Imported Upstream version 2.4.10upstream/2.4.10

29 files changed, 299 insertions, 108 deletions

diff --git a/src/mongo/client/distlock.cpp b/src/mongo/client/distlock.cpp
index 0e38a64f953..a2f56e564bf 100644
--- a/src/mongo/client/distlock.cpp
+++ b/src/mongo/client/distlock.cpp
@@ -131,19 +131,32 @@ namespace mongo {
                         continue;
                     }
 
-                    // remove really old entries from the lockpings collection if they're not holding a lock
-                    // (this may happen if an instance of a process was taken down and no new instance came up to
-                    // replace it for a quite a while)
-                    // if the lock is taken, the take-over mechanism should handle the situation
-                    auto_ptr<DBClientCursor> c = conn->query( LocksType::ConfigNS , BSONObj() );
-                    // TODO:  Would be good to make clear whether query throws or returns empty on errors
-                    uassert( 16060, str::stream() << "cannot query locks collection on config server " << conn.getHost(), c.get() );
+                    // Remove really old entries from the lockpings collection if they're not
+                    // holding a lock. This may happen if an instance of a process was taken down
+                    // and no new instance came up to replace it for a quite a while.
+                    // NOTE this is NOT the same as the standard take-over mechanism, which forces
+                    // the lock entry.
+                    BSONObj fieldsToReturn = BSON( LocksType::state() << 1 <<
+                                                   LocksType::process() << 1 );
+                    auto_ptr<DBClientCursor> activeLocks =
+                        conn->query( LocksType::ConfigNS,
+                                     BSON( LocksType::state() << GT << 0 ) );
+
+                    uassert( 16060,
+                             str::stream() << "cannot query locks collection on config server "
+                                           << conn.getHost(),
+                             activeLocks.get() );
 
                     set<string> pids;
-                    while ( c->more() ) {
-                        BSONObj lock = c->next();
-                        if ( ! lock[LocksType::process()].eoo() ) {
-                            pids.insert( lock[LocksType::process()].valuestrsafe() );
+                    while ( activeLocks->more() ) {
+                        BSONObj lock = activeLocks->nextSafe();
+
+                        if ( !lock[LocksType::process()].eoo() ) {
+                            pids.insert( lock[LocksType::process()].str() );
+                        }
+                        else {
+                            warning() << "found incorrect lock document during lock ping cleanup: "
+                                      << lock.toString() << endl;
                         }
                     }
 
diff --git a/src/mongo/client/distlock.h b/src/mongo/client/distlock.h
index 183826afcce..774d94a0df8 100644
--- a/src/mongo/client/distlock.h
+++ b/src/mongo/client/distlock.h
@@ -57,12 +57,19 @@ namespace mongo {
     };
 
     /**
-     * The distributed lock is a configdb backed way of synchronizing system-wide tasks. A task must be identified by a
-     * unique name across the system (e.g., "balancer"). A lock is taken by writing a document in the configdb's locks
-     * collection with that name.
+     * The distributed lock is a configdb backed way of synchronizing system-wide tasks. A task
+     * must be identified by a unique name across the system (e.g., "balancer"). A lock is taken
+     * by writing a document in the configdb's locks collection with that name.
      *
-     * To be maintained, each taken lock needs to be revalidated ("pinged") within a pre-established amount of time. This
-     * class does this maintenance automatically once a DistributedLock object was constructed.
+     * To be maintained, each taken lock needs to be revalidated ("pinged") within a
+     * pre-established amount of time. This class does this maintenance automatically once a
+     * DistributedLock object was constructed. The ping procedure records the local time to
+     * the ping document, but that time is untrusted and is only used as a point of reference
+     * of whether the ping was refreshed or not. Ultimately, the clock a configdb is the source
+     * of truth when determining whether a ping is still fresh or not. This is achieved by
+     * (1) remembering the ping document time along with config server time when unable to
+     * take a lock, and (2) ensuring all config servers report similar times and have similar
+     * time rates (the difference in times must start and stay small).
      */
     class DistributedLock {
     public:
@@ -147,9 +154,13 @@ namespace mongo {
         const ConnectionString& getRemoteConnection();
 
         /**
-         * Check the skew between a cluster of servers
+         * Checks the skew among a cluster of servers and returns true if the min and max clock
+         * times among the servers are within maxClockSkew.
          */
-        static bool checkSkew( const ConnectionString& cluster, unsigned skewChecks = NUM_LOCK_SKEW_CHECKS, unsigned long long maxClockSkew = MAX_LOCK_CLOCK_SKEW, unsigned long long maxNetSkew = MAX_LOCK_NET_SKEW );
+        static bool checkSkew( const ConnectionString& cluster,
+                               unsigned skewChecks = NUM_LOCK_SKEW_CHECKS,
+                               unsigned long long maxClockSkew = MAX_LOCK_CLOCK_SKEW,
+                               unsigned long long maxNetSkew = MAX_LOCK_NET_SKEW );
 
         /**
          * Get the remote time from a server or cluster
diff --git a/src/mongo/client/gridfs.cpp b/src/mongo/client/gridfs.cpp
index e2d1038d6ee..6d58ec690c2 100644
--- a/src/mongo/client/gridfs.cpp
+++ b/src/mongo/client/gridfs.cpp
@@ -37,7 +37,7 @@
 
 namespace mongo {
 
-    const unsigned DEFAULT_CHUNK_SIZE = 256 * 1024;
+    const unsigned DEFAULT_CHUNK_SIZE = 255 * 1024;
 
