X-Git-Url: http://xvm.mit.edu/gitweb/invirt/third/libt4.git/blobdiff_plain/5d99dbf06a14904944f5593c63705934bdfdcfb7..603bac8fcb3697f283e6537d81b4a92e457ebbad:/rsm.cc

diff --git a/rsm.cc b/rsm.cc
index 843418a..81f0e4c 100644
--- a/rsm.cc
+++ b/rsm.cc
@@ -86,15 +86,10 @@
 #include "rsm.h"
 #include "rsm_client.h"
 
-rsm::rsm(string _first, string _me) :
+rsm::rsm(const string & _first, const string & _me) :
     stf(0), primary(_first), insync (false), inviewchange (true), vid_commit(0),
     partitioned (false), dopartition(false), break1(false), break2(false)
 {
-    last_myvs.vid = 0;
-    last_myvs.seqno = 0;
-    myvs = last_myvs;
-    myvs.seqno = 1;
-
     cfg = new config(_first, _me, this);
 
     if (_first == _me) {
@@ -111,7 +106,7 @@ rsm::rsm(string _first, string _me) :
     rsmrpc->reg(rsm_protocol::joinreq, &rsm::joinreq, this);
 
     // tester must be on different port, otherwise it may partition itself
-    testsvr = new rpcs((uint32_t)stoi(_me) + 1);
+    testsvr = new rpcs((in_port_t)stoi(_me) + 1);
     testsvr->reg(rsm_test_protocol::net_repair, &rsm::test_net_repairreq, this);
     testsvr->reg(rsm_test_protocol::breakpoint, &rsm::breakpointreq, this);
 
@@ -140,7 +135,7 @@ void rsm::recovery() [[noreturn]] {
                 commit_change(cfg->view_id(), ml);
             } else {
                 ml.unlock();
-                this_thread::sleep_for(seconds(30)); // XXX make another node in cfg primary?
+                this_thread::sleep_for(seconds(3)); // XXX make another node in cfg primary?
                 ml.lock();
             }
         }
@@ -186,7 +181,7 @@ bool rsm::sync_with_backups(lock & rsm_mutex_lock) {
     insync = true;
     cfg->get_view(vid_insync, backups);
     backups.erase(find(backups.begin(), backups.end(), cfg->myaddr()));
-    LOG("rsm::sync_with_backups " << backups);
+    LOG("backups " << backups);
     sync_cond.wait(rsm_mutex_lock);
     insync = false;
     return true;
@@ -204,39 +199,39 @@ bool rsm::sync_with_primary(lock & rsm_mutex_lock) {
 }
 
 
-/**
- * Call to transfer state from m to the local node.
- * Assumes that rsm_mutex is already held.
- */
-bool rsm::statetransfer(string m, lock & rsm_mutex_lock)
+//
+// Call to transfer state from m to the local node.
+// Assumes that rsm_mutex is already held.
+//
+bool rsm::statetransfer(const string & m, lock & rsm_mutex_lock)
 {
     rsm_protocol::transferres r;
     handle h(m);
     int ret = 0;
-    LOG("rsm::statetransfer: contact " << m << " w. my last_myvs(" << last_myvs.vid << "," << last_myvs.seqno << ")");
+    LOG("contact " << m << " w. my last_myvs(" << last_myvs.vid << "," << last_myvs.seqno << ")");
     rpcc *cl;
     {
         rsm_mutex_lock.unlock();
         cl = h.safebind();
         if (cl) {
-            ret = cl->call_timeout(rsm_protocol::transferreq, rpcc::to(1000),
+            ret = cl->call_timeout(rsm_protocol::transferreq, rpcc::to(100),
                     r, cfg->myaddr(), last_myvs, vid_insync);
         }
         rsm_mutex_lock.lock();
     }
     if (cl == 0 || ret != rsm_protocol::OK) {
-        LOG("rsm::statetransfer: couldn't reach " << m << " " << hex << cl << " " << dec << ret);
+        LOG("couldn't reach " << m << " " << hex << cl << " " << dec << ret);
         return false;
     }
     if (stf && last_myvs != r.last) {
         stf->unmarshal_state(r.state);
     }
     last_myvs = r.last;
-    LOG("rsm::statetransfer transfer from " << m << " success, vs(" << last_myvs.vid << "," << last_myvs.seqno << ")");
+    LOG("transfer from " << m << " success, vs(" << last_myvs.vid << "," << last_myvs.seqno << ")");
     return true;
 }
 
