectSubmission.currentTagData.tagValue + 1, tagData.tagValue);
 }
 
 HWTEST_F(DrmDirectSubmissionTest, whenCreateDirectSubmissionThenValidObjectIsReturned) {