Reduced timeouts by 10x

[invirt/third/libt4.git] / rsm.cc

diff --git a/rsm.cc b/rsm.cc
index 66c8d97..81f0e4c 100644 (file)
--- a/rsm.cc
+++ b/rsm.cc
@@ -78,28 +78,18 @@
 // The rule is that a module releases its internal locks before it
 // upcalls, but can keep its locks when calling down.
 
 // The rule is that a module releases its internal locks before it
 // upcalls, but can keep its locks when calling down.
 

-#include <fstream>
-#include <iostream>
-#include <algorithm>

 #include <sys/types.h>
 #include <unistd.h>
 
 #include <sys/types.h>
 #include <unistd.h>
 

+#include "types.h"

 #include "handle.h"
 #include "rsm.h"
 #include "handle.h"
 #include "rsm.h"

-#include "tprintf.h"
-#include "lang/verify.h"

 #include "rsm_client.h"
 #include "rsm_client.h"

-#include "lock.h"

 
 

-rsm::rsm(std::string _first, std::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)
 {
     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) {
     cfg = new config(_first, _me, this);
 
     if (_first == _me) {

@@ -116,13 +106,13 @@ rsm::rsm(std::string _first, std::string _me) :
     rsmrpc->reg(rsm_protocol::joinreq, &rsm::joinreq, this);
 
     // tester must be on different port, otherwise it may partition itself
     rsmrpc->reg(rsm_protocol::joinreq, &rsm::joinreq, this);
 
     // tester must be on different port, otherwise it may partition itself

-    testsvr = new rpcs(atoi(_me.c_str()) + 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);
 
     {
         lock ml(rsm_mutex);
     testsvr->reg(rsm_test_protocol::net_repair, &rsm::test_net_repairreq, this);
     testsvr->reg(rsm_test_protocol::breakpoint, &rsm::breakpointreq, this);
 
     {
         lock ml(rsm_mutex);

-        std::thread(&rsm::recovery, this).detach();
+        thread(&rsm::recovery, this).detach();

     }
 }
 
     }
 }
 

@@ -132,7 +122,7 @@ void rsm::reg1(int proc, handler *h) {
 }
 
 // The recovery thread runs this function
 }
 
 // The recovery thread runs this function

-void rsm::recovery() {
+void rsm::recovery() [[noreturn]] {

     bool r = true;
     lock ml(rsm_mutex);
 
     bool r = true;
     lock ml(rsm_mutex);
 

@@ -140,23 +130,23 @@ void rsm::recovery() {
         while (!cfg->ismember(cfg->myaddr(), vid_commit)) {
             // XXX iannucci 2013/09/15 -- I don't understand whether accessing
             // cfg->view_id in this manner involves a race.  I suspect not.
         while (!cfg->ismember(cfg->myaddr(), vid_commit)) {
             // XXX iannucci 2013/09/15 -- I don't understand whether accessing
             // cfg->view_id in this manner involves a race.  I suspect not.

-            if (join(primary)) {
-                tprintf("recovery: joined\n");
-                commit_change_wo(cfg->view_id());
+            if (join(primary, ml)) {
+                LOG("recovery: joined");
+                commit_change(cfg->view_id(), ml);

             } else {
                 ml.unlock();
             } else {
                 ml.unlock();

-                std::this_thread::sleep_for(std::chrono::seconds(30)); // XXX make another node in cfg primary?
+                this_thread::sleep_for(seconds(3)); // XXX make another node in cfg primary?

                 ml.lock();
             }
         }
         vid_insync = vid_commit;
                 ml.lock();
             }
         }
         vid_insync = vid_commit;

-        tprintf("recovery: sync vid_insync %d\n", vid_insync);
+        LOG("recovery: sync vid_insync " << vid_insync);

         if (primary == cfg->myaddr()) {
         if (primary == cfg->myaddr()) {

-            r = sync_with_backups();
+            r = sync_with_backups(ml);

