|author||Denys Vlasenko <firstname.lastname@example.org>||2013-07-27 12:35:51 (GMT)|
|committer||Denys Vlasenko <email@example.com>||2013-07-27 12:35:51 (GMT)|
Signed-off-by: Denys Vlasenko <firstname.lastname@example.org>
1 files changed, 18 insertions, 1 deletions
diff --git a/docs/tcp.txt b/docs/tcp.txt
index 7951e1c..c867c6d 100644
@@ -39,11 +39,28 @@ Solution #1: block until sending is done:
Solution #2: tell kernel that you are done sending.
-This makes kernel send FIN, not RST:
+This makes kernel send FIN after all data is written:
+However, experiments on Linux 3.9.4 show that kernel can return from
+shutdown() and from close() before all data is sent,
+and if peer sends any data to us after this, kernel stll responds with
+RST before all our data is sent.
+In practice the protocol in use often does not allow peer to send
+such data to us, in which case this solution is acceptable.
+If you know that peer is going to close its end after it sees our FIN
+(as EOF), it might be a good idea to perform a read after shutdown().
+When read finishes with 0-sized result, we conclude that peer received all
+the data, saw EOF, and closed its end.
+However, this incurs small performance penalty (we run for a longer time)
+and requires safeguards (nonblocking reads, timeouts etc) against
+malicious peers which don't close the connection.