1: #!/usr/bin/perl
2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
5: # $Id: loncnew,v 1.83 2007/04/11 00:10:45 albertel Exp $
6: #
7: # Copyright Michigan State University Board of Trustees
8: #
9: # This file is part of the LearningOnline Network with CAPA (LON-CAPA).
10: ## LON-CAPA is free software; you can redistribute it and/or modify
11: # it under the terms of the GNU General Public License as published by
12: # the Free Software Foundation; either version 2 of the License, or
13: # (at your option) any later version.
14: #
15: # LON-CAPA is distributed in the hope that it will be useful,
16: # but WITHOUT ANY WARRANTY; without even the implied warranty of
17: # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18: # GNU General Public License for more details.
19: #
20: # You should have received a copy of the GNU General Public License
21: # along with LON-CAPA; if not, write to the Free Software
22: # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23: #
24: # /home/httpd/html/adm/gpl.txt
25: #
26: # http://www.lon-capa.org/
27: #
28: #
29: # new lonc handles n request out bver m connections to londs.
30: # This module is based on the Event class.
31: # Development iterations:
32: # - Setup basic event loop. (done)
33: # - Add timer dispatch. (done)
34: # - Add ability to accept lonc UNIX domain sockets. (done)
35: # - Add ability to create/negotiate lond connections (done).
36: # - Add general logic for dispatching requests and timeouts. (done).
37: # - Add support for the lonc/lond requests. (done).
38: # - Add logging/status monitoring. (done)
39: # - Add Signal handling - HUP restarts. USR1 status report. (done)
40: # - Add Configuration file I/O (done).
41: # - Add management/status request interface. (done)
42: # - Add deferred request capability. (done)
43: # - Detect transmission timeouts. (done)
44: #
45:
46: use strict;
47: use lib "/home/httpd/lib/perl/";
48: use Event qw(:DEFAULT );
49: use POSIX qw(:signal_h);
50: use POSIX;
51: use IO::Socket;
52: use IO::Socket::INET;
53: use IO::Socket::UNIX;
54: use IO::File;
55: use IO::Handle;
56: use Socket;
57: use Crypt::IDEA;
58: use LONCAPA::Queue;
59: use LONCAPA::Stack;
60: use LONCAPA::LondConnection;
61: use LONCAPA::LondTransaction;
62: use LONCAPA::Configuration;
63: use Fcntl qw(:flock);
64:
65:
66: # Read the httpd configuration file to get perl variables
67: # normally set in apache modules:
68:
69: my $perlvarref = LONCAPA::Configuration::read_conf('loncapa.conf');
70: my %perlvar = %{$perlvarref};
71:
72: #
73: # parent and shared variables.
74:
75: my %ChildPid; # by pid -> host.
76: my %ChildHost; # by host.
77: my %listening_to; # Socket->host table for who the parent
78: # is listening to.
79: my %parent_dispatchers; # host-> listener watcher events.
80:
81: my %parent_handlers; # Parent signal handlers...
82:
83: my $MaxConnectionCount = 10; # Will get from config later.
84: my $ClientConnection = 0; # Uniquifier for client events.
85:
86: my $DebugLevel = 0;
87: my $NextDebugLevel= 2; # So Sigint can toggle this.
88: my $IdleTimeout= 600; # Wait 10 minutes before pruning connections.
89:
90: my $LogTransactions = 0; # When True, all transactions/replies get logged.
91: my $executable = $0; # Get the full path to me.
92:
93: #
94: # The variables below are only used by the child processes.
95: #
96: my $RemoteHost; # Name of host child is talking to.
97: my $RemoteHostId; # default lonid of host child is talking to.
98: my @all_host_ids;
99: my $UnixSocketDir= $perlvar{'lonSockDir'};
100: my $IdleConnections = Stack->new(); # Set of idle connections
101: my %ActiveConnections; # Connections to the remote lond.
102: my %ActiveTransactions; # LondTransactions in flight.
103: my %ActiveClients; # Serial numbers of active clients by socket.
104: my $WorkQueue = Queue->new(); # Queue of pending transactions.
105: my $ConnectionCount = 0;
106: my $IdleSeconds = 0; # Number of seconds idle.
107: my $Status = ""; # Current status string.
108: my $RecentLogEntry = "";
109: my $ConnectionRetries=5; # Number of connection retries allowed.
110: my $ConnectionRetriesLeft=5; # Number of connection retries remaining.
111: my $LondVersion = "unknown"; # Version of lond we talk with.
112: my $KeyMode = ""; # e.g. ssl, local, insecure from last connect.
113: my $LondConnecting = 0; # True when a connection is being built.
114:
115:
116:
117: my $I_am_child = 0; # True if this is the child process.
118:
119: #
120: # The hash below gives the HTML format for log messages
121: # given a severity.
122: #
123: my %LogFormats;
124:
125: $LogFormats{"CRITICAL"} = "<font color='red'>CRITICAL: %s</font>";
126: $LogFormats{"SUCCESS"} = "<font color='green'>SUCCESS: %s</font>";
127: $LogFormats{"INFO"} = "<font color='yellow'>INFO: %s</font>";
128: $LogFormats{"WARNING"} = "<font color='blue'>WARNING: %s</font>";
129: $LogFormats{"DEFAULT"} = " %s ";
130:
131:
132: # UpdateStatus;
133: # Update the idle status display to show how many connections
134: # are left, retries and other stuff.
135: #
136: sub UpdateStatus {
137: if ($ConnectionRetriesLeft > 0) {
138: ShowStatus(GetServerHost()." Connection count: ".$ConnectionCount
139: ." Retries remaining: ".$ConnectionRetriesLeft
140: ." ($KeyMode)");
141: } else {
142: ShowStatus(GetServerHost()." >> DEAD <<");
143: }
144: }
145:
146:
147: =pod
148:
149: =head2 LogPerm
150:
151: Makes an entry into the permanent log file.
152:
153: =cut
154:
155: sub LogPerm {
156: my $message=shift;
157: my $execdir=$perlvar{'lonDaemons'};
158: my $now=time;
159: my $local=localtime($now);
160: my $fh=IO::File->new(">>$execdir/logs/lonnet.perm.log");
161: print $fh "$now:$message:$local\n";
162: }
163:
164: =pod
165:
166: =head2 Log
167:
168: Logs a message to the log file.
169: Parameters:
170:
171: =item severity
172:
173: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
174: format string used to format the message. if the severity is
175: not a defined severity the Default format string is used.
176:
177: =item message
178:
179: The base message. In addtion to the format string, the message
180: will be appended to a string containing the name of our remote
181: host and the time will be formatted into the message.
182:
183: =cut
184:
185: sub Log {
186:
187: my ($severity, $message) = @_;
188:
189: if(!$LogFormats{$severity}) {
190: $severity = "DEFAULT";
191: }
192:
193: my $format = $LogFormats{$severity};
194:
195: # Put the window dressing in in front of the message format:
196:
197: my $now = time;
198: my $local = localtime($now);
199: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
200: $finalformat = $finalformat.$format."\n";
201:
202: # open the file and put the result.
203:
204: my $execdir = $perlvar{'lonDaemons'};
205: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
206: my $msg = sprintf($finalformat, $message);
207: $RecentLogEntry = $msg;
208: print $fh $msg;
209:
210:
211: }
212:
213:
214: =pod
215:
216: =head2 GetPeerName
217:
218: Returns the name of the host that a socket object is connected to.
219:
220: =cut
221:
222: sub GetPeername {
223:
224:
225: my ($connection, $AdrFamily) = @_;
226:
227: my $peer = $connection->peername();
228: my $peerport;
229: my $peerip;
230: if($AdrFamily == AF_INET) {
231: ($peerport, $peerip) = sockaddr_in($peer);
232: my $peername = gethostbyaddr($peerip, $AdrFamily);
233: return $peername;
234: } elsif ($AdrFamily == AF_UNIX) {
235: my $peerfile;
236: ($peerfile) = sockaddr_un($peer);
237: return $peerfile;
238: }
239: }
240: =pod
241:
242: =head2 Debug
243:
244: Invoked to issue a debug message.