         } else {
         } else {

-            r = sync_with_primary();
+            r = sync_with_primary(ml);

         }
         }

-        tprintf("recovery: sync done\n");
+        LOG("recovery: sync done");

 
         // If there was a commited viewchange during the synchronization, restart
         // the recovery
 
         // If there was a commited viewchange during the synchronization, restart
         // the recovery

@@ -168,157 +158,136 @@ void rsm::recovery() {
             myvs.seqno = 1;
             inviewchange = false;
         }
             myvs.seqno = 1;
             inviewchange = false;
         }

-        tprintf("recovery: go to sleep %d %d\n", insync, inviewchange);
+        LOG("recovery: go to sleep " << insync << " " << inviewchange);

         recovery_cond.wait(ml);
     }
 }
 
         recovery_cond.wait(ml);
     }
 }
 

-template <class A>
-std::ostream & operator<<(std::ostream &o, const std::vector<A> &d) {
-    o << "[";
-    for (typename std::vector<A>::const_iterator i=d.begin(); i!=d.end(); i++) {
-        o << *i;
-        if (i+1 != d.end())
-            o << ", ";
-    }
-    o << "]";
-    return o;
-}
-
-bool rsm::sync_with_backups() {
-    adopt_lock ml(rsm_mutex);
-    ml.unlock();
+bool rsm::sync_with_backups(lock & rsm_mutex_lock) {
+    rsm_mutex_lock.unlock();

     {
     {

-        // Make sure that the state of lock_server_cache_rsm is stable during
+        // Make sure that the state of lock_server is stable during

         // synchronization; otherwise, the primary's state may be more recent
         // than replicas after the synchronization.
         // synchronization; otherwise, the primary's state may be more recent
         // than replicas after the synchronization.

-        lock ml(invoke_mutex);
+        lock invoke_mutex_lock(invoke_mutex);

         // By acquiring and releasing the invoke_mutex once, we make sure that
         // By acquiring and releasing the invoke_mutex once, we make sure that

-        // the state of lock_server_cache_rsm will not be changed until all
+        // the state of lock_server will not be changed until all

         // replicas are synchronized. The reason is that client_invoke arrives
         // after this point of time will see inviewchange == true, and returns
         // BUSY.
     }
         // replicas are synchronized. The reason is that client_invoke arrives
         // after this point of time will see inviewchange == true, and returns
         // BUSY.
     }

-    ml.lock();
+    rsm_mutex_lock.lock();

     // Start accepting synchronization request (statetransferreq) now!
     insync = true;
     cfg->get_view(vid_insync, backups);
     backups.erase(find(backups.begin(), backups.end(), cfg->myaddr()));
     // Start accepting synchronization request (statetransferreq) now!
     insync = true;
     cfg->get_view(vid_insync, backups);
     backups.erase(find(backups.begin(), backups.end(), cfg->myaddr()));

-    LOG("rsm::sync_with_backups " << backups);
-    sync_cond.wait(ml);
+    LOG("backups " << backups);
+    sync_cond.wait(rsm_mutex_lock);

     insync = false;
     return true;
 }
 
 
     insync = false;
     return true;
 }
 
 

-bool rsm::sync_with_primary() {
+bool rsm::sync_with_primary(lock & rsm_mutex_lock) {

     // Remember the primary of vid_insync
     // Remember the primary of vid_insync

-    std::string m = primary;
+    string m = primary;

     while (vid_insync == vid_commit) {
     while (vid_insync == vid_commit) {

-        if (statetransfer(m))
+        if (statetransfer(m, rsm_mutex_lock))

             break;
     }
             break;
     }

-    return statetransferdone(m);
+    return statetransferdone(m, rsm_mutex_lock);

 }
 
 
 }
 
 

