branches/zip: When adding a page to the buffer pool, add it to

buf_pool->page_hash and buf_pool->LRU before releasing buf_pool->mutex.

buf_page_init_for_read(), buf_page_create(): Allocate the compressed
page after the block has been added to the buffer pool.  Document the
reason for this.

buf/buf0buf.c

@@ -2231,15 +2231,6 @@ err_exit:
 
 	ut_ad(block);
 
-	if (zip_size) {
-		void*	data;
-		page_zip_set_size(&block->page.zip, zip_size);
-		mutex_exit(&block->mutex);
-		data = buf_buddy_alloc(zip_size, TRUE);
-		mutex_enter(&block->mutex);
-		block->page.zip.data = data;
-	}
-
 	buf_page_init(space, offset, block);
 
 	/* The block must be put to the LRU list, to the old blocks */
@@ -2250,6 +2241,21 @@ err_exit:
 
 	buf_pool->n_pend_reads++;
 
+	if (zip_size) {
+		void*	data;
+		page_zip_set_size(&block->page.zip, zip_size);
+		mutex_exit(&block->mutex);
+		/* buf_pool->mutex may be released and reacquired by
+		buf_buddy_alloc().  Thus, we must release block->mutex
+		in order not to break the latching order in
+		the reacquisition of buf_pool->mutex.  We also must
+		defer this operation until after the block descriptor
+		has been added to buf_pool->LRU and buf_pool->page_hash. */
+		data = buf_buddy_alloc(zip_size, TRUE);
+		mutex_enter(&block->mutex);
+		block->page.zip.data = data;
+	}
+
 	/* We set a pass-type x-lock on the frame because then the same
 	thread which called for the read operation (and is running now at
 	this point of code) can wait for the read to complete by waiting
@@ -2326,11 +2332,6 @@ buf_page_create(
 
 	block = free_block;
 
-	if (zip_size) {
-		page_zip_set_size(&block->page.zip, zip_size);
-		block->page.zip.data = buf_buddy_alloc(zip_size, TRUE);
-	}
-
 	mutex_enter(&block->mutex);
 
 	buf_page_init(space, offset, block);
@@ -2341,6 +2342,21 @@ buf_page_create(
 	buf_block_buf_fix_inc(block, __FILE__, __LINE__);
 	buf_pool->n_pages_created++;
 
+	if (zip_size) {
+		void*	data;
+		page_zip_set_size(&block->page.zip, zip_size);
+		mutex_exit(&block->mutex);
+		/* buf_pool->mutex may be released and reacquired by
+		buf_buddy_alloc().  Thus, we must release block->mutex
+		in order not to break the latching order in
+		the reacquisition of buf_pool->mutex.  We also must
+		defer this operation until after the block descriptor
+		has been added to buf_pool->LRU and buf_pool->page_hash. */
+		data = buf_buddy_alloc(zip_size, TRUE);
+		mutex_enter(&block->mutex);
+		block->page.zip.data = data;
+	}
+
 	mutex_exit(&(buf_pool->mutex));
 
 	mtr_memo_push(mtr, block, MTR_MEMO_BUF_FIX);