245:
246: =cut
247:
248: sub Debug {
249:
250: my ($level, $message) = @_;
251:
252: if ($level <= $DebugLevel) {
253: Log("INFO", "-Debug- $message host = $RemoteHost");
254: }
255: }
256:
257: sub SocketDump {
258:
259: my ($level, $socket) = @_;
260:
261: if($level <= $DebugLevel) {
262: $socket->Dump(-1); # Ensure it will get dumped.
263: }
264: }
265:
266: =pod
267:
268: =head2 ShowStatus
269:
270: Place some text as our pid status.
271: and as what we return in a SIGUSR1
272:
273: =cut
274:
275: sub ShowStatus {
276: my $state = shift;
277: my $now = time;
278: my $local = localtime($now);
279: $Status = $local.": ".$state;
280: $0='lonc: '.$state.' '.$local;
281: }
282:
283: =pod
284:
285: =head2 SocketTimeout
286:
287: Called when an action on the socket times out. The socket is
288: destroyed and any active transaction is failed.
289:
290:
291: =cut
292:
293: sub SocketTimeout {
294: my $Socket = shift;
295: Log("WARNING", "A socket timeout was detected");
296: Debug(5, " SocketTimeout called: ");
297: $Socket->Dump(0);
298: if(exists($ActiveTransactions{$Socket})) {
299: FailTransaction($ActiveTransactions{$Socket});
300: }
301: KillSocket($Socket); # A transaction timeout also counts as
302: # a connection failure:
303: $ConnectionRetriesLeft--;
304: if($ConnectionRetriesLeft <= 0) {
305: Log("CRITICAL", "Host marked DEAD: ".GetServerHost());
306: $LondConnecting = 0;
307: }
308:
309: }
310:
311: #
312: # This function should be called by the child in all cases where it must
313: # exit. The child process must create a lock file for the AF_UNIX socket
314: # in order to prevent connection requests from lonnet in the time between
315: # process exit and the parent picking up the listen again.
316: #
317: # Parameters:
318: # exit_code - Exit status value, however see the next parameter.
319: # message - If this optional parameter is supplied, the exit
320: # is via a die with this message.
321: #
322: sub child_exit {
323: my ($exit_code, $message) = @_;
324:
325: # Regardless of how we exit, we may need to do the lock thing:
326:
327: #
328: # Create a lock file since there will be a time window
329: # between our exit and the parent's picking up the listen
330: # during which no listens will be done on the
331: # lonnet client socket.
332: #
333: my $lock_file = &GetLoncSocketPath().".lock";
334: open(LOCK,">$lock_file");
335: print LOCK "Contents not important";
336: close(LOCK);
337: unlink(&GetLoncSocketPath());
338:
339: if ($message) {
340: die($message);
341: } else {
342: exit($exit_code);
343: }
344: }
345: #----------------------------- Timer management ------------------------
346:
347: =pod
348:
349: =head2 Tick
350:
351: Invoked each timer tick.
352:
353: =cut
354:
355:
356: sub Tick {
357: my ($Event) = @_;
358: my $clock_watcher = $Event->w;
359:
360: my $client;
361: UpdateStatus();
362:
363: # Is it time to prune connection count:
364:
365:
366: if($IdleConnections->Count() &&
367: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
368: $IdleSeconds++;
369: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
370: my $Socket = $IdleConnections->pop();
371: KillSocket($Socket);
372: $IdleSeconds = 0; # Otherwise all connections get trimmed to fast.
373: UpdateStatus();
374: if(($ConnectionCount == 0)) {
375: &child_exit(0);
376:
377: }
378: }
379: } else {
380: $IdleSeconds = 0; # Reset idle count if not idle.
381: }
382: #
383: # For each inflight transaction, tick down its timeout counter.
384: #
385:
386: foreach my $item (keys %ActiveConnections) {
387: my $State = $ActiveConnections{$item}->data->GetState();
388: if ($State ne 'Idle') {
389: Debug(5,"Ticking Socket $State $item");
390: $ActiveConnections{$item}->data->Tick();
391: }
392: }
393: # Do we have work in the queue, but no connections to service them?
394: # If so, try to make some new connections to get things going again.
395: #
396: # Note this code is dead now...
397: #
398: my $Requests = $WorkQueue->Count();
399: if (($ConnectionCount == 0) && ($Requests > 0) && (!$LondConnecting)) {
400: if ($ConnectionRetriesLeft > 0) {
401: Debug(5,"Work but no connections, Make a new one");
402: my $success;
403: $success = &MakeLondConnection;
404: if($success == 0) { # All connections failed:
405: Debug(5,"Work in queue failed to make any connectiouns\n");
406: EmptyQueue(); # Fail pending transactions with con_lost.
407: CloseAllLondConnections(); # Should all be closed but....
408: }
409: } else {
410: $LondConnecting = 0;
411: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
412: Debug(5,"Work in queue, but gave up on connections..flushing\n");
413: EmptyQueue(); # Connections can't be established.
414: CloseAllLondConnections(); # Should all already be closed but...
415: }
416:
417: }
418: if ($ConnectionCount == 0) {
419: $KeyMode = "";
420: $clock_watcher->cancel();
421: }
422: &UpdateStatus();
423: }
424:
425: =pod
426:
427: =head2 SetupTimer
428:
429: Sets up a 1 per sec recurring timer event. The event handler is used to:
430:
431: =item
432:
433: Trigger timeouts on communications along active sockets.
434:
435: =item
436:
437: Trigger disconnections of idle sockets.
438:
439: =cut
440:
441: sub SetupTimer {
442: Debug(6, "SetupTimer");
443: Event->timer(interval => 1, cb => \&Tick );
444: }
445:
446: =pod
447:
448: =head2 ServerToIdle
449:
450: This function is called when a connection to the server is
451: ready for more work.
452:
453: If there is work in the Work queue the top element is dequeued
454: and the connection will start to work on it. If the work queue is
455: empty, the connection is pushed on the idle connection stack where
456: it will either get another work unit, or alternatively, if it sits there
457: long enough, it will be shut down and released.
458:
459: =cut
460:
461: sub ServerToIdle {
462: my $Socket = shift; # Get the socket.
463: $KeyMode = $Socket->{AuthenticationMode};
464: delete($ActiveTransactions{$Socket}); # Server has no transaction
465:
466: &Debug(5, "Server to idle");
467:
468: # If there's work to do, start the transaction:
469:
470: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
471: if ($reqdata ne undef) {
472: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
473: &StartRequest($Socket, $reqdata);
474:
475: } else {
476:
477: # There's no work waiting, so push the server to idle list.
478: &Debug(5, "No new work requests, server connection going idle");
479: $IdleConnections->push($Socket);
480: }
481: }
482:
483: =pod
484:
485: =head2 ClientWritable
486:
487: Event callback for when a client socket is writable.
488:
489: This callback is established when a transaction reponse is
490: avaiable from lond. The response is forwarded to the unix socket
491: as it becomes writable in this sub.
492:
493: Parameters:
494:
495: =item Event
496:
497: The event that has been triggered. Event->w->data is
498: the data and Event->w->fd is the socket to write.
499:
500: =cut
501:
502: sub ClientWritable {
503: my $Event = shift;
504: my $Watcher = $Event->w;
505: my $Data = $Watcher->data;
506: my $Socket = $Watcher->fd;
507:
508: # Try to send the data:
509:
510: &Debug(6, "ClientWritable writing".$Data);
511: &Debug(9, "Socket is: ".$Socket);
512:
513: if($Socket->connected) {
514: my $result = $Socket->send($Data, 0);
515:
516: # $result undefined: the write failed.
517: # otherwise $result is the number of bytes written.
518: # Remove that preceding string from the data.
519: # If the resulting data is empty, destroy the watcher
520: # and set up a read event handler to accept the next
521: # request.
522:
523: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
524: if($result ne undef) {
525: &Debug(9, "send result was defined");
526: if($result == length($Data)) { # Entire string sent.