-/**
- * Call to transfer state from m to the local node.
- * Assumes that rsm_mutex is already held.
- */
-bool rsm::statetransfer(std::string m)
+//
+// 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;
 {
     rsm_protocol::transferres r;
     handle h(m);
     int ret = 0;

-    tprintf("rsm::statetransfer: contact %s w. my last_myvs(%d,%d)\n",
-            m.c_str(), last_myvs.vid, last_myvs.seqno);
+    LOG("contact " << m << " w. my last_myvs(" << last_myvs.vid << "," << last_myvs.seqno << ")");

     rpcc *cl;
     {
     rpcc *cl;
     {

-        adopt_lock ml(rsm_mutex);
-        ml.unlock();
+        rsm_mutex_lock.unlock();

         cl = h.safebind();
         if (cl) {
         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);
         }
                     r, cfg->myaddr(), last_myvs, vid_insync);
         }

-        ml.lock();
+        rsm_mutex_lock.lock();

     }
     if (cl == 0 || ret != rsm_protocol::OK) {
     }
     if (cl == 0 || ret != rsm_protocol::OK) {

-        tprintf("rsm::statetransfer: couldn't reach %s %lx %d\n", m.c_str(),
-                (long unsigned) cl, 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;
         return false;
     }
     if (stf && last_myvs != r.last) {
         stf->unmarshal_state(r.state);
     }
     last_myvs = r.last;

-    tprintf("rsm::statetransfer transfer from %s success, vs(%d,%d)\n",
-            m.c_str(), last_myvs.vid, last_myvs.seqno);
+    LOG("transfer from " << m << " success, vs(" << last_myvs.vid << "," << last_myvs.seqno << ")");

     return true;
 }
 
     return true;
 }
 

-bool rsm::statetransferdone(std::string m) {
-    adopt_lock ml(rsm_mutex);
-    ml.unlock();
+bool rsm::statetransferdone(const string & m, lock & rsm_mutex_lock) {
+    rsm_mutex_lock.unlock();

     handle h(m);
     rpcc *cl = h.safebind();
     bool done = false;
     if (cl) {
         int r;
     handle h(m);
     rpcc *cl = h.safebind();
     bool done = false;
     if (cl) {
         int r;

-        rsm_protocol::status ret = cl->call(rsm_protocol::transferdonereq, r, cfg->myaddr(), vid_insync);
+        auto ret = (rsm_protocol::status)cl->call(rsm_protocol::transferdonereq, r, cfg->myaddr(), vid_insync);

         done = (ret == rsm_protocol::OK);
     }
         done = (ret == rsm_protocol::OK);
     }

-    ml.lock();
+    rsm_mutex_lock.lock();

     return done;
 }
 
 
     return done;
 }
 
 

-bool rsm::join(std::string m) {
+bool rsm::join(const string & m, lock & rsm_mutex_lock) {

     handle h(m);
     int ret = 0;
     handle h(m);
     int ret = 0;

-    rsm_protocol::joinres r;
+    string log;

 
 

-    tprintf("rsm::join: %s mylast (%d,%d)\n", m.c_str(), last_myvs.vid,
-            last_myvs.seqno);
+    LOG("contacting " << m << " mylast (" << last_myvs.vid << "," << last_myvs.seqno << ")");

     rpcc *cl;
     {
     rpcc *cl;
     {

-        adopt_lock ml(rsm_mutex);
-        ml.unlock();
+        rsm_mutex_lock.unlock();

         cl = h.safebind();
         if (cl != 0) {
         cl = h.safebind();
         if (cl != 0) {

-            ret = cl->call_timeout(rsm_protocol::joinreq, rpcc::to(120000), r,
+            ret = cl->call_timeout(rsm_protocol::joinreq, rpcc::to(12000), log,

                     cfg->myaddr(), last_myvs);
         }
                     cfg->myaddr(), last_myvs);
         }

-        ml.lock();
+        rsm_mutex_lock.lock();