     GridFSChunk::GridFSChunk( BSONObj o ) {
         _data = o;
diff --git a/src/mongo/db/db.cpp b/src/mongo/db/db.cpp
index ec337c89885..046101a29bf 100644
--- a/src/mongo/db/db.cpp
+++ b/src/mongo/db/db.cpp
@@ -77,6 +77,7 @@ namespace mongo {
     extern int diagLogging;
     extern unsigned lenForNewNsFiles;
     extern int lockFile;
+    extern bool checkNsFilesOnLoad;
     extern string repairpath;
 
     static void setupSignalHandlers();
@@ -321,6 +322,8 @@ namespace mongo {
         Client::GodScope gs;
         LOG(1) << "enter repairDatabases (to check pdfile version #)" << endl;
 
+        checkNsFilesOnLoad = false; // we are mainly just checking the header - don't scan the whole .ns file for every db here.
+
         Lock::GlobalWrite lk;
         vector< string > dbNames;
         getDatabaseNames( dbNames );
@@ -361,12 +364,12 @@ namespace mongo {
                 }
             }
             else {
-                if (h->versionMinor == PDFILE_VERSION_MINOR_22_AND_OLDER) {
-                    const string systemIndexes = cc().database()->name + ".system.indexes";
-                    shared_ptr<Cursor> cursor(theDataFileMgr.findAll(systemIndexes));
-                    for ( ; cursor && cursor->ok(); cursor->advance()) {
-                        const BSONObj index = cursor->current();
-                        const BSONObj key = index.getObjectField("key");
+                const string systemIndexes = cc().database()->name + ".system.indexes";
+                shared_ptr<Cursor> cursor(theDataFileMgr.findAll(systemIndexes));
+                for ( ; cursor && cursor->ok(); cursor->advance()) {
+                    const BSONObj index = cursor->current();
+                    const BSONObj key = index.getObjectField("key");
+                    if (h->versionMinor == PDFILE_VERSION_MINOR_22_AND_OLDER) {
                         const string plugin = IndexPlugin::findPluginName(key);
                         if (IndexPlugin::existedBefore24(plugin))
                             continue;
@@ -377,6 +380,21 @@ namespace mongo {
                               << "http://dochub.mongodb.org/core/upgrade-2.4"
                               << startupWarningsLog;
                     }
+                    else {
+                        verify(h->versionMinor == PDFILE_VERSION_MINOR_24_AND_NEWER);
+                        try {
+                            IndexSpec(key, index, IndexSpec::RulesFor24);
+                        }
+                        catch (const DBException& e) {
+                            error() << "Encountered an unrecognized index spec: " << index << endl;
+                            error() << "Reason this index spec could not be loaded: \"" << e.what()
+                                    << "\"" << endl;
+                            error() << "The index cannot be used with this version of MongoDB; "
+                                    << "exiting..." << endl;
+                            cc().shutdown();
+                            dbexit(EXIT_UNCAUGHT);
+                        }
+                    }
                 }
                 Database::closeDatabase( dbName.c_str(), dbpath );
             }
@@ -389,6 +407,8 @@ namespace mongo {
             cc().shutdown();
             dbexit( EXIT_CLEAN );
         }
+
+        checkNsFilesOnLoad = true;
     }
 
     void clearTmpFiles() {
@@ -528,39 +548,47 @@ namespace mongo {
     /// warn if readahead > 256KB (gridfs chunk size)
     static void checkReadAhead(const string& dir) {
 #ifdef __linux__
-        const dev_t dev = getPartition(dir);
-
-        // This path handles the case where the filesystem uses the whole device (including LVM)
-        string path = str::stream() <<
-            "/sys/dev/block/" << major(dev) << ':' << minor(dev) << "/queue/read_ahead_kb";
-
-        if (!boost::filesystem::exists(path)){
-            // This path handles the case where the filesystem is on a partition.
-            path = str::stream()
-                << "/sys/dev/block/" << major(dev) << ':' << minor(dev) // this is a symlink
-                << "/.." // parent directory of a partition is for the whole device
-                << "/queue/read_ahead_kb";
-        }
+        try { 
+            const dev_t dev = getPartition(dir);
+
+            // This path handles the case where the filesystem uses the whole device (including LVM)
+            string path = str::stream() <<
+                "/sys/dev/block/" << major(dev) << ':' << minor(dev) << "/queue/read_ahead_kb";
+
+            if (!boost::filesystem::exists(path)){
+                // This path handles the case where the filesystem is on a partition.
+                path = str::stream()
+                    << "/sys/dev/block/" << major(dev) << ':' << minor(dev) // this is a symlink
+                    << "/.." // parent directory of a partition is for the whole device
+                    << "/queue/read_ahead_kb";
+            }
 
-        if (boost::filesystem::exists(path)) {
-            ifstream file (path.c_str());
-            if (file.is_open()) {
-                int kb;
-                file >> kb;
-                if (kb > 256) {
-                    log() << startupWarningsLog;
+            if (boost::filesystem::exists(path)) {
+                ifstream file (path.c_str());
+                if (file.is_open()) {
+                    int kb;
+                    file >> kb;
+                    if (kb > 256) {
+                        log() << startupWarningsLog;
 
-                    log() << "** WARNING: Readahead for " << dir << " is set to " << kb << "KB"
+                        log() << "** WARNING: Readahead for " << dir << " is set to " << kb << "KB"
                             << startupWarningsLog;
 
-                    log() << "**          We suggest setting it to 256KB (512 sectors) or less"
+                        log() << "**          We suggest setting it to 256KB (512 sectors) or less"
                             << startupWarningsLog;
 