527: &Debug(9, "ClientWritable data all written");
528: $Watcher->cancel();
529: #
530: # Set up to read next request from socket:
531:
532: my $descr = sprintf("Connection to lonc client %d",
533: $ActiveClients{$Socket});
534: Event->io(cb => \&ClientRequest,
535: poll => 'r',
536: desc => $descr,
537: data => "",
538: fd => $Socket);
539:
540: } else { # Partial string sent.
541: $Watcher->data(substr($Data, $result));
542: if($result == 0) { # client hung up on us!!
543: # Log("INFO", "lonc pipe client hung up on us!");
544: $Watcher->cancel;
545: $Socket->shutdown(2);
546: $Socket->close();
547: }
548: }
549:
550: } else { # Error of some sort...
551:
552: # Some errnos are possible:
553: my $errno = $!;
554: if($errno == POSIX::EWOULDBLOCK ||
555: $errno == POSIX::EAGAIN ||
556: $errno == POSIX::EINTR) {
557: # No action taken?
558: } else { # Unanticipated errno.
559: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
560: $Watcher->cancel; # Stop the watcher.
561: $Socket->shutdown(2); # Kill connection
562: $Socket->close(); # Close the socket.
563: }
564:
565: }
566: } else {
567: $Watcher->cancel(); # A delayed request...just cancel.
568: }
569: }
570:
571: =pod
572:
573: =head2 CompleteTransaction
574:
575: Called when the reply data has been received for a lond
576: transaction. The reply data must now be sent to the
577: ultimate client on the other end of the Unix socket. This is
578: done by setting up a writable event for the socket with the
579: data the reply data.
580:
581: Parameters:
582:
583: =item Socket
584:
585: Socket on which the lond transaction occured. This is a
586: LondConnection. The data received is in the TransactionReply member.
587:
588: =item Transaction
589:
590: The transaction that is being completed.
591:
592: =cut
593:
594: sub CompleteTransaction {
595: &Debug(5,"Complete transaction");
596:
597: my ($Socket, $Transaction) = @_;
598:
599: if (!$Transaction->isDeferred()) { # Normal transaction
600: my $data = $Socket->GetReply(); # Data to send.
601: if($LogTransactions) {
602: Log("SUCCESS", "Reply from lond: '$data'");
603: }
604: StartClientReply($Transaction, $data);
605: } else { # Delete deferred transaction file.
606: Log("SUCCESS", "A delayed transaction was completed");
607: LogPerm("S:$Transaction->getClient() :".$Transaction->getRequest());
608: unlink $Transaction->getFile();
609: }
610: }
611:
612: =pod
613:
614: =head1 StartClientReply
615:
616: Initiates a reply to a client where the reply data is a parameter.
617:
618: =head2 parameters:
619:
620: =item Transaction
621:
622: The transaction for which we are responding to the client.
623:
624: =item data
625:
626: The data to send to apached client.
627:
628: =cut
629:
630: sub StartClientReply {
631:
632: my ($Transaction, $data) = @_;
633:
634: my $Client = $Transaction->getClient();
635:
636: &Debug(8," Reply was: ".$data);
637: my $Serial = $ActiveClients{$Client};
638: my $desc = sprintf("Connection to lonc client %d",
639: $Serial);
640: Event->io(fd => $Client,
641: poll => "w",
642: desc => $desc,
643: cb => \&ClientWritable,
644: data => $data);
645: }
646:
647: =pod
648:
649: =head2 FailTransaction
650:
651: Finishes a transaction with failure because the associated lond socket
652: disconnected. There are two possibilities:
653: - The transaction is deferred: in which case we just quietly
654: delete the transaction since there is no client connection.
655: - The transaction is 'live' in which case we initiate the sending
656: of "con_lost" to the client.
657:
658: Deleting the transaction means killing it from the %ActiveTransactions hash.
659:
660: Parameters:
661:
662: =item client
663:
664: The LondTransaction we are failing.
665:
666:
667: =cut
668:
669: sub FailTransaction {
670: my $transaction = shift;
671:
672: # If the socket is dead, that's already logged.
673:
674: if ($ConnectionRetriesLeft > 0) {
675: Log("WARNING", "Failing transaction "
676: .$transaction->getLoggableRequest());
677: }
678: Debug(1, "Failing transaction: ".$transaction->getLoggableRequest());
679: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
680: my $client = $transaction->getClient();
681: Debug(1," Replying con_lost to ".$transaction->getRequest());
682: StartClientReply($transaction, "con_lost\n");
683: }
684:
685: }
686:
687: =pod
688:
689: =head1 EmptyQueue
690:
691: Fails all items in the work queue with con_lost.
692: Note that each item in the work queue is a transaction.
693:
694: =cut
695:
696: sub EmptyQueue {
697: $ConnectionRetriesLeft--; # Counts as connection failure too.
698: while($WorkQueue->Count()) {
699: my $request = $WorkQueue->dequeue(); # This is a transaction
700: FailTransaction($request);
701: }
702: }
703:
704: =pod
705:
706: =head2 CloseAllLondConnections
707:
708: Close all connections open on lond prior to exit e.g.
709:
710: =cut
711:
712: sub CloseAllLondConnections {
713: foreach my $Socket (keys %ActiveConnections) {
714: if(exists($ActiveTransactions{$Socket})) {
715: FailTransaction($ActiveTransactions{$Socket});
716: }
717: KillSocket($Socket);
718: }
719: }
720:
721: =pod
722:
723: =head2 KillSocket
724:
725: Destroys a socket. This function can be called either when a socket
726: has died of 'natural' causes or because a socket needs to be pruned due to
727: idleness. If the socket has died naturally, if there are no longer any
728: live connections a new connection is created (in case there are transactions
729: in the queue). If the socket has been pruned, it is never re-created.
730:
731: Parameters:
732:
733: =item Socket
734:
735: The socket to kill off.
736:
737: =item Restart
738:
739: nonzero if we are allowed to create a new connection.
740:
741: =cut
742:
743: sub KillSocket {
744: my $Socket = shift;
745:
746: Log("WARNING", "Shutting down a socket");
747: $Socket->Shutdown();
748:
749: # If the socket came from the active connection set,
750: # delete its transaction... note that FailTransaction should
751: # already have been called!!!
752: # otherwise it came from the idle set.
753: #
754:
755: if(exists($ActiveTransactions{$Socket})) {
756: delete ($ActiveTransactions{$Socket});
757: }
758: if(exists($ActiveConnections{$Socket})) {
759: delete($ActiveConnections{$Socket});
760: $ConnectionCount--;
761: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
762: }
763: # If the connection count has gone to zero and there is work in the
764: # work queue, the work all gets failed with con_lost.
765: #
766: if($ConnectionCount == 0) {
767: EmptyQueue();
768: CloseAllLondConnections; # Should all already be closed but...
769: }
770: }
771:
772: =pod
773:
774: =head2 LondReadable
775:
776: This function is called whenever a lond connection
777: is readable. The action is state dependent:
778:
779: =head3 State=Initialized
780:
781: We''re waiting for the challenge, this is a no-op until the
782: state changes.
783:
784: =head3 State=Challenged
785:
786: The challenge has arrived we need to transition to Writable.
787: The connection must echo the challenge back.
788:
789: =head3 State=ChallengeReplied
790:
791: The challenge has been replied to. The we are receiveing the
792: 'ok' from the partner.
793:
794: =head3 State=ReadingVersionString
795:
796: We have requested the lond version and are reading the
797: version back. Upon completion, we'll store the version away
798: for future use(?).
799:
800: =head3 State=HostSet
801:
802: We have selected the domain name of our peer (multhomed hosts)
803: and are getting the reply (presumably ok) back.
804:
805: =head3 State=RequestingKey
806:
807: The ok has been received and we need to send the request for
808: an encryption key. Transition to writable for that.
809:
810: =head3 State=ReceivingKey
811:
812: The the key has been requested, now we are reading the new key.
813:
814: =head3 State=Idle
815:
816: The encryption key has been negotiated or we have finished
817: reading data from the a transaction. If the callback data has
818: a client as well as the socket iformation, then we are
819: doing a transaction and the data received is relayed to the client
820: before the socket is put on the idle list.
