Use std::align_alloc in file_data_loader (pytorch#10660)

Summary:

Issue with aligned buffers: P1800967583

The alignment requested is 16, and std::max_align_t is also 16. This means we do not need to pad the size to meet any alignment.

However, the buffer we get from malloc is aligned to 8, not 16. When we try to align the buffer, we overflow and error out.

Seems like malloc is not guaranteed to return 8 or 16 byte-aligned buffers, so also a bit hard to test definitively. So far we've only seen this when the buffer size is small (size 2, 4)
```
The malloc(), calloc(), realloc(), and reallocarray() functions
return a pointer to the allocated memory, which is suitably
aligned for any type that fits into the requested size or less.
```

Use std::aligned_alloc (C++17) to ensure buffer is aligned.

Reviewed By: larryliu0820

Differential Revision: D74041198

Author: lucylq

extension/data_loader/file_data_loader.cpp

@@ -21,6 +21,7 @@
 
 #include <executorch/runtime/core/error.h>
 #include <executorch/runtime/core/result.h>
+#include <executorch/runtime/platform/compiler.h>
 #include <executorch/runtime/platform/log.h>
 
 // Some platforms (e.g. Xtensa) do not support pread() that we use to read the
@@ -49,20 +50,6 @@ namespace {
 static bool is_power_of_2(size_t value) {
   return value > 0 && (value & ~(value - 1)) == value;
 }
-
-/**
- * Returns the next alignment for a given pointer.
- */
-static uint8_t* align_pointer(void* ptr, size_t alignment) {
-  intptr_t addr = reinterpret_cast<intptr_t>(ptr);
-  if ((addr & (alignment - 1)) == 0) {
-    // Already aligned.
-    return reinterpret_cast<uint8_t*>(ptr);
-  }
-  // Bump forward.
-  addr = (addr | (alignment - 1)) + 1;
-  return reinterpret_cast<uint8_t*>(addr);
-}
 } // namespace
 
 FileDataLoader::~FileDataLoader() {
@@ -129,13 +116,13 @@ namespace {
 /**
  * FreeableBuffer::FreeFn-compatible callback.
  *
- * `context` is actually a ptrdiff_t value (not a pointer) that contains the
- * offset in bytes between `data` and the actual pointer to free.
+ * `context` is the original buffer pointer. It is allocated with
+ * ET_ALIGNED_ALLOC, and must be freed with ET_ALIGNED_FREE.
+ *
+ * `data` and `size` are unused.
  */
 void FreeSegment(void* context, void* data, ET_UNUSED size_t size) {
-  ptrdiff_t offset = reinterpret_cast<ptrdiff_t>(context);
-  ET_DCHECK_MSG(offset >= 0, "Unexpected offset %ld", (long int)offset);
-  std::free(static_cast<uint8_t*>(data) - offset);
+  ET_ALIGNED_FREE(context);
 }
 } // namespace
 
@@ -163,57 +150,26 @@ Result<FreeableBuffer> FileDataLoader::load(
   }
 
   // Allocate memory for the FreeableBuffer.
-  size_t alloc_size = size;
-  if (alignment_ > alignof(std::max_align_t)) {
-    // malloc() will align to smaller values, but we must manually align to
-    // larger values.
-    alloc_size += alignment_;
-  }
-  void* buffer = std::malloc(alloc_size);
-  if (buffer == nullptr) {
+  void* aligned_buffer = ET_ALIGNED_ALLOC(alignment_, size);
+  if (aligned_buffer == nullptr) {
     ET_LOG(
         Error,
-        "Reading from %s at offset %zu: malloc(%zd) failed",
+        "Reading from %s at offset %zu: ET_ALIGNED_ALLOC(%zd, %zd) failed",
         file_name_,
         offset,
+        alignment_,
         size);
     return Error::MemoryAllocationFailed;
   }
 
-  // Align.
-  void* aligned_buffer = align_pointer(buffer, alignment_);
-
-  // Assert that the alignment didn't overflow the buffer.
-  ET_DCHECK_MSG(
-      reinterpret_cast<uintptr_t>(aligned_buffer) + size <=
-          reinterpret_cast<uintptr_t>(buffer) + alloc_size,
-      "aligned_buffer %p + size %zu > buffer %p + alloc_size %zu",
-      aligned_buffer,
-      size,
-      buffer,
-      alloc_size);
-
   auto err = load_into(offset, size, segment_info, aligned_buffer);
   if (err != Error::Ok) {
-    // Free `buffer`, which is what malloc() gave us, not `aligned_buffer`.
-    std::free(buffer);
+    ET_ALIGNED_FREE(aligned_buffer);
     return err;
   }
 
-  // We can't naively free this pointer, since it may not be what malloc() gave
-  // us. Pass the offset to the real buffer as context. This is the number of
-  // bytes that need to be subtracted from the FreeableBuffer::data() pointer to
-  // find the actual pointer to free.
-  return FreeableBuffer(
-      aligned_buffer,
-      size,
-      FreeSegment,
-      /*free_fn_context=*/
-      reinterpret_cast<void*>(
-          // Using signed types here because it will produce a signed ptrdiff_t
-          // value, though for us it will always be non-negative.
-          reinterpret_cast<intptr_t>(aligned_buffer) -
-          reinterpret_cast<intptr_t>(buffer)));
+  // Pass the aligned_buffer pointer as context to FreeSegment.
+  return FreeableBuffer(aligned_buffer, size, FreeSegment, aligned_buffer);
 }
 
 Result<size_t> FileDataLoader::size() const {

runtime/platform/compiler.h

@@ -171,6 +171,44 @@
 using ssize_t = ptrdiff_t;
 #endif
 
+/**
+ * Platform-specific aligned memory allocation and deallocation.
+ *
+ * Usage:
+ *   void* ptr = ET_ALIGNED_ALLOC(alignment, size);
+ *   // use ptr...
+ *   ET_ALIGNED_FREE(ptr);
+ *
+ * Note: alignment must be a power of 2 and size must be an integral multiple of
+ * alignment.
+ */
+#if defined(_MSC_VER)
+#include <malloc.h>
+#define ET_ALIGNED_ALLOC(alignment, size) \
+  _aligned_malloc(((size + alignment - 1) & ~(alignment - 1)), (alignment))
+#define ET_ALIGNED_FREE(ptr) _aligned_free(ptr)
+#elif defined(__APPLE__)
+#include <stdlib.h> // For posix_memalign and free
+inline void* et_apple_aligned_alloc(size_t alignment, size_t size) {
+  void* ptr = nullptr;
+  // The address of the allocated memory must be a multiple of sizeof(void*).
+  alignment = std::max(alignment, sizeof(void*));
+  if (posix_memalign(
+          &ptr, alignment, (size + alignment - 1) & ~(alignment - 1)) != 0) {
+    return nullptr;
+  }
+  return ptr;
+}
+#define ET_ALIGNED_ALLOC(alignment, size) \
+  et_apple_aligned_alloc((alignment), (size))
+#define ET_ALIGNED_FREE(ptr) free(ptr)
+#else
+// Linux and other posix systems.
+#define ET_ALIGNED_ALLOC(alignment, size) \
+  ::aligned_alloc(alignment, (size + alignment - 1) & ~(alignment - 1))
+#define ET_ALIGNED_FREE(ptr) free(ptr)
+#endif
+
 // DEPRECATED: Use the non-underscore-prefixed versions instead.
 // TODO(T199005537): Remove these once all users have stopped using them.
 #define __ET_DEPRECATED ET_DEPRECATED