-                    log() << "**          http://dochub.mongodb.org/core/readahead"
+                        log() << "**          http://dochub.mongodb.org/core/readahead"
                             << startupWarningsLog;
+                    }
                 }
             }
         }
+        catch (const std::exception& e) {
+            log() << "unable to validate readahead settings due to error: " << e.what()
+                << startupWarningsLog;
+            log() << "for more information, see http://dochub.mongodb.org/core/readahead"
+                << startupWarningsLog;
+        }
 #endif // __linux__
     }
 
@@ -1415,6 +1443,7 @@ namespace mongo {
         sigaddset( &asyncSignals, SIGINT );
         sigaddset( &asyncSignals, SIGTERM );
         sigaddset( &asyncSignals, SIGUSR1 );
+        sigaddset( &asyncSignals, SIGXCPU );
 
         set_terminate( myterminate );
         set_new_handler( my_new_handler );
diff --git a/src/mongo/db/geo/s2index.cpp b/src/mongo/db/geo/s2index.cpp
index d52f0ad2991..412d425f2f8 100644
--- a/src/mongo/db/geo/s2index.cpp
+++ b/src/mongo/db/geo/s2index.cpp
@@ -365,6 +365,11 @@ namespace mongo {
             uassert(16688, "finestIndexedLevel must be <= 30", params.finestIndexedLevel <= 30);
             uassert(16689, "finestIndexedLevel must be >= coarsestIndexedLevel",
                     params.finestIndexedLevel >= params.coarsestIndexedLevel);
+            massert(17289,
+                    str::stream() << "unsupported geo index version { "
+                                  << spec->info["2dsphereIndexVersion"]
+                                  << " }, only support versions: [1]",
+                    configValueWithDefault(spec, "2dsphereIndexVersion", 1) == 1);
 
             // Categorize the fields we're indexing and make sure we have a geo field.
             int geoFields = 0;
@@ -388,6 +393,15 @@ namespace mongo {
             return new S2IndexType(SPHERE_2D_NAME, this, spec, params);
         }
 
+        virtual BSONObj adjustIndexSpec( const BSONObj& spec ) const {
+            BSONElement indexVersionElt = spec["2dsphereIndexVersion"];
+            uassert(17290,
+                    str::stream() << "unsupported geo index version { " << indexVersionElt
+                                  << " }, only support versions: [1]",
+                    indexVersionElt.eoo() || indexVersionElt.numberInt() == 1);
+            return spec;
+        }
+
         int configValueWithDefault(const IndexSpec* spec, const string& name, int def) const {
             BSONElement e = spec->info[name];
             if (e.isNumber()) { return e.numberInt(); }
diff --git a/src/mongo/db/index.cpp b/src/mongo/db/index.cpp
index f7346978313..e4366977081 100644
--- a/src/mongo/db/index.cpp
+++ b/src/mongo/db/index.cpp
@@ -324,6 +324,8 @@ namespace mongo {
         getDur().writingInt(dfh->versionMinor) = PDFILE_VERSION_MINOR_24_AND_NEWER;
     }
 
+    extern BSONObj id_obj; // { _id : 1 }
+
     bool prepareToBuildIndex(const BSONObj& io,
                              bool mayInterrupt,
                              bool god,
@@ -334,9 +336,11 @@ namespace mongo {
 
         // the collection for which we are building an index
         sourceNS = io.getStringField("ns");
+        NamespaceString nss(sourceNS);
         uassert(10096, "invalid ns to index", sourceNS.find( '.' ) != string::npos);
-        massert(10097, str::stream() << "bad table to index name on add index attempt current db: " << cc().database()->name << "  source: " << sourceNS ,
-                cc().database()->name == nsToDatabase(sourceNS));
+        uassert(17072, "cannot create indexes on the system.indexes collection",
+                !nss.isSystemDotIndexes());
+        massert(10097, str::stream() << "bad table to index name on add index attempt current db: " << cc().database()->name << "  source: " << sourceNS , cc().database()->name == nsToDatabase(sourceNS));
 
         // logical name of the index.  todo: get rid of the name, we don't need it!
         const char *name = io.getStringField("name");
@@ -393,10 +397,10 @@ namespace mongo {
            all be treated as the same pattern.
         */
         if ( IndexDetails::isIdIndexPattern(key) ) {
-            if( !god ) {
-                ensureHaveIdIndex( sourceNS.c_str(), mayInterrupt );
-                return false;
-            }
+            //if( !god ) {
+            //ensureHaveIdIndex( sourceNS.c_str(), mayInterrupt );
+            //return false;
+            //}
         }
         else {
             /* is buildIndexes:false set for this replica set member?
@@ -435,7 +439,14 @@ namespace mongo {
             }
             // idea is to put things we use a lot earlier
             b.append("v", v);
-            b.append(o["key"]);
+            if ( IndexDetails::isIdIndexPattern(o["key"].Obj()) ) {
+                b.append("name", "_id_");
+                b.append("key", id_obj);
+            }
+            else {
+                b.append( o["name"] );
+                b.append(o["key"]);
+            }
             if( o["unique"].trueValue() )
                 b.appendBool("unique", true); // normalize to bool true in case was int 1 or something...
             b.append(o["ns"]);
@@ -446,7 +457,7 @@ namespace mongo {
                 while ( i.more() ) {
                     BSONElement e = i.next();
                     string s = e.fieldName();
-                    if( s != "_id" && s != "v" && s != "ns" && s != "unique" && s != "key" )
+                    if( s != "_id" && s != "v" && s != "ns" && s != "unique" && s != "key" && s != "name" )
                         b.append(e);
                 }
             }
diff --git a/src/mongo/db/instance.cpp b/src/mongo/db/instance.cpp
index 926cdeecdb6..b95bcb1bfea 100644
--- a/src/mongo/db/instance.cpp
+++ b/src/mongo/db/instance.cpp
@@ -792,16 +792,19 @@ namespace mongo {
             }
         }
 
