Merge pull request #2197 from elBoberido/iox-2196-fix-memory-order-in…

…-mpmc-loffli-fence-synchronization

iox-#2196 Fix memory order in 'MpmcLoFFLi' fence synchronization

doc/website/release-notes/iceoryx-unreleased.md

@@ -117,7 +117,8 @@
 - Add public functions to create an 'access_rights' object from integer values [#2108](https://github.com/eclipse-iceoryx/iceoryx/issues/2108)
 - Fix `historyRequest` may be larger than `queueCapacity` during creating a subscriber [#2121](https://github.com/eclipse-iceoryx/iceoryx/issues/2121)
 - Unable to acquire file status due to an unknown failure [#2023](https://github.com/eclipse-iceoryx/iceoryx/issues/2023)
-- Bug in 'ListenerImpl' [#2137](https://github.com/eclipse-iceoryx/iceoryx/issues/2137)
+- Bug in `ListenerImpl` [#2137](https://github.com/eclipse-iceoryx/iceoryx/issues/2137)
+- Wrong memory order in `MpmcLoFFLi` fence synchronization [#2196](https://github.com/eclipse-iceoryx/iceoryx/issues/2196)
 
 **Refactoring:**
 

iceoryx_hoofs/concurrent/buffer/source/mpmc_loffli.cpp

@@ -74,7 +74,12 @@ bool MpmcLoFFLi::pop(Index_t& index) noexcept
     m_nextFreeIndex.get()[index] = m_invalidIndex;
 
     /// we need to synchronize m_nextFreeIndex with push so that we can perform a validation
-    /// check right before push to avoid double free's
+    /// check right before push to avoid double free's;
+    /// a simple fence without explicit atomic store should be sufficient since the index needs to be transferred to
+    /// another thread and the most simple method would be a relaxed store of the index in the 'pop' thread and a
+    /// relaxed load of the same atomic in the 'push' thread which would be the atomic in the fence to fence
+    /// synchronization; other mechanism would involve stronger synchronizations and implicitly also synchronize
+    /// m_nextFreeIndex
     std::atomic_thread_fence(std::memory_order_release);
 
     return true;
@@ -83,7 +88,7 @@ bool MpmcLoFFLi::pop(Index_t& index) noexcept
 bool MpmcLoFFLi::push(const Index_t index) noexcept
 {
     /// we synchronize with m_nextFreeIndex in pop to perform the validity check
-    std::atomic_thread_fence(std::memory_order_release);
+    std::atomic_thread_fence(std::memory_order_acquire);
 
     /// we want to avoid double free's therefore we check if the index was acquired
     /// in pop and the push argument "index" is valid

iceoryx_posh/include/iceoryx_posh/internal/popo/used_chunk_list.inl

@@ -53,7 +53,6 @@ bool UsedChunkList<Capacity>::insert(mepoo::SharedChunk chunk) noexcept
         // set freeListHead to the next free entry
         m_freeListHead = nextFree;
 
-        /// @todo iox-#623 can we do this cheaper with a global fence in cleanup?
         m_synchronizer.clear(std::memory_order_release);
         return true;
     }
@@ -92,7 +91,6 @@ bool UsedChunkList<Capacity>::remove(const mepoo::ChunkHeader* chunkHeader, mepo
                 m_listIndices[current] = m_freeListHead;
                 m_freeListHead = current;
 
-                /// @todo iox-#623 can we do this cheaper with a global fence in cleanup?
                 m_synchronizer.clear(std::memory_order_release);
                 return true;
             }