     }
 
     if (cl == 0 || ret != rsm_protocol::OK) {
     }
 
     if (cl == 0 || ret != rsm_protocol::OK) {

-        tprintf("rsm::join: couldn't reach %s %p %d\n", m.c_str(),
-                cl, ret);
+        LOG("couldn't reach " << m << " " << hex << cl << " " << dec << ret);

         return false;
     }
         return false;
     }

-    tprintf("rsm::join: succeeded %s\n", r.log.c_str());
-    cfg->restore(r.log);
+    LOG("succeeded " << log);
+    cfg->restore(log);

     return true;
 }
 
     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);
 void rsm::commit_change(unsigned vid) {
     lock ml(rsm_mutex);

-    commit_change_wo(vid);
+    commit_change(vid, ml);

     if (cfg->ismember(cfg->myaddr(), vid_commit))
         breakpoint2();
 }
 
     if (cfg->ismember(cfg->myaddr(), vid_commit))
         breakpoint2();
 }
 

-void rsm::commit_change_wo(unsigned vid) {
+void rsm::commit_change(unsigned vid, lock &) {

     if (vid <= vid_commit)
         return;
     if (vid <= vid_commit)
         return;

-    tprintf("commit_change: new view (%d)  last vs (%d,%d) %s insync %d\n",
-            vid, last_myvs.vid, last_myvs.seqno, primary.c_str(), insync);
+    LOG("commit_change: new view (" << vid << ") last vs (" << last_myvs.vid << "," <<
+            last_myvs.seqno << ") " << primary << " insync " << insync);

     vid_commit = vid;
     inviewchange = true;
     set_primary(vid);
     vid_commit = vid;
     inviewchange = true;
     set_primary(vid);

@@ -329,18 +298,14 @@ void rsm::commit_change_wo(unsigned vid) {
 }
 
 
 }
 
 

-void rsm::execute(int procno, std::string req, std::string &r) {
-    tprintf("execute\n");
+void rsm::execute(int procno, const string & req, string & r) {
+    LOG("execute");

     handler *h = procs[procno];
     VERIFY(h);
     handler *h = procs[procno];
     VERIFY(h);

-    unmarshall args(req);
+    unmarshall args(req, false);

     marshall rep;
     marshall rep;

-    std::string reps;
-    rsm_protocol::status ret = (*h)(args, rep);
-    marshall rep1;
-    rep1 << ret;
-    rep1 << rep.str();
-    r = rep1.str();
+    auto ret = (rsm_protocol::status)(*h)(args, rep);
+    r = marshall{ret, rep.content()}.content();

 }
 
 //
 }
 
 //

@@ -349,14 +314,14 @@ void rsm::execute(int procno, std::string req, std::string &r) {
 // number, and invokes it on all members of the replicated state
 // machine.
 //
 // number, and invokes it on all members of the replicated state
 // machine.
 //

-rsm_client_protocol::status rsm::client_invoke(std::string &r, int procno, std::string req) {
-    LOG("rsm::client_invoke: procno 0x" << std::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);
     lock ml(invoke_mutex);

-    std::vector<std::string> m;
-    std::string myaddr;
+    vector<string> m;
+    string myaddr;

     viewstamp vs;
     {
     viewstamp vs;
     {

-        lock ml(rsm_mutex);
+        lock ml2(rsm_mutex);

         LOG("Checking for inviewchange");
         if (inviewchange)
             return rsm_client_protocol::BUSY;
         LOG("Checking for inviewchange");
         if (inviewchange)
             return rsm_client_protocol::BUSY;

@@ -381,14 +346,14 @@ rsm_client_protocol::status rsm::client_invoke(std::string &r, int procno, std::
             rpcc *cl = h.safebind();
             if (!cl)
                 return rsm_client_protocol::BUSY;
             rpcc *cl = h.safebind();
             if (!cl)
                 return rsm_client_protocol::BUSY;