-        theDataFileMgr.insertWithObjMod(ns,
-                                        // May be modified in the call to add an _id field.
-                                        js,
-                                        // Only permit interrupting an (index build) insert if the
-                                        // insert comes from a socket client request rather than a
-                                        // parent operation using the client interface.  The parent
-                                        // operation might not support interrupts.
-                                        cc().curop()->parent() == NULL,
-                                        false);
-        logOp("i", ns, js);
+        DiskLoc dl = theDataFileMgr.
+            insertWithObjMod(ns,
+                             // May be modified in the call to add an _id field.
+                             js,
+                             // Only permit interrupting an (index build) insert if the
+                             // insert comes from a socket client request rather than a
+                             // parent operation using the client interface.  The parent
+                             // operation might not support interrupts.
+                             cc().curop()->parent() == NULL,
+                             false);
+        if (!dl.isNull()) {
+            logOp("i", ns, js);
+        }
     }
 
     NOINLINE_DECL void insertMulti(bool keepGoing, const char *ns, vector<BSONObj>& objs, CurOp& op) {
diff --git a/src/mongo/db/namespace_details.cpp b/src/mongo/db/namespace_details.cpp
index 0fc7bdf5eb5..382522bb50a 100644
--- a/src/mongo/db/namespace_details.cpp
+++ b/src/mongo/db/namespace_details.cpp
@@ -122,6 +122,29 @@ namespace mongo {
     }
 #endif
 
+    void NamespaceDetails::onLoad(const Namespace& k) {
+
+        if( k.isExtra() ) {
+            /* overflow storage for indexes - so don't treat as a NamespaceDetails object. */
+            return;
+        }
+
+        if( indexBuildsInProgress ) {
+            verify( Lock::isW() ); // TODO(erh) should this be per db?
+            if( indexBuildsInProgress ) {
+                log() << "indexBuildsInProgress was " << indexBuildsInProgress << " for " << k
+                      << ", indicating an abnormal db shutdown" << endl;
+                getDur().writingInt( indexBuildsInProgress ) = 0;
+            }
+        }
+    }
+
+    static void namespaceOnLoadCallback(const Namespace& k, NamespaceDetails& v) {
+        v.onLoad(k);
+    }
+
+    bool checkNsFilesOnLoad = true;
+
     NOINLINE_DECL void NamespaceIndex::_init() {
         verify( !ht );
 
@@ -174,6 +197,9 @@ namespace mongo {
 
         verify( len <= 0x7fffffff );
         ht = new HashTable<Namespace,NamespaceDetails>(p, (int) len, "namespace index");
+        if( checkNsFilesOnLoad )
+            ht->iterAll(namespaceOnLoadCallback);
+
     }
 
     static void namespaceGetNamespacesCallback( const Namespace& k , NamespaceDetails& v , void * extra ) {
diff --git a/src/mongo/db/namespace_details.h b/src/mongo/db/namespace_details.h
index 0e7e324da67..3f381605250 100644
--- a/src/mongo/db/namespace_details.h
+++ b/src/mongo/db/namespace_details.h
@@ -405,6 +405,7 @@ namespace mongo {
         /** Make all linked Extra objects writeable as well */
         NamespaceDetails *writingWithExtra();
 
+        void onLoad( const Namespace& k );
     private:
         DiskLoc _alloc(const char *ns, int len);
         void maybeComplain( const char *ns, int len ) const;
diff --git a/src/mongo/db/namespacestring.h b/src/mongo/db/namespacestring.h
index d108bbaac72..96d78230b39 100644
--- a/src/mongo/db/namespacestring.h
+++ b/src/mongo/db/namespacestring.h
@@ -46,6 +46,7 @@ namespace mongo {
 
         bool isSystem() const { return strncmp(coll.c_str(), "system.", 7) == 0; }
         bool isCommand() const { return coll == "$cmd"; }
+        bool isSystemDotIndexes() const { return strncmp(coll.c_str(), "system.indexes", 14) == 0; }
 