-bool rsm::statetransferdone(string m, lock & rsm_mutex_lock) {
+bool rsm::statetransferdone(const string & m, lock & rsm_mutex_lock) {
     rsm_mutex_lock.unlock();
     handle h(m);
     rpcc *cl = h.safebind();
@@ -251,36 +246,36 @@ bool rsm::statetransferdone(string m, lock & rsm_mutex_lock) {
 }
 
 
-bool rsm::join(string m, lock & rsm_mutex_lock) {
+bool rsm::join(const string & m, lock & rsm_mutex_lock) {
     handle h(m);
     int ret = 0;
     string log;
 
-    LOG("rsm::join: " << m << " mylast (" << last_myvs.vid << "," << last_myvs.seqno << ")");
+    LOG("contacting " << m << " mylast (" << last_myvs.vid << "," << last_myvs.seqno << ")");
     rpcc *cl;
     {
         rsm_mutex_lock.unlock();
         cl = h.safebind();
         if (cl != 0) {
-            ret = cl->call_timeout(rsm_protocol::joinreq, rpcc::to(120000), log,
+            ret = cl->call_timeout(rsm_protocol::joinreq, rpcc::to(12000), log,
                     cfg->myaddr(), last_myvs);
         }
         rsm_mutex_lock.lock();
     }
 
     if (cl == 0 || ret != rsm_protocol::OK) {
-        LOG("rsm::join: couldn't reach " << m << " " << hex << cl << " " << dec << ret);
+        LOG("couldn't reach " << m << " " << hex << cl << " " << dec << ret);
         return false;
     }
-    LOG("rsm::join: succeeded " << log);
+    LOG("succeeded " << log);
     cfg->restore(log);
     return true;
 }
 
-/*
- * Config informs rsm whenever it has successfully
- * completed a view change
- */
+//
+// Config informs rsm whenever it has successfully
+// completed a view change
+//
 void rsm::commit_change(unsigned vid) {
     lock ml(rsm_mutex);
     commit_change(vid, ml);
@@ -303,18 +298,14 @@ void rsm::commit_change(unsigned vid, lock &) {
 }
 
 
-void rsm::execute(int procno, string req, string &r) {
+void rsm::execute(int procno, const string & req, string & r) {
     LOG("execute");
     handler *h = procs[procno];
     VERIFY(h);
     unmarshall args(req, false);
     marshall rep;
-    string reps;
     auto ret = (rsm_protocol::status)(*h)(args, rep);
-    marshall rep1;
-    rep1 << ret;
-    rep1 << rep.content();
-    r = rep1.content();
+    r = marshall{ret, rep.content()}.content();
 }
 
 //
@@ -323,8 +314,8 @@ void rsm::execute(int procno, string req, string &r) {
 // number, and invokes it on all members of the replicated state
 // machine.
 //
-rsm_client_protocol::status rsm::client_invoke(string &r, int procno, string req) {
-    LOG("rsm::client_invoke: procno 0x" << hex << procno);
+rsm_client_protocol::status rsm::client_invoke(string & r, int procno, const string & req) {
+    LOG("invoke procno 0x" << hex << procno);
     lock ml(invoke_mutex);
     vector<string> m;
     string myaddr;
@@ -356,7 +347,7 @@ rsm_client_protocol::status rsm::client_invoke(string &r, int procno, string req
             if (!cl)
                 return rsm_client_protocol::BUSY;
             int ignored_rval;
-            auto ret = (rsm_protocol::status)cl->call_timeout(rsm_protocol::invoke, rpcc::to(1000), ignored_rval, procno, vs, req);
+            auto ret = (rsm_protocol::status)cl->call_timeout(rsm_protocol::invoke, rpcc::to(100), ignored_rval, procno, vs, req);
             LOG("Invoke returned " << ret);
             if (ret != rsm_protocol::OK)
                 return rsm_client_protocol::BUSY;
@@ -377,8 +368,8 @@ rsm_client_protocol::status rsm::client_invoke(string &r, int procno, string req
 // the replica must execute requests in order (with no gaps)
 // according to requests' seqno
 