821:
822: =head3 State=SendingRequest
823:
824: I do not think this state can be received here, but if it is,
825: the appropriate thing to do is to transition to writable, and send
826: the request.
827:
828: =head3 State=ReceivingReply
829:
830: We finished sending the request to the server and now transition
831: to readable to receive the reply.
832:
833: The parameter to this function are:
834:
835: The event. Implicit in this is the watcher and its data. The data
836: contains at least the lond connection object and, if a
837: transaction is in progress, the socket attached to the local client.
838:
839: =cut
840:
841: sub LondReadable {
842:
843: my $Event = shift;
844: my $Watcher = $Event->w;
845: my $Socket = $Watcher->data;
846: my $client = undef;
847:
848: &Debug(6,"LondReadable called state = ".$Socket->GetState());
849:
850:
851: my $State = $Socket->GetState(); # All action depends on the state.
852:
853: SocketDump(6, $Socket);
854: my $status = $Socket->Readable();
855:
856: &Debug(2, "Socket->Readable returned: $status");
857:
858: if($status != 0) {
859: # bad return from socket read. Currently this means that
860: # The socket has become disconnected. We fail the transaction.
861:
862: Log("WARNING",
863: "Lond connection lost.");
864: if(exists($ActiveTransactions{$Socket})) {
865: FailTransaction($ActiveTransactions{$Socket});
866: } else {
867: # Socket is connecting and failed... need to mark
868: # no longer connecting.
869:
870: $LondConnecting = 0;
871: }
872: $Watcher->cancel();
873: KillSocket($Socket);
874: $ConnectionRetriesLeft--; # Counts as connection failure
875: return;
876: }
877: SocketDump(6,$Socket);
878:
879: $State = $Socket->GetState(); # Update in case of transition.
880: &Debug(6, "After read, state is ".$State);
881:
882: if($State eq "Initialized") {
883:
884:
885: } elsif ($State eq "ChallengeReceived") {
886: # The challenge must be echoed back; The state machine
887: # in the connection takes care of setting that up. Just
888: # need to transition to writable:
889:
890: $Watcher->cb(\&LondWritable);
891: $Watcher->poll("w");
892:
893: } elsif ($State eq "ChallengeReplied") {
894:
895: } elsif ($State eq "RequestingVersion") {
896: # Need to ask for the version... that is writiability:
897:
898: $Watcher->cb(\&LondWritable);
899: $Watcher->poll("w");
900:
901: } elsif ($State eq "ReadingVersionString") {
902: # Read the rest of the version string...
903: } elsif ($State eq "SetHost") {
904: # Need to request the actual domain get set...
905:
906: $Watcher->cb(\&LondWritable);
907: $Watcher->poll("w");
908: } elsif ($State eq "HostSet") {
909: # Reading the 'ok' from the peer.
910:
911: } elsif ($State eq "RequestingKey") {
912: # The ok was received. Now we need to request the key
913: # That requires us to be writable:
914:
915: $Watcher->cb(\&LondWritable);
916: $Watcher->poll("w");
917:
918: } elsif ($State eq "ReceivingKey") {
919:
920: } elsif ($State eq "Idle") {
921:
922: # This is as good a spot as any to get the peer version
923: # string:
924:
925: if($LondVersion eq "unknown") {
926: $LondVersion = $Socket->PeerVersion();
927: Log("INFO", "Connected to lond version: $LondVersion");
928: }
929: # If necessary, complete a transaction and then go into the
930: # idle queue.
931: # Note that a trasition to idle indicates a live lond
932: # on the other end so reset the connection retries.
933: #
934: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
935: $Watcher->cancel();
936: if(exists($ActiveTransactions{$Socket})) {
937: Debug(5,"Completing transaction!!");
938: CompleteTransaction($Socket,
939: $ActiveTransactions{$Socket});
940: } else {
941: Log("SUCCESS", "Connection ".$ConnectionCount." to "
942: .$RemoteHost." now ready for action");
943: }
944: ServerToIdle($Socket); # Next work unit or idle.
945:
946: #
947: $LondConnecting = 0; # Best spot I can think of for this.
948: #
949:
950: } elsif ($State eq "SendingRequest") {
951: # We need to be writable for this and probably don't belong
952: # here inthe first place.
953:
954: Debug(6, "SendingRequest state encountered in readable");
955: $Watcher->poll("w");
956: $Watcher->cb(\&LondWritable);
957:
958: } elsif ($State eq "ReceivingReply") {
959:
960:
961: } else {
962: # Invalid state.
963: Debug(4, "Invalid state in LondReadable");
964: }
965: }
966:
967: =pod
968:
969: =head2 LondWritable
970:
971: This function is called whenever a lond connection
972: becomes writable while there is a writeable monitoring
973: event. The action taken is very state dependent:
974:
975: =head3 State = Connected
976:
977: The connection is in the process of sending the 'init' hailing to the
978: lond on the remote end. The connection object''s Writable member is
979: called. On error, ConnectionError is called to destroy the connection
980: and remove it from the ActiveConnections hash
981:
982: =head3 Initialized
983:
984: 'init' has been sent, writability monitoring is removed and
985: readability monitoring is started with LondReadable as the callback.
986:
987: =head3 ChallengeReceived
988:
989: The connection has received the who are you challenge from the remote
990: system, and is in the process of sending the challenge
991: response. Writable is called.
992:
993: =head3 ChallengeReplied
994:
995: The connection has replied to the initial challenge The we switch to
996: monitoring readability looking for the server to reply with 'ok'.
997:
998: =head3 RequestingKey
999:
1000: The connection is in the process of requesting its encryption key.
1001: Writable is called.
1002:
1003: =head3 ReceivingKey
1004:
1005: The connection has sent the request for a key. Switch to readability
1006: monitoring to accept the key
1007:
1008: =head3 SendingRequest
1009:
1010: The connection is in the process of sending a request to the server.
1011: This request is part of a client transaction. All the states until
1012: now represent the client setup protocol. Writable is called.
1013:
1014: =head3 ReceivingReply
1015:
1016: The connection has sent a request. Now it must receive a reply.
1017: Readability monitoring is requested.
1018:
1019: This function is an event handler and therefore receives as
1020: a parameter the event that has fired. The data for the watcher
1021: of this event is a reference to a list of one or two elements,
1022: depending on state. The first (and possibly only) element is the
1023: socket. The second (present only if a request is in progress)
1024: is the socket on which to return a reply to the caller.
1025:
1026: =cut
1027:
1028: sub LondWritable {
1029: my $Event = shift;
1030: my $Watcher = $Event->w;
1031: my $Socket = $Watcher->data;
1032: my $State = $Socket->GetState();
1033:
1034: Debug(6,"LondWritable State = ".$State."\n");
1035:
1036:
1037: # Figure out what to do depending on the state of the socket:
1038:
1039:
1040:
1041:
1042: SocketDump(6,$Socket);
1043:
1044: # If the socket is writable, we must always write.
1045: # Only by writing will we undergo state transitions.
1046: # Old logic wrote in state specific code below, however
1047: # That forces us at least through another invocation of
1048: # this function after writability is possible again.
1049: # This logic also factors out common code for handling
1050: # write failures... in all cases, write failures
1051: # Kill the socket.
1052: # This logic makes the branches of the >big< if below
1053: # so that the writing states are actually NO-OPs.
1054:
1055: if ($Socket->Writable() != 0) {
1056: # The write resulted in an error.
1057: # We'll treat this as if the socket got disconnected:
1058: Log("WARNING", "Connection to ".$RemoteHost.
1059: " has been disconnected");
1060: if(exists($ActiveTransactions{$Socket})) {
1061: FailTransaction($ActiveTransactions{$Socket});
1062: } else {
1063: # In the process of conneting, so need to turn that off.
1064:
1065: $LondConnecting = 0;
1066: }
1067: $Watcher->cancel();
1068: KillSocket($Socket);
1069: return;
1070: }
1071:
1072:
1073:
1074: if ($State eq "Connected") {
1075:
1076: # "init" is being sent...
