From f3ffd7070c8440a477af417606fd4a36b7bad234 Mon Sep 17 00:00:00 2001
From: German Andryeyev <56892148+gandryey@users.noreply.github.com>
Date: Wed, 3 Dec 2025 11:49:40 -0500
Subject: [PATCH] rocr: Change hsaKmtQueueRingDoorbell interface (#2068)

WSL uses the call just for the thread wake-up, however under Windows
KMD needs the actual value (SWDEV-568592). The interface is changed
to avoid programming of a modified write_ptr value, which somewhat
changes the client's logic.
---
 .../rocr-runtime/runtime/hsa-runtime/core/inc/thunk_loader.h   | 3 ++-
 .../runtime/hsa-runtime/core/runtime/amd_aql_queue.cpp         | 2 +-
 .../runtime/hsa-runtime/core/runtime/amd_blit_sdma.cpp         | 2 +-
 3 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/inc/thunk_loader.h b/projects/rocr-runtime/runtime/hsa-runtime/core/inc/thunk_loader.h
index 3c0c5ee4f5..a7b344a2ce 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/inc/thunk_loader.h
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/inc/thunk_loader.h
@@ -323,7 +323,8 @@ class ThunkLoader {
                                       HsaPcSamplingTraceId traceId);
     typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtPcSamplingSupport))(void);
     typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtModelEnabled))(bool* enable);
-    typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtQueueRingDoorbell))(HSA_QUEUEID QueueId);
+    typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtQueueRingDoorbell))(HSA_QUEUEID QueueId, \
+                                      HSAuint64 value);
     typedef HSAKMT_STATUS (HSAKMT_DEF(hsaKmtAisReadWriteFile))(void *devicePtr, \
                                       HSAuint64 size, \
                                       HSAint32 fd, \
diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_aql_queue.cpp b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_aql_queue.cpp
index 979e27c871..396edeff7e 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_aql_queue.cpp
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_aql_queue.cpp
@@ -471,7 +471,7 @@ uint64_t AqlQueue::AddWriteIndexRelease(uint64_t value) {
 void AqlQueue::StoreRelaxed(hsa_signal_value_t value) {
   if (core::Runtime::runtime_singleton_->thunkLoader()->IsDTIF() ||
         core::Runtime::runtime_singleton_->thunkLoader()->IsDXG()) {
-    HSAKMT_CALL(hsaKmtQueueRingDoorbell(queue_id_));
+    HSAKMT_CALL(hsaKmtQueueRingDoorbell(queue_id_, value));
   } else {
     // Hardware doorbell supports AQL semantics.
     _mm_sfence();
diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_blit_sdma.cpp b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_blit_sdma.cpp
index e423e47cc2..3067f827d3 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_blit_sdma.cpp
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_blit_sdma.cpp
@@ -785,7 +785,7 @@ void BlitSdma<useGCR>::UpdateWriteAndDoorbellRegister(uint64_t curr_index, uint6
       *reinterpret_cast<uint64_t*>(queue_resource_.Queue_DoorBell) = new_index;
       if (core::Runtime::runtime_singleton_->thunkLoader()->IsDXG() ||
           core::Runtime::runtime_singleton_->thunkLoader()->IsDTIF()) {
-        HSAKMT_CALL(hsaKmtQueueRingDoorbell(queue_resource_.QueueId));
+        HSAKMT_CALL(hsaKmtQueueRingDoorbell(queue_resource_.QueueId, new_index));
       }
 
       atomic::Store(&cached_commit_index_, new_index, std::memory_order_release);