From b1cf2104756e4b97ba30891c9bec0ed68a182002 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@redhat.com>
Date: Sun, 24 Aug 2014 21:01:07 -0700
Subject: [PATCH] mon/Paxos: WRITING != WRITING_PREVIOUS

This distinction is important: the update-previous state should not be
writeable, as reflected by PaxosService::is_writeable().

Signed-off-by: Sage Weil <sage@redhat.com>
---
 src/mon/Monitor.cc      |  4 ++--
 src/mon/Paxos.cc        | 14 ++++++++++----
 src/mon/Paxos.h         |  9 +++++++++
 src/mon/PaxosService.cc |  3 ++-
 4 files changed, 23 insertions(+), 7 deletions(-)

diff --git a/src/mon/Monitor.cc b/src/mon/Monitor.cc
index f9327ef9e2a6b..ced69a5b09662 100644
--- a/src/mon/Monitor.cc
+++ b/src/mon/Monitor.cc
@@ -781,7 +781,7 @@ void Monitor::shutdown()
 
 void Monitor::bootstrap()
 {
-  if (paxos->is_writing()) {
+  if (paxos->is_writing() || paxos->is_writing_previous()) {
     dout(10) << "bootstrap flushing pending write" << dendl;
     lock.Unlock();
     store->flush();
@@ -2387,7 +2387,7 @@ void Monitor::handle_command(MMonCommand *m)
   }
 
   if (prefix == "scrub") {
-    while (paxos->is_writing()) {
+    while (paxos->is_writing() || paxos->is_writing_previous()) {
       lock.Unlock();
       store->flush();
       lock.Lock();
diff --git a/src/mon/Paxos.cc b/src/mon/Paxos.cc
index c85beaabbf41a..e5208527055e4 100644
--- a/src/mon/Paxos.cc
+++ b/src/mon/Paxos.cc
@@ -796,7 +796,8 @@ void Paxos::accept_timeout()
   dout(1) << "accept timeout, calling fresh election" << dendl;
   accept_timeout_event = 0;
   assert(mon->is_leader());
-  assert(is_updating() || is_updating_previous() || is_writing());
+  assert(is_updating() || is_updating_previous() || is_writing() ||
+	 is_writing_previous());
   logger->inc(l_paxos_accept_timeout);
   mon->bootstrap();
 }
@@ -839,7 +840,12 @@ void Paxos::commit_start()
 
   get_store()->queue_transaction(t, new C_Committed(this));
 
-  state = STATE_WRITING;
+  if (is_updating_previous())
+    state = STATE_WRITING_PREVIOUS;
+  else if (is_updating())
+    state = STATE_WRITING;
+  else
+    assert(0);
 }
 
 void Paxos::commit_finish()
@@ -889,7 +895,7 @@ void Paxos::commit_finish()
   // WRITING -> REFRESH
   // among other things, this lets do_refresh() -> mon->bootstrap() know
   // it doesn't need to flush the store queue
-  assert(state == STATE_WRITING);
+  assert(is_writing() || is_writing_previous());
   state = STATE_REFRESH;
 
   if (do_refresh()) {
@@ -1339,7 +1345,7 @@ void Paxos::restart()
   cancel_events();
   new_value.clear();
 
-  if (is_writing()) {
+  if (is_writing() || is_writing_previous()) {
     dout(10) << __func__ << " flushing" << dendl;
     mon->lock.Unlock();
     mon->store->flush();
diff --git a/src/mon/Paxos.h b/src/mon/Paxos.h
index aa591de7658e0..dcba45805b0cf 100644
--- a/src/mon/Paxos.h
+++ b/src/mon/Paxos.h
@@ -228,6 +228,10 @@ public:
      * writeable.
      */
     STATE_WRITING,
+    /*
+     * Leader/Peon is writing a new commit from a previous round.
+     */
+    STATE_WRITING_PREVIOUS,
     // leader: refresh following a commit
     STATE_REFRESH,
   };
@@ -253,6 +257,8 @@ public:
       return "updating-previous";
     case STATE_WRITING:
       return "writing";
+    case STATE_WRITING_PREVIOUS:
+      return "writing-previous";
     case STATE_REFRESH:
       return "refresh";
     default:
@@ -298,6 +304,9 @@ public:
   /// @return 'true' if we are writing an update to disk
   bool is_writing() const { return state == STATE_WRITING; }
 
+  /// @return 'true' if we are writing an update-previous to disk
+  bool is_writing_previous() const { return state == STATE_WRITING_PREVIOUS; }
+
   /// @return 'true' if we are refreshing an update just committed
   bool is_refresh() const { return state == STATE_REFRESH; }
 
diff --git a/src/mon/PaxosService.cc b/src/mon/PaxosService.cc
index aba8c959a94da..92257209565f8 100644
--- a/src/mon/PaxosService.cc
+++ b/src/mon/PaxosService.cc
@@ -53,7 +53,8 @@ bool PaxosService::dispatch(PaxosServiceMessage *m)
   // make sure the client is still connected.  note that a proxied
   // connection will be disconnected with a null message; don't drop
   // those.  also ignore loopback (e.g., log) messages.
-  if (!m->get_connection()->is_connected() &&
+  if (m->get_connection() &&
+      !m->get_connection()->is_connected() &&
       m->get_connection() != mon->con_self &&
       m->get_connection()->get_messenger() != NULL) {
     dout(10) << " discarding message from disconnected client "
-- 
2.39.5