PostgreSQL数据库缓冲区管理器——Buffer Pool初始化

PostgreSQL缓冲区管理器由三层组成，即缓冲表层、缓冲区描述符层和缓冲池层。缓冲表层是一个散列表，它存储着页面的buffer_tag与描述符的buffer_id之间的映射关系；缓冲区描述符层是一个由缓冲区描述符组成的数组（每个描述符与缓冲池槽一一对应，并保存着相应槽的元数据）；缓冲池层是一个数组（每个槽都存储一个数据文件页，数组槽的索引为buffer_id）。

缓冲池只是一个用于存储关系数据文件（例如表或索引）页面的简单数组。缓冲池数组的序号索引就是buffer_id。缓冲池槽的大小为8KB，等于页面大小，因而每个槽都能存储整个页面Page(src/include/storage/bufpage.h)。如上图（熊灿灿大神出品）中Shared Buffers中的Buffer Pool框图，其包含的页面数量定义在NBuffers GUC参数中。

Buffer Pool

Buffer Pool初始化代码定义在src/backend/storage/buffer/buf_init.c文件中，BufferShmemSize函数用于计算buffer pool共享内存的大小（包含data pages、buffer descriptprs、hash tables等）。最好在BufferDesc中包含I/O锁，但这会将BufferDesc的大小增加到多个缓存行，并且基准测试表明，保持每个BufferDesc在缓存行边界上对齐对性能很重要。因此，相反，I/O锁的阵列被分配在一个单独的部分中。因为这些锁不是高度竞争的，所以我们用最小的填充来布置阵列。

Size BufferShmemSize(void) {
  Size    size = 0; 
  size = add_size(size, mul_size(NBuffers, sizeof(BufferDescPadded))); /* size of buffer descriptors */ <--- 缓冲区描述符层 
  size = add_size(size, PG_CACHE_LINE_SIZE); /* to allow aligning buffer descriptors */   
  size = add_size(size, mul_size(NBuffers, sizeof(LWLockMinimallyPadded))); /* It would be nice to include the I/O locks in the BufferDesc, but that would increase the size of a BufferDesc to more than one cache line, and benchmarking has shown that keeping every BufferDesc aligned on a cache line boundary is important for performance.  So, instead, the array of I/O locks is allocated in a separate tranche.  Because those locks are not highly contended, we lay out the array with minimal padding. */
  size = add_size(size, PG_CACHE_LINE_SIZE);  /* to allow aligning the above */

  size = add_size(size, mul_size(NBuffers, BLCKSZ)); /* size of data pages */ <--- 缓冲池层

  size = add_size(size, StrategyShmemSize()); /* size of stuff controlled by freelist.c */ <-- 缓冲表层+freelist
  
  size = add_size(size, mul_size(NBuffers, sizeof(CkptSortItem)));  /* size of checkpoint sort array in bufmgr.c */
  return size;
}

从InitBufferPool中可以看出Buffer Descriptors数组指针为BufferDescriptors、Buffer Blocks数组指针为BufferBlocks。共享内存申请完毕后，对BufferDescriptors进行初始化，其中需要将buf_id设置为元素在数组中的序号，将freeNext设置为下一个元组在数组中的序号（最后一个设置为FREENEXT_END_OF_LIST），初始化buffer_content和buffer_ioLWLock（buffer_content位于BufferDesc中，buffer_io位于BufferIOLWLockArray中）；初始化缓冲表层（大小为NBuffers + NUM_BUFFER_PARTITIONS），为StrategyControl申请共享内存（设置firstFreeBuffer为0，lastFreeBuffer为NBuffers - 1，设置bgwprocno为-1[No pending notification]）。

void InitBufferPool(void) {
  bool    foundBufs,foundDescs,foundIOLocks,foundBufCkpt; 
  BufferDescriptors = (BufferDescPadded *)ShmemInitStruct("Buffer Descriptors",NBuffers * sizeof(BufferDescPadded),&foundDescs); /* Align descriptors to a cacheline boundary. */
  BufferIOLWLockArray = (LWLockMinimallyPadded *)ShmemInitStruct("Buffer IO Locks",NBuffers * (Size) sizeof(LWLockMinimallyPadded),&foundIOLocks); /* Align lwlocks to cacheline boundary */
  
  BufferBlocks = (char *)ShmemInitStruct("Buffer Blocks",NBuffers * (Size) BLCKSZ, &foundBufs);

  LWLockRegisterTranche(LWTRANCHE_BUFFER_IO_IN_PROGRESS, "buffer_io");
  LWLockRegisterTranche(LWTRANCHE_BUFFER_CONTENT, "buffer_content");
  
  CkptBufferIds = (CkptSortItem *)ShmemInitStruct("Checkpoint BufferIds",NBuffers * sizeof(CkptSortItem), &foundBufCkpt); /* The array used to sort to-be-checkpointed buffer ids is located in shared memory, to avoid having to allocate significant amounts of memory at runtime. As that'd be in the middle of a checkpoint, or when the checkpointer is restarted, memory allocation failures would be painful. */

  if (foundDescs || foundBufs || foundIOLocks || foundBufCkpt){/* should find all of these, or none of them */
    Assert(foundDescs && foundBufs && foundIOLocks && foundBufCkpt);/* note: this path is only taken in EXEC_BACKEND case */
  }else{    
    for (int i = 0; i < NBuffers; i++) { /* Initialize all the buffer headers. */
      BufferDesc *buf = GetBufferDescriptor(i);
      CLEAR_BUFFERTAG(buf->tag);
      pg_atomic_init_u32(&buf->state, 0);
      buf->wait_backend_pid = 0;
      buf->buf_id = i;
      /* Initially link all the buffers together as unused. Subsequent management of this list is done by freelist.c. */
      buf->freeNext = i + 1;
      LWLockInitialize(BufferDescriptorGetContentLock(buf), LWTRANCHE_BUFFER_CONTENT);
      LWLockInitialize(BufferDescriptorGetIOLock(buf), LWTRANCHE_BUFFER_IO_IN_PROGRESS);
    }   
    GetBufferDescriptor(NBuffers - 1)->freeNext = FREENEXT_END_OF_LIST; /* Correct last entry of linked list */
  }
  
  StrategyInitialize(!foundDescs); /* Init other shared buffer-management stuff */  
  WritebackContextInit(&BackendWritebackContext, &backend_flush_after); /* Initialize per-backend file flush context */
}

• IO_IN_PROGRESS – this is a flag in the buffer descriptor. It must be set when an IO is initiated and cleared at the end of the IO. It is there to make sure that one process doesn’t start to use a buffer while another is faulting it in. see WaitIO and related routines. IO_IN_PROGRESS——这是缓冲区描述符中的一个标志。必须在IO启动时设置，并在IO结束时清除。它的作用是确保一个进程在另一个进程将缓冲区插入时不会开始使用缓冲区。请参阅WaitIO和相关例程。
• refcount – Counts the number of processes holding pins on a buffer. A buffer is pinned during IO and immediately after a BufferAlloc(). Pins must be released before end of transaction. For efficiency the shared refcount isn’t increased if an individual backend pins a buffer multiple times. Check the PrivateRefCount infrastructure in bufmgr.c.refcount—统计缓冲区上持有管脚的进程数。缓冲区在IO期间和BufferAlloc（）之后被固定。必须在交易结束前释放PIN。为了提高效率，如果单个后端多次固定缓冲区，共享引用计数不会增加。检查bufmgr.c中的PrivateRefCount基础结构。