1077:
1078: } elsif ($State eq "Initialized") {
1079:
1080: # Now that init was sent, we switch
1081: # to watching for readability:
1082:
1083: $Watcher->cb(\&LondReadable);
1084: $Watcher->poll("r");
1085:
1086: } elsif ($State eq "ChallengeReceived") {
1087: # We received the challenge, now we
1088: # are echoing it back. This is a no-op,
1089: # we're waiting for the state to change
1090:
1091: } elsif ($State eq "ChallengeReplied") {
1092: # The echo was sent back, so we switch
1093: # to watching readability.
1094:
1095: $Watcher->cb(\&LondReadable);
1096: $Watcher->poll("r");
1097: } elsif ($State eq "RequestingVersion") {
1098: # Sending the peer a version request...
1099:
1100: } elsif ($State eq "ReadingVersionString") {
1101: # Transition to read since we have sent the
1102: # version command and now just need to read the
1103: # version string from the peer:
1104:
1105: $Watcher->cb(\&LondReadable);
1106: $Watcher->poll("r");
1107:
1108: } elsif ($State eq "SetHost") {
1109: # Setting the remote domain...
1110:
1111: } elsif ($State eq "HostSet") {
1112: # Back to readable to get the ok.
1113:
1114: $Watcher->cb(\&LondReadable);
1115: $Watcher->poll("r");
1116:
1117:
1118: } elsif ($State eq "RequestingKey") {
1119: # At this time we're requesting the key.
1120: # again, this is essentially a no-op.
1121:
1122: } elsif ($State eq "ReceivingKey") {
1123: # Now we need to wait for the key
1124: # to come back from the peer:
1125:
1126: $Watcher->cb(\&LondReadable);
1127: $Watcher->poll("r");
1128:
1129: } elsif ($State eq "SendingRequest") {
1130:
1131: # At this time we are sending a request to the
1132: # peer... write the next chunk:
1133:
1134:
1135: } elsif ($State eq "ReceivingReply") {
1136: # The send has completed. Wait for the
1137: # data to come in for a reply.
1138: Debug(8,"Writable sent request/receiving reply");
1139: $Watcher->cb(\&LondReadable);
1140: $Watcher->poll("r");
1141:
1142: } else {
1143: # Control only passes here on an error:
1144: # the socket state does not match any
1145: # of the known states... so an error
1146: # must be logged.
1147:
1148: &Debug(4, "Invalid socket state ".$State."\n");
1149: }
1150:
1151: }
1152:
1153: =pod
1154:
1155: =cut
1156:
1157:
1158: sub QueueDelayed {
1159: Debug(3,"QueueDelayed called");
1160:
1161: my $path = "$perlvar{'lonSockDir'}/delayed";
1162:
1163: Debug(4, "Delayed path: ".$path);
1164: opendir(DIRHANDLE, $path);
1165:
1166: my $host_id_re = '(?:'.join('|',map {quotemeta($_)} (@all_host_ids)).')';
1167: my @alldelayed = grep(/\.$host_id_re$/, readdir(DIRHANDLE));
1168: closedir(DIRHANDLE);
1169: foreach my $dfname (sort(@alldelayed)) {
1170: my $reqfile = "$path/$dfname";
1171: my ($host_id) = ($dfname =~ /\.([^.]*)$/);
1172: Debug(4, "queueing ".$reqfile." for $host_id");
1173: my $Handle = IO::File->new($reqfile);
1174: my $cmd = <$Handle>;
1175: chomp $cmd; # There may or may not be a newline...
1176: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1177: my $Transaction = LondTransaction->new("sethost:$host_id:$cmd");
1178: $Transaction->SetDeferred($reqfile);
1179: QueueTransaction($Transaction);
1180: }
1181:
1182: }
1183:
1184: =pod
1185:
1186: =head2 MakeLondConnection
1187:
1188: Create a new lond connection object, and start it towards its initial
1189: idleness. Once idle, it becomes elligible to receive transactions
1190: from the work queue. If the work queue is not empty when the
1191: connection is completed and becomes idle, it will dequeue an entry and
1192: start off on it.
1193:
1194: =cut
1195:
1196: sub MakeLondConnection {
1197: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1198: .GetServerPort());
1199:
1200: my $Connection = LondConnection->new(&GetServerHost(),
1201: &GetServerPort(),
1202: &GetHostId());
1203:
1204: if($Connection eq undef) { # Needs to be more robust later.
1205: Log("CRITICAL","Failed to make a connection with lond.");
1206: $ConnectionRetriesLeft--;
1207: return 0; # Failure.
1208: } else {
1209:
1210: $LondConnecting = 1; # Connection in progress.
1211: # The connection needs to have writability
1212: # monitored in order to send the init sequence
1213: # that starts the whole authentication/key
1214: # exchange underway.
1215: #
1216: my $Socket = $Connection->GetSocket();
1217: if($Socket eq undef) {
1218: &child_exit(-1, "did not get a socket from the connection");
1219: } else {
1220: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1221: }
1222:
1223: $Connection->SetTimeoutCallback(\&SocketTimeout);
1224:
1225: my $event = Event->io(fd => $Socket,
1226: poll => 'w',
1227: cb => \&LondWritable,
1228: data => $Connection,
1229: desc => 'Connection to lond server');
1230: $ActiveConnections{$Connection} = $event;
1231: if ($ConnectionCount == 0) {
1232: &SetupTimer; # Need to handle timeouts with connections...
1233: }
1234: $ConnectionCount++;
1235: Debug(4, "Connection count = ".$ConnectionCount);
1236: if($ConnectionCount == 1) { # First Connection:
1237: QueueDelayed;
1238: }
1239: Log("SUCESS", "Created connection ".$ConnectionCount
1240: ." to host ".GetServerHost());
1241: return 1; # Return success.
1242: }
1243:
1244: }
1245:
1246: =pod
1247:
1248: =head2 StartRequest
1249:
1250: Starts a lond request going on a specified lond connection.
1251: parameters are:
1252:
1253: =item $Lond
1254:
1255: Connection to the lond that will send the transaction and receive the
1256: reply.
1257:
1258: =item $Client
1259:
1260: Connection to the client that is making this request We got the
1261: request from this socket, and when the request has been relayed to
1262: lond and we get a reply back from lond it will get sent to this
1263: socket.
1264:
1265: =item $Request
1266:
1267: The text of the request to send.
1268:
1269: =cut
1270:
1271: sub StartRequest {
1272:
1273: my ($Lond, $Request) = @_;
1274:
1275: Debug(6, "StartRequest: ".$Request->getRequest());
1276:
1277: my $Socket = $Lond->GetSocket();
1278:
1279: $Request->Activate($Lond);
1280: $ActiveTransactions{$Lond} = $Request;
1281:
1282: $Lond->InitiateTransaction($Request->getRequest());
1283: my $event = Event->io(fd => $Socket,
1284: poll => "w",
1285: cb => \&LondWritable,
1286: data => $Lond,
1287: desc => "lond transaction connection");
1288: $ActiveConnections{$Lond} = $event;
1289: Debug(8," Start Request made watcher data with ".$event->data."\n");
1290: }
1291:
1292: =pod
1293:
1294: =head2 QueueTransaction
1295:
1296: If there is an idle lond connection, it is put to work doing this
1297: transaction. Otherwise, the transaction is placed in the work queue.
1298: If placed in the work queue and the maximum number of connections has
1299: not yet been created, a new connection will be started. Our goal is
1300: to eventually have a sufficient number of connections that the work
1301: queue will typically be empty. parameters are:
1302:
1303: =item Socket
1304:
1305: open on the lonc client.
1306:
1307: =item Request
1308:
1309: data to send to the lond.
1310:
1311: =cut
1312:
1313: sub QueueTransaction {
1314:
1315: my $requestData = shift; # This is a LondTransaction.
1316: my $cmd = $requestData->getRequest();
1317:
1318: Debug(6,"QueueTransaction: ".$cmd);
1319:
1320: my $LondSocket = $IdleConnections->pop();
1321: if(!defined $LondSocket) { # Need to queue request.
