Fix unaligned load and stores: (#4528) (#4531)

Misaligned load and store operations are supported by both Intel and ARM
CPUs. However, in C++, these operations are undefined behavior (UB).
Substituting these operations with a `memcpy` fixes this UB. The
compiled assembly code is equivalent to the original, so there is no
performance penalty to using memcpy.

For context: The unaligned load and store operations fixed here were
originally introduced in the slab allocator (#4218).

src/ripple/basics/SlabAllocator.h

@@ -21,16 +21,18 @@
 #define RIPPLE_BASICS_SLABALLOCATOR_H_INCLUDED
 
 #include <ripple/beast/type_name.h>
-#include <algorithm>
-#include <atomic>
-#include <cassert>
-#include <cstdint>
-#include <mutex>
 
 #include <boost/align.hpp>
 #include <boost/container/static_vector.hpp>
 #include <boost/predef.h>
 
+#include <algorithm>
+#include <atomic>
+#include <cassert>
+#include <cstdint>
+#include <cstring>
+#include <mutex>
+
 #if BOOST_OS_LINUX
 #include <sys/mman.h>
 #endif
@@ -76,7 +78,9 @@ class SlabAllocator
 
             while (data + item <= p_ + size_)
             {
-                *reinterpret_cast<std::uint8_t**>(data) = l_;
+                // Use memcpy to avoid unaligned UB
+                // (will optimize to equivalent code)
+                std::memcpy(data, &l_, sizeof(std::uint8_t*));
                 l_ = data;
                 data += item;
             }
@@ -115,7 +119,11 @@ class SlabAllocator
                 ret = l_;
 
                 if (ret)
-                    l_ = *reinterpret_cast<std::uint8_t**>(ret);
+                {
+                    // Use memcpy to avoid unaligned UB
+                    // (will optimize to equivalent code)
+                    std::memcpy(&l_, ret, sizeof(std::uint8_t*));
+                }
             }
 
             return ret;
@@ -136,7 +144,10 @@ class SlabAllocator
             assert(own(ptr));
 
             std::lock_guard l(m_);
-            *reinterpret_cast<std::uint8_t**>(ptr) = l_;
+
+            // Use memcpy to avoid unaligned UB
+            // (will optimize to equivalent code)
+            std::memcpy(ptr, &l_, sizeof(std::uint8_t*));
             l_ = ptr;
         }
     };

src/ripple/basics/impl/partitioned_unordered_map.cpp

@@ -31,7 +31,11 @@ namespace ripple {
 static std::size_t
 extract(uint256 const& key)
 {
-    return *reinterpret_cast<std::size_t const*>(key.data());
+    std::size_t result;
+    // Use memcpy to avoid unaligned UB
+    // (will optimize to equivalent code)
+    std::memcpy(&result, key.data(), sizeof(std::size_t));
+    return result;
 }
 
 static std::size_t

src/ripple/beast/hash/impl/xxhash.cpp

@@ -33,6 +33,8 @@ You can contact the author at :
 
 #include <ripple/beast/hash/impl/xxhash.h>
 
+#include <cstring>
+
 //**************************************
 // Tuning parameters
 //**************************************
@@ -87,7 +89,7 @@ You can contact the author at :
 //**************************************
 // Includes & Memory related functions
 //**************************************
-//#include "xxhash.h"
+// #include "xxhash.h"
 // Modify the local functions below should you wish to use some other memory
 // routines for malloc(), free()
 #include <stdlib.h>
@@ -260,7 +262,13 @@ FORCE_INLINE U64
 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
 {
     if (align == XXH_unaligned)
-        return endian == XXH_littleEndian ? A64(ptr) : XXH_swap64(A64(ptr));
+    {
+        // Use memcpy to avoid unaligned UB
+        U64 tmp_aligned;
+        std::memcpy(&tmp_aligned, ptr, sizeof(U64));
+        return endian == XXH_littleEndian ? tmp_aligned
+                                          : XXH_swap64(tmp_aligned);
+    }
     else
         return endian == XXH_littleEndian ? *(U64*)ptr : XXH_swap64(*(U64*)ptr);
 }