-            rsm_protocol::status ret;
-            int r;
-            ret = cl->call_timeout(rsm_protocol::invoke, rpcc::to(1000), r, procno, vs, req);
+            int ignored_rval;
+            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;
             breakpoint1();
             LOG("Invoke returned " << ret);
             if (ret != rsm_protocol::OK)
                 return rsm_client_protocol::BUSY;
             breakpoint1();

-            partition1();
+            lock rsm_mutex_lock(rsm_mutex);
+            partition1(rsm_mutex_lock);

         }
     }
     execute(procno, req, r);
         }
     }
     execute(procno, req, r);

@@ -403,13 +368,13 @@ rsm_client_protocol::status rsm::client_invoke(std::string &r, int procno, std::
 // the replica must execute requests in order (with no gaps)
 // according to requests' seqno
 
 // the replica must execute requests in order (with no gaps)
 // according to requests' seqno
 

-rsm_protocol::status rsm::invoke(int &, int proc, viewstamp vs, std::string req) {
-    LOG("rsm::invoke: procno 0x" << std::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);
     lock ml(invoke_mutex);

-    std::vector<std::string> m;
-    std::string myaddr;
+    vector<string> m;
+    string myaddr;

     {
     {

-        lock ml(rsm_mutex);
+        lock ml2(rsm_mutex);

         // check if !inviewchange
         LOG("Checking for view change");
         if (inviewchange)
         // check if !inviewchange
         LOG("Checking for view change");
         if (inviewchange)

@@ -428,36 +393,34 @@ rsm_protocol::status rsm::invoke(int &, int proc, viewstamp vs, std::string req)
             return rsm_protocol::ERR;
         myvs++;
     }
             return rsm_protocol::ERR;
         myvs++;
     }

-    std::string r;
+    string r;

     execute(proc, req, r);
     last_myvs = vs;
     breakpoint1();
     return rsm_protocol::OK;
 }
 
     execute(proc, req, r);
     last_myvs = vs;
     breakpoint1();
     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, std::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);
         viewstamp last, unsigned vid) {
     lock ml(rsm_mutex);

-    int ret = rsm_protocol::OK;
-    tprintf("transferreq from %s (%d,%d) vs (%d,%d)\n", src.c_str(),
-            last.vid, last.seqno, last_myvs.vid, last_myvs.seqno);
-    if (!insync || vid != vid_insync) {
+    LOG("transferreq from " << src << " (" << last.vid << "," << last.seqno << ") vs (" <<
+            last_myvs.vid << "," << last_myvs.seqno << ")");
+    if (!insync || vid != vid_insync)

         return rsm_protocol::BUSY;
         return rsm_protocol::BUSY;

-    }

     if (stf && last != last_myvs)
         r.state = stf->marshal_state();
     r.last = last_myvs;
     if (stf && last != last_myvs)
         r.state = stf->marshal_state();
     r.last = last_myvs;

-    return ret;
+    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 &, std::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;
     lock ml(rsm_mutex);
     if (!insync || vid != vid_insync)
         return rsm_protocol::BUSY;

@@ -470,21 +433,22 @@ rsm_protocol::status rsm::transferdonereq(int &, std::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.
 // 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(rsm_protocol::joinres &r, std::string m, viewstamp last) {
-    int ret = rsm_protocol::OK;
+rsm_protocol::status rsm::joinreq(string & log, const string & m, viewstamp last) {
+    auto ret = rsm_protocol::OK;

 
     lock ml(rsm_mutex);
 
     lock ml(rsm_mutex);

-    tprintf("joinreq: src %s last (%d,%d) mylast (%d,%d)\n", m.c_str(),
-            last.vid, last.seqno, last_myvs.vid, last_myvs.seqno);
+    LOG("join request from " << m << "; last=(" << last.vid << "," << last.seqno << "), mylast=(" <<
+            last_myvs.vid << "," << last_myvs.seqno << ")");

     if (cfg->ismember(m, vid_commit)) {
     if (cfg->ismember(m, vid_commit)) {

-        tprintf("joinreq: is still a member\n");
-        r.log = cfg->dump();
+        LOG(m << " is still a member -- nothing to do");
+        log = cfg->dump();