1322: Debug(5,"Must queue...");
1323: $WorkQueue->enqueue($requestData);
1324: Debug(5, "Queue Transaction startnew $ConnectionCount $LondConnecting");
1325: if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) {
1326:
1327: if($ConnectionRetriesLeft > 0) {
1328: Debug(5,"Starting additional lond connection");
1329: if(&MakeLondConnection() == 0) {
1330: EmptyQueue(); # Fail transactions, can't make connection.
1331: CloseAllLondConnections; # Should all be closed but...
1332: }
1333: } else {
1334: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1335: $LondConnecting = 0;
1336: EmptyQueue(); # It's worse than that ... he's dead Jim.
1337: CloseAllLondConnections; # Should all be closed but..
1338: }
1339: }
1340: } else { # Can start the request:
1341: Debug(8,"Can start...");
1342: StartRequest($LondSocket, $requestData);
1343: }
1344: }
1345:
1346: #-------------------------- Lonc UNIX socket handling ---------------------
1347:
1348: =pod
1349:
1350: =head2 ClientRequest
1351: Callback that is called when data can be read from the UNIX domain
1352: socket connecting us with an apache server process.
1353:
1354: =cut
1355:
1356: sub ClientRequest {
1357: Debug(6, "ClientRequest");
1358: my $event = shift;
1359: my $watcher = $event->w;
1360: my $socket = $watcher->fd;
1361: my $data = $watcher->data;
1362: my $thisread;
1363:
1364: Debug(9, " Watcher named: ".$watcher->desc);
1365:
1366: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1367: Debug(8, "rcv: data length = ".length($thisread)
1368: ." read =".$thisread);
1369: unless (defined $rv && length($thisread)) {
1370: # Likely eof on socket.
1371: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1372: close($socket);
1373: $watcher->cancel();
1374: delete($ActiveClients{$socket});
1375: return;
1376: }
1377: Debug(8,"Data: ".$data." this read: ".$thisread);
1378: $data = $data.$thisread; # Append new data.
1379: $watcher->data($data);
1380: if($data =~ /\n$/) { # Request entirely read.
1381: if($data eq "close_connection_exit\n") {
1382: Log("CRITICAL",
1383: "Request Close Connection ... exiting");
1384: CloseAllLondConnections();
1385: exit;
1386: }
1387: Debug(8, "Complete transaction received: ".$data);
1388: if($LogTransactions) {
1389: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1390: }
1391: my $Transaction = LondTransaction->new($data);
1392: $Transaction->SetClient($socket);
1393: QueueTransaction($Transaction);
1394: $watcher->cancel(); # Done looking for input data.
1395: }
1396:
1397: }
1398:
1399: #
1400: # Accept a connection request for a client (lonc child) and
1401: # start up an event watcher to keep an eye on input from that
1402: # Event. This can be called both from NewClient and from
1403: # ChildProcess.
1404: # Parameters:
1405: # $socket - The listener socket.
1406: # Returns:
1407: # NONE
1408: # Side Effects:
1409: # An event is made to watch the accepted connection.
1410: # Active clients hash is updated to reflect the new connection.
1411: # The client connection count is incremented.
1412: #
1413: sub accept_client {
1414: my ($socket) = @_;
1415:
1416: Debug(8, "Entering accept for lonc UNIX socket\n");
1417: my $connection = $socket->accept(); # Accept the client connection.
1418: Debug(8,"Connection request accepted from "
1419: .GetPeername($connection, AF_UNIX));
1420:
1421:
1422: my $description = sprintf("Connection to lonc client %d",
1423: $ClientConnection);
1424: Debug(9, "Creating event named: ".$description);
1425: Event->io(cb => \&ClientRequest,
1426: poll => 'r',
1427: desc => $description,
1428: data => "",
1429: fd => $connection);
1430: $ActiveClients{$connection} = $ClientConnection;
1431: $ClientConnection++;
1432: }
1433:
1434: =pod
1435:
1436: =head2 NewClient
1437:
1438: Callback that is called when a connection is received on the unix
1439: socket for a new client of lonc. The callback is parameterized by the
1440: event.. which is a-priori assumed to be an io event, and therefore has
1441: an fd member that is the Listener socket. We Accept the connection
1442: and register a new event on the readability of that socket:
1443:
1444: =cut
1445:
1446: sub NewClient {
1447: Debug(6, "NewClient");
1448: my $event = shift; # Get the event parameters.
1449: my $watcher = $event->w;
1450: my $socket = $watcher->fd; # Get the event' socket.
1451:
1452: &accept_client($socket);
1453: }
1454:
1455: =pod
1456:
1457: =head2 GetLoncSocketPath
1458:
1459: Returns the name of the UNIX socket on which to listen for client
1460: connections.
1461:
1462: =head2 Parameters:
1463:
1464: host (optional) - Name of the host socket to return.. defaults to
1465: the return from GetServerHost().
1466:
1467: =cut
1468:
1469: sub GetLoncSocketPath {
1470:
1471: my $host = GetServerHost(); # Default host.
1472: if (@_) {
1473: ($host) = @_; # Override if supplied.
1474: }
1475: return $UnixSocketDir."/".$host;
1476: }
1477:
1478: =pod
1479:
1480: =head2 GetServerHost
1481:
1482: Returns the host whose lond we talk with.
1483:
1484: =cut
1485:
1486: sub GetServerHost {
1487: return $RemoteHost; # Setup by the fork.
1488: }
1489:
1490: =pod
1491:
1492: =head2 GetServerId
1493:
1494: Returns the hostid whose lond we talk with.
1495:
1496: =cut
1497:
1498: sub GetHostId {
1499: return $RemoteHostId; # Setup by the fork.
1500: }
1501:
1502: =pod
1503:
1504: =head2 GetServerPort
1505:
1506: Returns the lond port number.
1507:
1508: =cut
1509:
1510: sub GetServerPort {
1511: return $perlvar{londPort};
1512: }
1513:
1514: =pod
1515:
1516: =head2 SetupLoncListener
1517:
1518: Setup a lonc listener event. The event is called when the socket
1519: becomes readable.. that corresponds to the receipt of a new
1520: connection. The event handler established will accept the connection
1521: (creating a communcations channel), that int turn will establish
1522: another event handler to subess requests.
1523:
1524: =head2 Parameters:
1525:
1526: host (optional) Name of the host to set up a unix socket to.
1527:
1528: =cut
1529:
1530: sub SetupLoncListener {
1531: my ($host,$SocketName) = @_;
1532: if (!$host) { $host = &GetServerHost(); }
1533: if (!$SocketName) { $SocketName = &GetLoncSocketPath($host); }
1534:
1535:
1536: unlink($SocketName);
1537:
1538: my $socket;
1539: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1540: Listen => 250,
1541: Type => SOCK_STREAM)) {
1542: if($I_am_child) {
1543: &child_exit(-1, "Failed to create a lonc listener socket");
1544: } else {
1545: die "Failed to create a lonc listner socket";
1546: }
1547: }
1548: return $socket;
1549: }
1550:
1551: #
1552: # Toggle transaction logging.
1553: # Implicit inputs:
1554: # LogTransactions
1555: # Implicit Outputs:
1556: # LogTransactions
1557: sub ToggleTransactionLogging {
1558: print STDERR "Toggle transaction logging...\n";
1559: if(!$LogTransactions) {
1560: $LogTransactions = 1;
1561: } else {
1562: $LogTransactions = 0;
1563: }
1564:
1565:
1566: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1567: }
1568:
1569: =pod
1570:
1571: =head2 ChildStatus
1572:
1573: Child USR1 signal handler to report the most recent status
1574: into the status file.
1575:
1576: We also use this to reset the retries count in order to allow the
1577: client to retry connections with a previously dead server.
1578:
1579: =cut
1580:
1581: sub ChildStatus {
1582: my $event = shift;
1583: my $watcher = $event->w;
1584:
1585: Debug(2, "Reporting child status because : ".$watcher->data);
1586: my $docdir = $perlvar{'lonDocRoot'};
1587:
1588: open(LOG,">>$docdir/lon-status/loncstatus.txt");
1589: flock(LOG,LOCK_EX);
1590: print LOG $$."\t".$RemoteHost."\t".$Status."\t".