-rsm_protocol::status rsm::invoke(int &, int proc, viewstamp vs, string req) {
-    LOG("rsm::invoke: procno 0x" << hex << proc);
+rsm_protocol::status rsm::invoke(int &, int proc, viewstamp vs, const string & req) {
+    LOG("invoke procno 0x" << hex << proc);
     lock ml(invoke_mutex);
     vector<string> m;
     string myaddr;
@@ -409,10 +400,10 @@ rsm_protocol::status rsm::invoke(int &, int proc, viewstamp vs, string req) {
     return rsm_protocol::OK;
 }
 
-/**
- * RPC handler: Send back the local node's state to the caller
- */
-rsm_protocol::status rsm::transferreq(rsm_protocol::transferres &r, string src,
+//
+// RPC handler: Send back the local node's state to the caller
+//
+rsm_protocol::status rsm::transferreq(rsm_protocol::transferres &r, const string & src,
         viewstamp last, unsigned vid) {
     lock ml(rsm_mutex);
     LOG("transferreq from " << src << " (" << last.vid << "," << last.seqno << ") vs (" <<
@@ -425,11 +416,11 @@ rsm_protocol::status rsm::transferreq(rsm_protocol::transferres &r, string src,
     return rsm_protocol::OK;
 }
 
-/**
- * RPC handler: Inform the local node (the primary) that node m has synchronized
- * for view vid
- */
-rsm_protocol::status rsm::transferdonereq(int &, string m, unsigned vid) {
+//
+// RPC handler: Inform the local node (the primary) that node m has synchronized
+// for view vid
+//
+rsm_protocol::status rsm::transferdonereq(int &, const string & m, unsigned vid) {
     lock ml(rsm_mutex);
     if (!insync || vid != vid_insync)
         return rsm_protocol::BUSY;
@@ -442,7 +433,7 @@ rsm_protocol::status rsm::transferdonereq(int &, string m, unsigned vid) {
 // a node that wants to join an RSM as a server sends a
 // joinreq to the RSM's current primary; this is the
 // handler for that RPC.
-rsm_protocol::status rsm::joinreq(string & log, string m, viewstamp last) {
+rsm_protocol::status rsm::joinreq(string & log, const string & m, viewstamp last) {
     auto ret = rsm_protocol::OK;
 
     lock ml(rsm_mutex);
@@ -476,18 +467,18 @@ rsm_protocol::status rsm::joinreq(string & log, string m, viewstamp last) {
     return ret;
 }
 
-/*
- * RPC handler: Send back all the nodes this local knows about to client
- * so the client can switch to a different primary
- * when it existing primary fails
- */
+//
+// RPC handler: Send back all the nodes this local knows about to client
+// so the client can switch to a different primary
+// when it existing primary fails
+//
 rsm_client_protocol::status rsm::client_members(vector<string> &r, int) {
     vector<string> m;
     lock ml(rsm_mutex);
     cfg->get_view(vid_commit, m);
     m.push_back(primary);
     r = m;
-    LOG("rsm::client_members return " << m << " m " << primary);
+    LOG("return " << m << " m " << primary);
     return rsm_client_protocol::OK;
 }
 
@@ -533,7 +524,7 @@ void rsm::net_repair(bool heal, lock &) {
     for (unsigned i  = 0; i < m.size(); i++) {
         if (m[i] != cfg->myaddr()) {
             handle h(m[i]);
-            LOG("rsm::net_repair: " << m[i] << " " << heal);
+            LOG("member " << m[i] << " " << heal);
             if (h.safebind()) h.safebind()->set_reachable(heal);
         }
     }
@@ -542,7 +533,7 @@ void rsm::net_repair(bool heal, lock &) {
 
 rsm_test_protocol::status rsm::test_net_repairreq(rsm_test_protocol::status &r, int heal) {
     lock ml(rsm_mutex);
-    LOG("rsm::test_net_repairreq: " << heal << " (dopartition " <<
+    LOG("heal " << heal << " (dopartition " <<
             dopartition << ", partitioned " << partitioned << ")");
     if (heal) {
         net_repair(heal, ml);
@@ -582,7 +573,7 @@ void rsm::partition1(lock & rsm_mutex_lock) {
 rsm_test_protocol::status rsm::breakpointreq(rsm_test_protocol::status &r, int b) {
     r = rsm_test_protocol::OK;
     lock ml(rsm_mutex);
-    LOG("rsm::breakpointreq: " << b);
+    LOG("breakpoint " << b);
     if (b == 1) break1 = true;
     else if (b == 2) break2 = true;
     else if (b == 3 || b == 4) cfg->breakpoint(b);