         /**
          * @return true if the namespace is valid. Special namespaces for internal use are considered as valid.
diff --git a/src/mongo/db/oplog.cpp b/src/mongo/db/oplog.cpp
index b88ad66ac03..33990367d92 100644
--- a/src/mongo/db/oplog.cpp
+++ b/src/mongo/db/oplog.cpp
@@ -95,7 +95,15 @@ namespace mongo {
                      */
             if( theReplSet ) {
                 if( !(theReplSet->lastOpTimeWritten<ts) ) {
-                    log() << "replSet error possible failover clock skew issue? " << theReplSet->lastOpTimeWritten.toString() << ' ' << endl;
+                    log() << "replication oplog stream went back in time. previous timestamp: "
+                        << theReplSet->lastOpTimeWritten << " newest timestamp: " << ts
+                        << ". attempting to sync directly from primary." << endl;
+                    std::string errmsg;
+                    BSONObjBuilder result;
+                    if (!theReplSet->forceSyncFrom(theReplSet->box.getPrimary()->fullName(),
+                                                   errmsg, result)) {
+                        log() << "Can't sync from primary: " << errmsg << endl;
+                    }
                 }
                 theReplSet->lastOpTimeWritten = ts;
                 theReplSet->lastH = h;
@@ -222,8 +230,15 @@ namespace mongo {
                      */
             if( theReplSet ) {
                 if( !(theReplSet->lastOpTimeWritten<ts) ) {
-                    log() << "replSet ERROR possible failover clock skew issue? " << theReplSet->lastOpTimeWritten << ' ' << ts << rsLog;
-                    log() << "replSet " << theReplSet->isPrimary() << rsLog;
+                    log() << "replication oplog stream went back in time. previous timestamp: "
+                        << theReplSet->lastOpTimeWritten << " newest timestamp: " << ts
+                        << ". attempting to sync directly from primary." << endl;
+                    std::string errmsg;
+                    BSONObjBuilder result;
+                    if (!theReplSet->forceSyncFrom(theReplSet->box.getPrimary()->fullName(),
+                                                   errmsg, result)) {
+                        log() << "Can't sync from primary: " << errmsg << endl;
+                    }
                 }
                 theReplSet->lastOpTimeWritten = ts;
                 theReplSet->lastH = hashNew;
diff --git a/src/mongo/db/pdfile.cpp b/src/mongo/db/pdfile.cpp
index 4085dfd2b3f..aa49b156a28 100644
--- a/src/mongo/db/pdfile.cpp
+++ b/src/mongo/db/pdfile.cpp
@@ -1316,8 +1316,10 @@ namespace mongo {
 
     void DataFileMgr::insertAndLog( const char *ns, const BSONObj &o, bool god, bool fromMigrate ) {
         BSONObj tmp = o;
-        insertWithObjMod( ns, tmp, false, god );
-        logOp( "i", ns, tmp, 0, 0, fromMigrate );
+        DiskLoc loc = insertWithObjMod( ns, tmp, false, god );
+        if (!loc.isNull()) {
+            logOp( "i", ns, tmp, 0, 0, fromMigrate );
+        }
     }
 
     /** @param o the object to insert. can be modified to add _id and thus be an in/out param
@@ -1475,16 +1477,20 @@ namespace mongo {
         }
 
         try {
-            IndexDetails& idx = tableToIndex->getNextIndexDetails(tabletoidxns.c_str());
-            // It's important that this is outside the inner try/catch so that we never try to call
-            // kill_idx on a half-formed disk loc (if this asserts).
-            getDur().writingDiskLoc(idx.info) = loc;
+            {
+                IndexDetails& idx = tableToIndex->getNextIndexDetails(tabletoidxns.c_str());
+                // It's important that this is outside the inner try/catch so that we never try to call
+                // kill_idx on a half-formed disk loc (if this asserts).
+                getDur().writingDiskLoc(idx.info) = loc;
+            }
 
             try {
+                IndexDetails& idx = tableToIndex->getNextIndexDetails(tabletoidxns.c_str());
                 getDur().writingInt(tableToIndex->indexBuildsInProgress) += 1;
                 buildAnIndex(tabletoidxns, tableToIndex, idx, background, mayInterrupt);
             }
             catch (DBException& e) {
+                log() << "error building index: " << e << endl;
                 // save our error msg string as an exception or dropIndexes will overwrite our message
                 LastError *le = lastError.get();
                 int savecode = 0;
@@ -1499,7 +1505,10 @@ namespace mongo {
                 }
 
                 // Recalculate the index # so we can remove it from the list in the next catch
+
                 idxNo = IndexBuildsInProgress::get(tabletoidxns.c_str(), idxName);
+                IndexDetails& idx = tableToIndex->idx(idxNo);
+
                 // roll back this index
                 idx.kill_idx();
 
@@ -1519,6 +1528,7 @@ namespace mongo {
                       << tableToIndex->nIndexes << endl;
                 // We cannot use idx here, as it may point to a different index entry if it was
                 // flipped during building
+
                 IndexDetails temp = tableToIndex->idx(idxNo);
                 *getDur().writing(&tableToIndex->idx(idxNo)) =
                     tableToIndex->idx(tableToIndex->nIndexes);
@@ -1531,6 +1541,7 @@ namespace mongo {
                 tableToIndex->setIndexIsMultikey(tabletoidxns.c_str(), tableToIndex->nIndexes,
                                                  tempMultikey);
 
+
                 idxNo = tableToIndex->nIndexes;
             }
 
@@ -1542,10 +1553,10 @@ namespace mongo {
             tableToIndex->addIndex(tabletoidxns.c_str());
             getDur().writingInt(tableToIndex->indexBuildsInProgress) -= 1;
 
-            IndexType* indexType = idx.getSpec().getType();
+            IndexType* indexType = tableToIndex->idx(idxNo).getSpec().getType();
             const IndexPlugin *plugin = indexType ? indexType->getPlugin() : NULL;
             if (plugin) {
-                plugin->postBuildHook( idx.getSpec() );
+                plugin->postBuildHook( tableToIndex->idx(idxNo).getSpec() );
             }
 
         }
@@ -1589,7 +1600,8 @@ namespace mongo {
         Lock::assertWriteLocked(ns);
         NamespaceDetails* nsd = nsdetails(ns);
 