1591: $RecentLogEntry."\n";
1592: #
1593: # Write out information about each of the connections:
1594: #
1595: if ($DebugLevel > 2) {
1596: print LOG "Active connection statuses: \n";
1597: my $i = 1;
1598: print STDERR "================================= Socket Status Dump:\n";
1599: foreach my $item (keys %ActiveConnections) {
1600: my $Socket = $ActiveConnections{$item}->data;
1601: my $state = $Socket->GetState();
1602: print LOG "Connection $i State: $state\n";
1603: print STDERR "---------------------- Connection $i \n";
1604: $Socket->Dump(-1); # Ensure it gets dumped..
1605: $i++;
1606: }
1607: }
1608: flock(LOG,LOCK_UN);
1609: close(LOG);
1610: $ConnectionRetriesLeft = $ConnectionRetries;
1611: UpdateStatus();
1612: }
1613:
1614: =pod
1615:
1616: =head2 SignalledToDeath
1617:
1618: Called in response to a signal that causes a chid process to die.
1619:
1620: =cut
1621:
1622:
1623: sub SignalledToDeath {
1624: my $event = shift;
1625: my $watcher= $event->w;
1626:
1627: Debug(2,"Signalled to death! via ".$watcher->data);
1628: my ($signal) = $watcher->data;
1629: chomp($signal);
1630: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1631: ."died through "."\"$signal\"");
1632: #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1633: # ."\"$signal\"");
1634: exit 0;
1635:
1636: }
1637:
1638: =pod
1639:
1640: =head2 ToggleDebug
1641:
1642: This sub toggles trace debugging on and off.
1643:
1644: =cut
1645:
1646: sub ToggleDebug {
1647: my $Current = $DebugLevel;
1648: $DebugLevel = $NextDebugLevel;
1649: $NextDebugLevel = $Current;
1650:
1651: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1652:
1653: }
1654:
1655: =pod
1656:
1657: =head2 ChildProcess
1658:
1659: This sub implements a child process for a single lonc daemon.
1660: Optional parameter:
1661: $socket - if provided, this is a socket already open for listen
1662: on the client socket. Otherwise, a new listen is set up.
1663:
1664: =cut
1665:
1666: sub ChildProcess {
1667: # We've inherited all the
1668: # events of our parent and those have to be cancelled or else
1669: # all holy bloody chaos will result.. trust me, I already made
1670: # >that< mistake.
1671:
1672: my $host = GetServerHost();
1673: foreach my $listener (keys %parent_dispatchers) {
1674: my $watcher = $parent_dispatchers{$listener};
1675: my $s = $watcher->fd;
1676: if ($listener ne $host) { # Close everyone but me.
1677: Debug(5, "Closing listen socket for $listener");
1678: $s->close();
1679: }
1680: Debug(5, "Killing watcher for $listener");
1681:
1682: $watcher->cancel();
1683: delete($parent_dispatchers{$listener});
1684:
1685: }
1686:
1687: # kill off the parent's signal handlers too!
1688: #
1689:
1690: for my $handler (keys %parent_handlers) {
1691: my $watcher = $parent_handlers{$handler};
1692: $watcher->cancel();
1693: delete($parent_handlers{$handler});
1694: }
1695:
1696: $I_am_child = 1; # Seems like in spite of it all I may still getting
1697: # parent event dispatches.. flag I'm a child.
1698:
1699:
1700: #
1701: # Signals must be handled by the Event framework...
1702: #
1703:
1704: Event->signal(signal => "QUIT",
1705: cb => \&SignalledToDeath,
1706: data => "QUIT");
1707: Event->signal(signal => "HUP",
1708: cb => \&ChildStatus,
1709: data => "HUP");
1710: Event->signal(signal => "USR1",
1711: cb => \&ChildStatus,
1712: data => "USR1");
1713: Event->signal(signal => "USR2",
1714: cb => \&ToggleTransactionLogging);
1715: Event->signal(signal => "INT",
1716: cb => \&ToggleDebug,
1717: data => "INT");
1718:
1719: # Figure out if we got passed a socket or need to open one to listen for
1720: # client requests.
1721:
1722: my ($socket) = @_;
1723: if (!$socket) {
1724:
1725: $socket = SetupLoncListener();
1726: }
1727: # Establish an event to listen for client connection requests.
1728:
1729:
1730: Event->io(cb => \&NewClient,
1731: poll => 'r',
1732: desc => 'Lonc Listener Unix Socket',
1733: fd => $socket);
1734:
1735: $Event::DebugLevel = $DebugLevel;
1736:
1737: Debug(9, "Making initial lond connection for ".$RemoteHost);
1738:
1739: # Setup the initial server connection:
1740:
1741: # &MakeLondConnection(); // let first work request do it.
1742:
1743: # need to accept the connection since the event may not fire.
1744:
1745: &accept_client($socket);
1746:
1747: Debug(9,"Entering event loop");
1748: my $ret = Event::loop(); # Start the main event loop.
1749:
1750:
1751: &child_exit (-1,"Main event loop exited!!!");
1752: }
1753:
1754: # Create a new child for host passed in:
1755:
1756: sub CreateChild {
1757: my ($host, $hostid) = @_;
1758:
1759: my $sigset = POSIX::SigSet->new(SIGINT);
1760: sigprocmask(SIG_BLOCK, $sigset);
1761: $RemoteHost = $host;
1762: Log("CRITICAL", "Forking server for ".$host);
1763: my $pid = fork;
1764: if($pid) { # Parent
1765: $RemoteHost = "Parent";
1766: $ChildPid{$pid} = $host;
1767: sigprocmask(SIG_UNBLOCK, $sigset);
1768: undef(@all_host_ids);
1769: } else { # child.
1770: $RemoteHostId = $hostid;
1771: ShowStatus("Connected to ".$RemoteHost);
1772: $SIG{INT} = 'DEFAULT';
1773: sigprocmask(SIG_UNBLOCK, $sigset);
1774: &ChildProcess(); # Does not return.
1775: }
1776: }
1777:
1778: # parent_client_connection:
1779: # Event handler that processes client connections for the parent process.
1780: # This sub is called when the parent is listening on a socket and
1781: # a connection request arrives. We must:
1782: # Start a child process to accept the connection request.
1783: # Kill our listen on the socket.
1784: # Parameter:
1785: # event - The event object that was created to monitor this socket.
1786: # event->w->fd is the socket.
1787: # Returns:
1788: # NONE
1789: #
1790: sub parent_client_connection {
1791: if ($I_am_child) {
1792: # Should not get here, but seem to anyway:
1793: &Debug(5," Child caught parent client connection event!!");
1794: my ($event) = @_;
1795: my $watcher = $event->w;
1796: $watcher->cancel(); # Try to kill it off again!!
1797: } else {
1798: &Debug(9, "parent_client_connection");
1799: my ($event) = @_;
1800: my $watcher = $event->w;
1801: my $socket = $watcher->fd;
1802: my $connection = $socket->accept(); # Accept the client connection.
1803: Event->io(cb => \&get_remote_hostname,
1804: poll => 'r',
1805: data => "",
1806: fd => $connection);
1807: }
1808: }
1809:
1810: sub get_remote_hostname {
1811: my ($event) = @_;
1812: my $watcher = $event->w;
1813: my $socket = $watcher->fd;
1814:
1815: my $thisread;
1816: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1817: Debug(8, "rcv: data length = ".length($thisread)." read =".$thisread);
1818: if (!defined($rv) || length($thisread) == 0) {
1819: # Likely eof on socket.
1820: Debug(5,"Client Socket closed on lonc for p_c_c");
1821: close($socket);
1822: $watcher->cancel();
1823: return;
1824: }
1825:
1826: my $data = $watcher->data().$thisread;
1827: $watcher->data($data);
1828: if($data =~ /\n$/) { # Request entirely read.
1829: chomp($data);
1830: } else {
1831: return;
1832: }
1833:
1834: &Debug(5,"Creating child for $data (parent_client_connection)");
1835: (my $hostname,my $lonid,@all_host_ids) = split(':',$data);
1836: $ChildHost{$hostname}++;
1837: if ($ChildHost{$hostname} == 1) {
1838: &CreateChild($hostname,$lonid);
1839: } else {
1840: &Log('WARNING',"Request for a second child on $hostname");
1841: }
1842: # Clean up the listen since now the child takes over until it exits.