     } else if (cfg->myaddr() != primary) {
     } else if (cfg->myaddr() != primary) {

-        tprintf("joinreq: busy\n");
+        LOG("but I, " << cfg->myaddr() << ", am not the primary, " << primary << "!");

         ret = rsm_protocol::BUSY;
     } else {
         // We cache vid_commit to avoid adding m to a view which already contains
         // m due to race condition
         ret = rsm_protocol::BUSY;
     } else {
         // We cache vid_commit to avoid adding m to a view which already contains
         // m due to race condition

+        LOG("calling down to config layer");

         unsigned vid_cache = vid_commit;
         bool succ;
         {
         unsigned vid_cache = vid_commit;
         bool succ;
         {

@@ -493,29 +457,28 @@ rsm_protocol::status rsm::joinreq(rsm_protocol::joinres &r, std::string m, views
             ml.lock();
         }
         if (cfg->ismember(m, cfg->view_id())) {
             ml.lock();
         }
         if (cfg->ismember(m, cfg->view_id())) {

-            r.log = cfg->dump();
-            tprintf("joinreq: ret %d log %s\n:", ret, r.log.c_str());
+            log = cfg->dump();
+            LOG("ret " << ret << " log " << log);

         } else {
         } else {

-            tprintf("joinreq: failed; proposer couldn't add %d\n", succ);
+            LOG("failed; proposer couldn't add " << succ);

             ret = rsm_protocol::BUSY;
         }
     }
     return ret;
 }
 
             ret = rsm_protocol::BUSY;
         }
     }
     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
- */
-rsm_client_protocol::status rsm::client_members(std::vector<std::string> &r, int i) {
-    std::vector<std::string> m;
+//
+// 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;
     lock ml(rsm_mutex);
     cfg->get_view(vid_commit, m);
     m.push_back(primary);
     r = m;

-    tprintf("rsm::client_members return %s m %s\n", print_members(m).c_str(),
-            primary.c_str());
+    LOG("return " << m << " m " << primary);

     return rsm_client_protocol::OK;
 }
 
     return rsm_client_protocol::OK;
 }
 

@@ -523,13 +486,13 @@ rsm_client_protocol::status rsm::client_members(std::vector<std::string> &r, int
 // otherwise, the lowest number node of the previous view.
 // caller should hold rsm_mutex
 void rsm::set_primary(unsigned vid) {
 // otherwise, the lowest number node of the previous view.
 // caller should hold rsm_mutex
 void rsm::set_primary(unsigned vid) {

-    std::vector<std::string> c, p;
+    vector<string> c, p;

     cfg->get_view(vid, c);
     cfg->get_view(vid - 1, p);
     VERIFY (c.size() > 0);
 
     if (isamember(primary,c)) {
     cfg->get_view(vid, c);
     cfg->get_view(vid - 1, p);
     VERIFY (c.size() > 0);
 
     if (isamember(primary,c)) {

-        tprintf("set_primary: primary stays %s\n", primary.c_str());
+        LOG("set_primary: primary stays " << primary);

         return;
     }
 
         return;
     }
 

@@ -537,7 +500,7 @@ void rsm::set_primary(unsigned vid) {
     for (unsigned i = 0; i < p.size(); i++) {
         if (isamember(p[i], c)) {
             primary = p[i];
     for (unsigned i = 0; i < p.size(); i++) {
         if (isamember(p[i], c)) {
             primary = p[i];