-        for (int i=offset; i<nsd->getTotalIndexCount(); i++) {
+        // offset is 0-based, so we subtract one from the index count
+        for (int i = offset; i < (nsd->getTotalIndexCount() - 1); i++) {
             if (i < NamespaceDetails::NIndexesMax-1) {
                 *getDur().writing(&nsd->idx(i)) = nsd->idx(i+1);
                 nsd->setIndexIsMultikey(ns, i, nsd->isMultikey(i+1));
diff --git a/src/mongo/db/pipeline/value.cpp b/src/mongo/db/pipeline/value.cpp
index 207dc4e37b8..6045207c764 100644
--- a/src/mongo/db/pipeline/value.cpp
+++ b/src/mongo/db/pipeline/value.cpp
@@ -513,19 +513,15 @@ namespace mongo {
     }
 
     string Value::coerceToString() const {
-        stringstream ss;
         switch(getType()) {
         case NumberDouble:
-            ss << _storage.doubleValue;
-            return ss.str();
+            return str::stream() << _storage.doubleValue;
 
         case NumberInt:
-            ss << _storage.intValue;
-            return ss.str();
+            return str::stream() << _storage.intValue;
 
         case NumberLong:
-            ss << _storage.longValue;
-            return ss.str();
+            return str::stream() << _storage.longValue;
 
         case Code:
         case Symbol:
@@ -533,8 +529,7 @@ namespace mongo {
             return getStringData().toString();
 
         case Timestamp:
-            ss << getTimestamp().toStringPretty();
-            return ss.str();
+            return getTimestamp().toStringPretty();
 
         case Date:
             return tmToISODateString(coerceToTm());
diff --git a/src/mongo/db/repl/consensus.cpp b/src/mongo/db/repl/consensus.cpp
index dcb31408c11..a2459477432 100644
--- a/src/mongo/db/repl/consensus.cpp
+++ b/src/mongo/db/repl/consensus.cpp
@@ -253,7 +253,6 @@ namespace mongo {
             try {
                 vote = yea(whoid);
                 dassert( hopeful->id() == whoid );
-                rs.relinquish();
                 log() << "replSet info voting yea for " <<  hopeful->fullName() << " (" << whoid << ')' << rsLog;
             }
             catch(VoteException&) {
diff --git a/src/mongo/db/repl/manager.cpp b/src/mongo/db/repl/manager.cpp
index cbc3a3e7baa..4e9708a1d19 100644
--- a/src/mongo/db/repl/manager.cpp
+++ b/src/mongo/db/repl/manager.cpp
@@ -80,11 +80,7 @@ namespace mongo {
         }
 
         if (rs->box.getState().primary()) {
-            // make sure exactly one primary steps down
-            if (rs->selfId() < m->id()) {
-                return;
-            }
-
+            log() << "stepping down; another primary seen in replicaset";
             rs->relinquish();
         }
 
diff --git a/src/mongo/db/repl/rs_sync.cpp b/src/mongo/db/repl/rs_sync.cpp
index e6e31943a9f..f904d86c5ad 100644
--- a/src/mongo/db/repl/rs_sync.cpp
+++ b/src/mongo/db/repl/rs_sync.cpp
@@ -585,13 +585,6 @@ namespace replset {
         bool golive = false;
 
         lock rsLock( this );
-        Lock::GlobalWrite writeLock;
-
-        // make sure we're not primary, secondary, rollback, or fatal already
-        if (box.getState().primary() || box.getState().secondary() ||
-            box.getState().fatal()) {
-            return false;
-        }
 
         if (_maintenanceMode > 0) {
             // we're not actually going live
@@ -603,6 +596,14 @@ namespace replset {
             return false;
         }
 
+        Lock::GlobalWrite writeLock;
+
+        // make sure we're not primary, secondary, rollback, or fatal already
+        if (box.getState().primary() || box.getState().secondary() ||
+            box.getState().fatal()) {
+            return false;
+        }
+
         minvalid = getMinValid();
         if( minvalid <= lastOpTimeWritten ) {
             golive=true;
diff --git a/src/mongo/dbtests/d_chunk_manager_tests.cpp b/src/mongo/dbtests/d_chunk_manager_tests.cpp
index e50961d3ee7..be7639f5794 100644
--- a/src/mongo/dbtests/d_chunk_manager_tests.cpp
+++ b/src/mongo/dbtests/d_chunk_manager_tests.cpp
@@ -362,6 +362,20 @@ namespace {
             ASSERT( cloned->belongsToMe( split1 ) );
             ASSERT( cloned->belongsToMe( split2 ) );
             ASSERT( ! cloned->belongsToMe( max ) );
+
+            ASSERT_FALSE( cloned->getNextChunk( BSON( "a" << MinKey << "b" << 0 ), &min, &max ));
+            ASSERT_EQUALS( BSON( "a" << 10 << "b" << 0 ), min );
+            ASSERT_EQUALS( BSON( "a" << 15 << "b" << 0 ), max );
+
+            ASSERT_FALSE( cloned->getNextChunk( BSON( "a" << 10 << "b" << 0 ), &min, &max ));
+            ASSERT_EQUALS( BSON( "a" << 15 << "b" << 0 ), min );
+            ASSERT_EQUALS( BSON( "a" << 18 << "b" << 0 ), max );
+
+            ASSERT_FALSE( cloned->getNextChunk( BSON( "a" << 15 << "b" << 0 ), &min, &max ));
+            ASSERT_EQUALS( BSON( "a" << 18 << "b" << 0 ), min );
+            ASSERT_EQUALS( BSON( "a" << 20 << "b" << 0 ), max );
+
+            ASSERT( cloned->getNextChunk( BSON( "a" << 18 << "b" << 0 ), &min, &max ));
         }
     };
 
diff --git a/src/mongo/s/balance.cpp b/src/mongo/s/balance.cpp
index 2cf9d06d131..7e12e1c47b8 100644
--- a/src/mongo/s/balance.cpp
+++ b/src/mongo/s/balance.cpp
@@ -467,6 +467,7 @@ namespace mongo {
                         conn.done();
                         warning() << "Skipping balancing round because data inconsistency"
                                   << " was detected amongst the config servers." << endl;
+                        sleepsecs( sleepTime );
                         continue;
                     }
 
diff --git a/src/mongo/s/commands_admin.cpp b/src/mongo/s/commands_admin.cpp
index 91577d33b97..d7061f6fa86 100644
--- a/src/mongo/s/commands_admin.cpp
+++ b/src/mongo/s/commands_admin.cpp
@@ -663,7 +663,7 @@ namespace mongo {
                 result << "collectionsharded" << ns;
 
