From 3d5f7ab8920bc9e6920519c4c2549fd1a9c6f671 Mon Sep 17 00:00:00 2001
From: Giulio rebuffo <giulio.rebuffo@gmail.com>
Date: Fri, 30 Oct 2020 23:16:10 +0100
Subject: [PATCH] added stage 3 documentation to rtd (#1328)

---
 docs/readthedocs/source/stagedsync.rst | 46 +++++++++++++++++++++++++-
 1 file changed, 45 insertions(+), 1 deletion(-)

diff --git a/docs/readthedocs/source/stagedsync.rst b/docs/readthedocs/source/stagedsync.rst
index 1a441734b..56cb61d89 100644
--- a/docs/readthedocs/source/stagedsync.rst
+++ b/docs/readthedocs/source/stagedsync.rst
@@ -81,4 +81,48 @@ The function above its used with ETL to extract blockHashes and blockNumber from
 
 Changes in DB:
 
-    * BlockHash => BlockNumber are written in bucket `dbutils.HeaderNumberPrefix`
\ No newline at end of file
+    * BlockHash => BlockNumber are written in bucket `dbutils.HeaderNumberPrefix`
+
+Stage 3 : Download Block Bodies
+===============================
+
+.. code-block:: go
+
+		{
+			ID: stages.Bodies,
+			Build: func(world StageParameters) *Stage {
+				return &Stage{
+					ID:          stages.Bodies,
+					Description: "Download block bodies",
+					ExecFunc: func(s *StageState, u Unwinder) error {
+						return spawnBodyDownloadStage(s, u, world.d, world.pid, world.prefetchedBlocks)
+					},
+					UnwindFunc: func(u *UnwindState, s *StageState) error {
+						return unwindBodyDownloadStage(u, world.db)
+					},
+				}
+			},
+		},
+
+This stage, downloads block bodies and put them into the database. This stage is divided into two processes:
+
+.. code-block:: go
+
+	func (d *Downloader) fetchBodies(from uint64) error
+
+	func (d *Downloader) processBodiesStage(logPrefix string, to uint64) error
+
+`fetchBodies` downloads the bodies from the peer and decode them from RLP format.
+
+`processBodiesStage` takes the bodies downloaded and those the following with them:
+
+	* Verifiy them.
+	* RLP-encode them.
+	* compress the rlp-encoded bodies using `snappy`.
+	* put the commpressed RLP into the database.
+
+in order for turbo-geth to reaccess the block bodies, it decompress the and rlp-decode them. the entries in the db for block bodies are a concatenation of [block number] + [block hash] in order to pre-sort them before inserting them into the database.
+
+Changes in DB:
+
+    * [block number] + [block hash] => Block bodies are written in bucket `dbutils.BlockBodyPrefix`
\ No newline at end of file