-            tprintf("set_primary: primary is %s\n", primary.c_str());
+            LOG("set_primary: primary is " << primary);

             return;
         }
     }
             return;
         }
     }

@@ -555,25 +518,25 @@ bool rsm::amiprimary() {
 // Simulate partitions
 
 // assumes caller holds rsm_mutex
 // Simulate partitions
 
 // assumes caller holds rsm_mutex

-void rsm::net_repair_wo(bool heal) {
-    std::vector<std::string> m;
+void rsm::net_repair(bool heal, lock &) {
+    vector<string> m;

     cfg->get_view(vid_commit, m);
     for (unsigned i  = 0; i < m.size(); i++) {
         if (m[i] != cfg->myaddr()) {
             handle h(m[i]);
     cfg->get_view(vid_commit, m);
     for (unsigned i  = 0; i < m.size(); i++) {
         if (m[i] != cfg->myaddr()) {
             handle h(m[i]);

-            tprintf("rsm::net_repair_wo: %s %d\n", m[i].c_str(), heal);
+            LOG("member " << m[i] << " " << heal);

             if (h.safebind()) h.safebind()->set_reachable(heal);
         }
     }
     rsmrpc->set_reachable(heal);
 }
 
             if (h.safebind()) h.safebind()->set_reachable(heal);
         }
     }
     rsmrpc->set_reachable(heal);
 }
 

-rsm_test_protocol::status rsm::test_net_repairreq(int &r, int heal) {
+rsm_test_protocol::status rsm::test_net_repairreq(rsm_test_protocol::status &r, int heal) {

     lock ml(rsm_mutex);
     lock ml(rsm_mutex);

-    tprintf("rsm::test_net_repairreq: %d (dopartition %d, partitioned %d)\n",
-            heal, dopartition, partitioned);
+    LOG("heal " << heal << " (dopartition " <<
+            dopartition << ", partitioned " << partitioned << ")");

     if (heal) {
     if (heal) {

-        net_repair_wo(heal);
+        net_repair(heal, ml);

         partitioned = false;
     } else {
         dopartition = true;
         partitioned = false;
     } else {
         dopartition = true;

@@ -587,30 +550,30 @@ rsm_test_protocol::status rsm::test_net_repairreq(int &r, int heal) {
 
 void rsm::breakpoint1() {
     if (break1) {
 
 void rsm::breakpoint1() {
     if (break1) {

-        tprintf("Dying at breakpoint 1 in rsm!\n");
+        LOG("Dying at breakpoint 1 in rsm!");

         exit(1);
     }
 }
 
 void rsm::breakpoint2() {
     if (break2) {
         exit(1);
     }
 }
 
 void rsm::breakpoint2() {
     if (break2) {

-        tprintf("Dying at breakpoint 2 in rsm!\n");
+        LOG("Dying at breakpoint 2 in rsm!");

         exit(1);
     }
 }
 
         exit(1);
     }
 }
 

-void rsm::partition1() {
+void rsm::partition1(lock & rsm_mutex_lock) {

     if (dopartition) {
     if (dopartition) {

-        net_repair_wo(false);
+        net_repair(false, rsm_mutex_lock);

         dopartition = false;
         partitioned = true;
     }
 }
 
         dopartition = false;
         partitioned = true;
     }
 }
 

-rsm_test_protocol::status rsm::breakpointreq(int &r, int b) {
+rsm_test_protocol::status rsm::breakpointreq(rsm_test_protocol::status &r, int b) {

     r = rsm_test_protocol::OK;
     lock ml(rsm_mutex);
     r = rsm_test_protocol::OK;
     lock ml(rsm_mutex);

-    tprintf("rsm::breakpointreq: %d\n", b);
+    LOG("breakpoint " << b);

     if (b == 1) break1 = true;
     else if (b == 2) break2 = true;
     else if (b == 3 || b == 4) cfg->breakpoint(b);
     if (b == 1) break1 = true;
     else if (b == 2) break2 = true;
     else if (b == 3 || b == 4) cfg->breakpoint(b);