                 // only initially move chunks when using a hashed shard key
-                if (isHashedShardKey) {
+                if (isHashedShardKey && isEmpty) {
 
                     // Reload the new config info.  If we created more than one initial chunk, then
                     // we need to move them around to balance.
diff --git a/src/mongo/s/config.cpp b/src/mongo/s/config.cpp
index b8698a10687..9e042938607 100644
--- a/src/mongo/s/config.cpp
+++ b/src/mongo/s/config.cpp
@@ -34,6 +34,8 @@
 #include "mongo/s/type_chunk.h"
 #include "mongo/s/type_collection.h"
 #include "mongo/s/type_database.h"
+#include "mongo/s/type_locks.h"
+#include "mongo/s/type_lockpings.h"
 #include "mongo/s/type_settings.h"
 #include "mongo/s/type_shard.h"
 #include "mongo/util/net/message.h"
@@ -985,6 +987,16 @@ namespace mongo {
 
             conn->get()->ensureIndex(ShardType::ConfigNS, BSON(ShardType::host() << 1), true);
 
+            conn->get()->ensureIndex(LocksType::ConfigNS, 
+                                     BSON( LocksType::lockID() << 1 ), true);
+
+            conn->get()->ensureIndex(LockpingsType::ConfigNS, 
+                                     BSON( LockpingsType::ping() << 1 ), false);
+
+            conn->get()->ensureIndex(LocksType::ConfigNS, 
+                                     BSON( LocksType::state() << 1 << 
+                                           LocksType::process() << 1 ), false);
+
             conn->done();
         }
         catch ( DBException& e ) {
diff --git a/src/mongo/s/d_chunk_manager.cpp b/src/mongo/s/d_chunk_manager.cpp
index 4eb360702bc..98a6d90d341 100644
--- a/src/mongo/s/d_chunk_manager.cpp
+++ b/src/mongo/s/d_chunk_manager.cpp
@@ -426,7 +426,7 @@ namespace mongo {
         BSONObj startKey = min;
         for ( vector<BSONObj>::const_iterator it = splitKeys.begin() ; it != splitKeys.end() ; ++it ) {
             BSONObj split = *it;
-            p->_chunksMap[min] = split.getOwned();
+            p->_chunksMap[startKey] = split.getOwned();
             p->_chunksMap.insert( make_pair( split.getOwned() , max.getOwned() ) );
             p->_version.incMinor();
             startKey = split;
diff --git a/src/mongo/s/server.cpp b/src/mongo/s/server.cpp
index e6aa728d7a3..5bcfe291519 100644
--- a/src/mongo/s/server.cpp
+++ b/src/mongo/s/server.cpp
@@ -197,6 +197,9 @@ namespace mongo {
 
         signal(SIGTERM, sighandler);
         signal(SIGINT, sighandler);
+#if defined(SIGXCPU)
+        signal(SIGXCPU, sighandler);
+#endif
 
 #if defined(SIGQUIT)
         signal( SIGQUIT , printStackAndExit );
diff --git a/src/mongo/shell/replsettest.js b/src/mongo/shell/replsettest.js
index a9c8c36b050..6b84da39a06 100644
--- a/src/mongo/shell/replsettest.js
+++ b/src/mongo/shell/replsettest.js
@@ -489,12 +489,13 @@ ReplSetTest.prototype.initiate = function( cfg , initCmd , timeout ) {
     }
 }
 
-ReplSetTest.prototype.reInitiate = function() {
+ReplSetTest.prototype.reInitiate = function(timeout) {
     var master  = this.nodes[0];
     var c = master.getDB("local")['system.replset'].findOne();
     var config  = this.getReplSetConfig();
+    var timeout = timeout || 60000;
     config.version = c.version + 1;
-    this.initiate( config , 'replSetReconfig' );
+    this.initiate( config , 'replSetReconfig', timeout );
 }
 
 ReplSetTest.prototype.getLastOpTimeWritten = function() {
diff --git a/src/mongo/shell/shell_utils_extended.cpp b/src/mongo/shell/shell_utils_extended.cpp
index 1e68882a3bc..2808bab7688 100644
--- a/src/mongo/shell/shell_utils_extended.cpp
+++ b/src/mongo/shell/shell_utils_extended.cpp
@@ -26,6 +26,7 @@
 #include "mongo/util/file.h"
 #include "mongo/util/md5.hpp"
 #include "mongo/util/net/sock.h"
+#include "mongo/util/scopeguard.h"
 #include "mongo/util/text.h"
 
 namespace mongo {
@@ -144,6 +145,7 @@ namespace mongo {
             stringstream ss;
             FILE* f = fopen(e.valuestrsafe(), "rb");
             uassert(CANT_OPEN_FILE, "couldn't open file", f );
+            ON_BLOCK_EXIT(fclose, f);
 
             md5digest d;
             md5_state_t st;
diff --git a/src/mongo/util/file_allocator.cpp b/src/mongo/util/file_allocator.cpp
index e59bce6934b..7876744f7d5 100644
--- a/src/mongo/util/file_allocator.cpp
+++ b/src/mongo/util/file_allocator.cpp
@@ -39,6 +39,7 @@
 #include "mongo/platform/posix_fadvise.h"
 #include "mongo/util/mongoutils/str.h"
 #include "mongo/util/paths.h"
+#include "mongo/util/processinfo.h"
 #include "mongo/util/time_support.h"
 #include "mongo/util/timer.h"
 
