From 20e053016dae4a38c9f616e2582a59701f3df6ae Mon Sep 17 00:00:00 2001
From: Terence Hampson <thampson@google.com>
Date: Thu, 15 Sep 2022 19:05:41 -0400
Subject: [PATCH] Disable MDNS lookup on first message deliver failure be
 default (#22678)

* Disable MDNS lookup on first message deliver failure

* Restyle

* Reorder config entry

* Address PR comments and clean up config comment

* Address PR comments
---
 src/messaging/BUILD.gn                        |  6 ++++++
 src/messaging/ReliableMessageMgr.cpp          |  2 ++
 src/messaging/ReliableMessageProtocolConfig.h | 20 +++++++++++++++++++
 3 files changed, 28 insertions(+)

diff --git a/src/messaging/BUILD.gn b/src/messaging/BUILD.gn
index f01b749d3f6d17..bf363c7239db71 100644
--- a/src/messaging/BUILD.gn
+++ b/src/messaging/BUILD.gn
@@ -17,6 +17,8 @@ import("//build_overrides/chip.gni")
 declare_args() {
   # Allows to change time between retries during the case session
   chip_case_retry_delta = ""
+
+  chip_config_resolve_peer_on_first_transmit_failure = ""
 }
 
 defines = []
@@ -26,6 +28,10 @@ if (chip_case_retry_delta != "") {
       [ "CHIP_CONFIG_MRP_LOCAL_ACTIVE_RETRY_INTERVAL=${chip_case_retry_delta}" ]
 }
 
+if (chip_config_resolve_peer_on_first_transmit_failure != "") {
+  defines += [ "CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE=${chip_config_resolve_peer_on_first_transmit_failure}" ]
+}
+
 source_set("messaging_mrp_config") {
   sources = [ "ReliableMessageProtocolConfig.h" ]
 
diff --git a/src/messaging/ReliableMessageMgr.cpp b/src/messaging/ReliableMessageMgr.cpp
index e745e54d46665e..f1b297b7b8d799 100644
--- a/src/messaging/ReliableMessageMgr.cpp
+++ b/src/messaging/ReliableMessageMgr.cpp
@@ -316,6 +316,7 @@ CHIP_ERROR ReliableMessageMgr::SendFromRetransTable(RetransTableEntry * entry)
 
     if (err == CHIP_NO_ERROR)
     {
+#if CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE
         const ExchangeManager * exchangeMgr = entry->ec->GetExchangeMgr();
         // TODO: investigate why in ReliableMessageMgr::CheckResendApplicationMessageWithPeerExchange unit test released exchange
         // context with mExchangeMgr==nullptr is used.
@@ -329,6 +330,7 @@ CHIP_ERROR ReliableMessageMgr::SendFromRetransTable(RetransTableEntry * entry)
                 mSessionUpdateDelegate->UpdatePeerAddress(entry->ec->GetSessionHandle()->GetPeer());
             }
         }
+#endif // CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE
     }
     else
     {
diff --git a/src/messaging/ReliableMessageProtocolConfig.h b/src/messaging/ReliableMessageProtocolConfig.h
index f05f4b05d9cc97..2e253c10bdc3ac 100644
--- a/src/messaging/ReliableMessageProtocolConfig.h
+++ b/src/messaging/ReliableMessageProtocolConfig.h
@@ -71,6 +71,26 @@ namespace chip {
 #define CHIP_CONFIG_RMP_DEFAULT_ACK_TIMEOUT (200_ms32)
 #endif // CHIP_CONFIG_RMP_DEFAULT_ACK_TIMEOUT
 
+/**
+ *  @def CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE
+ *
+ *  @brief
+ *    Should an address lookup of the peer happen on every first message that fails
+ *    to send on the link.
+ *
+ *  The default value to not perform lookup was selected because most implementations
+ *  of address lookup are not cache the and a request is sent on the link. Failing
+ *  to deliver the first message is far more likely to happen due to lossy link
+ *  than an actual address change where the peer did not reset. In the lossy link
+ *  situation, doing further DNS resolutions on a degraded link can exacerbate that
+ *  problem greatly. Additionally, every message that arrives from a peer updates the
+ *  address. If the peer has fallen off the link due to any other reason, a re-resolve
+ *  may not achieve an address that is reachable, even if a resolve response occurs.
+ */
+#ifndef CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE
+#define CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE 0
+#endif // CHIP_CONFIG_RESOLVE_PEER_ON_FIRST_TRANSMIT_FAILURE
+
 /**
  *  @def CHIP_CONFIG_RMP_RETRANS_TABLE_SIZE
  *