@@ -186,6 +187,15 @@ namespace mongo {
                      size - 1 == lseek(fd, size - 1, SEEK_SET) );
             uassert( 10442 ,  str::stream() << "Unable to allocate new file of size " << size << ' ' << errnoWithDescription(),
                      1 == write(fd, "", 1) );
+
+            // File expansion is completed here. Do not do the zeroing out on OS-es where there
+            // is no risk of triggering allocation-related bugs such as
+            // http://support.microsoft.com/kb/2731284.
+            //
+            if (!ProcessInfo::isDataFileZeroingNeeded()) {
+                return;
+            }
+
             lseek(fd, 0, SEEK_SET);
 
             const long z = 256 * 1024;
diff --git a/src/mongo/util/processinfo.h b/src/mongo/util/processinfo.h
index 8e9044d39e4..03fcdd85993 100644
--- a/src/mongo/util/processinfo.h
+++ b/src/mongo/util/processinfo.h
@@ -92,6 +92,11 @@ namespace mongo {
         bool hasNumaEnabled() const { return sysInfo().hasNuma; }
 
         /**
+         * Determine if file zeroing is necessary for newly allocated data files.
+         */
+        static bool isDataFileZeroingNeeded() { return systemInfo->fileZeroNeeded;  }
+
+        /**
          * Get extra system stats
          */
         void appendSystemDetails( BSONObjBuilder& details ) const {
@@ -145,12 +150,19 @@ namespace mongo {
             string cpuArch;
             bool hasNuma;
             BSONObj _extraStats;
+
+            // This is an OS specific value, which determines whether files should be zero-filled
+            // at allocation time in order to avoid Microsoft KB 2731284.
+            //
+            bool fileZeroNeeded;
+
             SystemInfo() :
                     addrSize( 0 ),
                     memSize( 0 ),
                     numCores( 0 ),
                     pageSize( 0 ),
-                    hasNuma( false ) { 
+                    hasNuma( false ),
+                    fileZeroNeeded (false) { 
                 // populate SystemInfo during construction
                 collectSystemInfo();
             }
diff --git a/src/mongo/util/processinfo_win32.cpp b/src/mongo/util/processinfo_win32.cpp
index 3f42eb12698..d6f88e88a59 100644
--- a/src/mongo/util/processinfo_win32.cpp
+++ b/src/mongo/util/processinfo_win32.cpp
@@ -100,7 +100,7 @@ namespace mongo {
     void ProcessInfo::SystemInfo::collectSystemInfo() {
         BSONObjBuilder bExtra;
         stringstream verstr;
-        OSVERSIONINFOEX osvi;   // os version 
+        OSVERSIONINFOEX osvi;   // os version
         MEMORYSTATUSEX mse;     // memory stats
         SYSTEM_INFO ntsysinfo;  //system stats
 
@@ -142,6 +142,15 @@ namespace mongo {
                             osName += "Windows 7";
                         else
                             osName += "Windows Server 2008 R2";
+
+                        // Windows 6.1 is either Windows 7 or Windows 2008 R2. There is no SP2 for
+                        // either of these two operating systems, but the check will hold if one
+                        // were released. This code assumes that SP2 will include fix for 
+                        // http://support.microsoft.com/kb/2731284.
+                        //
+                        if ((osvi.wServicePackMajor >= 0) && (osvi.wServicePackMajor < 2)) {
+                            fileZeroNeeded = true;
+                        }
                         break;
                     case 0:
                         if ( osvi.wProductType == VER_NT_WORKSTATION )
diff --git a/src/mongo/util/version.cpp b/src/mongo/util/version.cpp
index a5622e4dbbb..e27c3843a4a 100644
--- a/src/mongo/util/version.cpp
+++ b/src/mongo/util/version.cpp
@@ -47,7 +47,7 @@ namespace mongo {
      *      1.2.3-rc4-pre-
      * If you really need to do something else you'll need to fix _versionArray()
      */
-    const char versionString[] = "2.4.9";
+    const char versionString[] = "2.4.10";
 
     // See unit test for example outputs
     BSONArray toVersionArray(const char* version){
diff --git a/src/third_party/v8/src/spaces-inl.h b/src/third_party/v8/src/spaces-inl.h
index ed78fc7a15f..d84019084cc 100644
--- a/src/third_party/v8/src/spaces-inl.h
+++ b/src/third_party/v8/src/spaces-inl.h
@@ -164,7 +164,7 @@ Page* Page::Initialize(Heap* heap,
                        Executability executable,
                        PagedSpace* owner) {
   Page* page = reinterpret_cast<Page*>(chunk);
-  ASSERT(chunk->size() <= static_cast<size_t>(kPageSize));
+  ASSERT(page->area_size() <= kNonCodeObjectAreaSize);
   ASSERT(chunk->owner() == owner);
   owner->IncreaseCapacity(page->area_size());
   owner->Free(page->area_start(), page->area_size());