1843: $watcher->cancel(); # Nolonger listening to this event
1844: $socket->send("done\n");
1845: $socket->close();
1846: }
1847:
1848: # parent_listen:
1849: # Opens a socket and starts a listen for the parent process on a client UNIX
1850: # domain socket.
1851: #
1852: # This involves:
1853: # Creating a socket for listen.
1854: # Removing any socket lock file
1855: # Adding an event handler for this socket becoming readable
1856: # To the parent's event dispatcher.
1857: # Parameters:
1858: # loncapa_host - LonCAPA cluster name of the host represented by the client
1859: # socket.
1860: # Returns:
1861: # NONE
1862: #
1863: sub parent_listen {
1864: my ($loncapa_host) = @_;
1865: Debug(5, "parent_listen: $loncapa_host");
1866:
1867: my ($socket,$file);
1868: if (!$loncapa_host) {
1869: $loncapa_host = 'common_parent';
1870: $file = $perlvar{'lonSockCreate'};
1871: } else {
1872: $file = &GetLoncSocketPath($loncapa_host);
1873: }
1874: $socket = &SetupLoncListener($loncapa_host,$file);
1875:
1876: $listening_to{$socket} = $loncapa_host;
1877: if (!$socket) {
1878: die "Unable to create a listen socket for $loncapa_host";
1879: }
1880:
1881: my $lock_file = $file.".lock";
1882: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1883:
1884: my $watcher =
1885: Event->io(cb => \&parent_client_connection,
1886: poll => 'r',
1887: desc => "Parent listener unix socket ($loncapa_host)",
1888: data => "",
1889: fd => $socket);
1890: $parent_dispatchers{$loncapa_host} = $watcher;
1891:
1892: }
1893:
1894: sub parent_clean_up {
1895: my ($loncapa_host) = @_;
1896: Debug(-1, "parent_clean_up: $loncapa_host");
1897:
1898: my $socket_file = &GetLoncSocketPath($loncapa_host);
1899: unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
1900: my $lock_file = $socket_file.".lock";
1901: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1902: }
1903:
1904:
1905:
1906: # This sub initiates a listen on the common unix domain lonc client socket.
1907: # loncnew starts up with no children, and only spawns off children when a
1908: # connection request occurs on the common client unix socket. The spawned
1909: # child continues to run until it has been idle a while at which point it
1910: # eventually exits and once more the parent picks up the listen.
1911: #
1912: # Parameters:
1913: # NONE
1914: # Implicit Inputs:
1915: # The configuration file that has been read in by LondConnection.
1916: # Returns:
1917: # NONE
1918: #
1919: sub listen_on_common_socket {
1920: Debug(5, "listen_on_common_socket");
1921: &parent_listen();
1922: }
1923:
1924: # server_died is called whenever a child process exits.
1925: # Since this is dispatched via a signal, we must process all
1926: # dead children until there are no more left. The action
1927: # is to:
1928: # - Remove the child from the bookeeping hashes
1929: # - Re-establish a listen on the unix domain socket associated
1930: # with that host.
1931: # Parameters:
1932: # The event, but we don't actually care about it.
1933: sub server_died {
1934: &Debug(9, "server_died called...");
1935:
1936: while(1) { # Loop until waitpid nowait fails.
1937: my $pid = waitpid(-1, WNOHANG);
1938: if($pid <= 0) {
1939: return; # Nothing left to wait for.
1940: }
1941: # need the host to restart:
1942:
1943: my $host = $ChildPid{$pid};
1944: if($host) { # It's for real...
1945: &Debug(9, "Caught sigchild for $host");
1946: delete($ChildPid{$pid});
1947: delete($ChildHost{$host});
1948: &parent_clean_up($host);
1949:
1950: } else {
1951: &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
1952: }
1953: }
1954:
1955: }
1956:
1957: #
1958: # Parent process logic pass 1:
1959: # For each entry in the hosts table, we will
1960: # fork off an instance of ChildProcess to service the transactions
1961: # to that host. Each pid will be entered in a global hash
1962: # with the value of the key, the host.
1963: # The parent will then enter a loop to wait for process exits.
1964: # Each exit gets logged and the child gets restarted.
1965: #
1966:
1967: #
1968: # Fork and start in new session so hang-up isn't going to
1969: # happen without intent.
1970: #
1971:
1972:
1973:
1974:
1975:
1976:
1977: ShowStatus("Forming new session");
1978: my $childpid = fork;
1979: if ($childpid != 0) {
1980: sleep 4; # Give child a chacne to break to
1981: exit 0; # a new sesion.
1982: }
1983: #
1984: # Write my pid into the pid file so I can be located
1985: #
1986:
1987: ShowStatus("Parent writing pid file:");
1988: my $execdir = $perlvar{'lonDaemons'};
1989: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1990: print PIDSAVE "$$\n";
1991: close(PIDSAVE);
1992:
1993:
1994:
1995: if (POSIX::setsid() < 0) {
1996: print "Could not create new session\n";
1997: exit -1;
1998: }
1999:
2000: ShowStatus("Forking node servers");
2001:
2002: Log("CRITICAL", "--------------- Starting children ---------------");
2003:
2004: LondConnection::ReadConfig; # Read standard config files.
2005:
2006: $RemoteHost = "[parent]";
2007: &listen_on_common_socket();
2008:
2009: $RemoteHost = "Parent Server";
2010:
2011: # Maintain the population:
2012:
2013: ShowStatus("Parent keeping the flock");
2014:
2015:
2016: # We need to setup a SIGChild event to handle the exit (natural or otherwise)
2017: # of the children.
2018:
2019: Event->signal(cb => \&server_died,
2020: desc => "Child exit handler",
2021: signal => "CHLD");
2022:
2023:
2024: # Set up all the other signals we set up.
2025:
2026: $parent_handlers{INT} = Event->signal(cb => \&Terminate,
2027: desc => "Parent INT handler",
2028: signal => "INT");
2029: $parent_handlers{TERM} = Event->signal(cb => \&Terminate,
2030: desc => "Parent TERM handler",
2031: signal => "TERM");
2032: $parent_handlers{HUP} = Event->signal(cb => \&KillThemAll,
2033: desc => "Parent HUP handler.",
2034: signal => "HUP");
2035: $parent_handlers{USR1} = Event->signal(cb => \&CheckKids,
2036: desc => "Parent USR1 handler",
2037: signal => "USR1");
2038: $parent_handlers{USR2} = Event->signal(cb => \&UpdateKids,
2039: desc => "Parent USR2 handler.",
2040: signal => "USR2");
2041:
2042: # Start procdesing events.
2043:
2044: $Event::DebugLevel = $DebugLevel;
2045: Debug(9, "Parent entering event loop");
2046: my $ret = Event::loop();
2047: die "Main Event loop exited: $ret";
2048:
2049: =pod
2050:
2051: =head1 CheckKids
2052:
2053: Since kids do not die as easily in this implementation
2054: as the previous one, there is no need to restart the
2055: dead ones (all dead kids get restarted when they die!!)
2056: The only thing this function does is to pass USR1 to the
2057: kids so that they report their status.
2058:
2059: =cut
2060:
2061: sub CheckKids {
2062: Debug(2, "Checking status of children");
2063: my $docdir = $perlvar{'lonDocRoot'};
2064: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
2065: my $now=time;
2066: my $local=localtime($now);
2067: print $fh "LONC status $local - parent $$ \n\n";
2068: foreach my $host (keys %parent_dispatchers) {
2069: print $fh "LONC Parent process listening for $host\n";
2070: }
2071: foreach my $pid (keys %ChildPid) {
2072: Debug(2, "Sending USR1 -> $pid");
2073: kill 'USR1' => $pid; # Tell Child to report status.
2074: }
2075:
2076: }
2077:
2078: =pod
2079:
2080: =head1 UpdateKids
2081:
2082: parent's SIGUSR2 handler. This handler:
2083:
2084: =item
2085:
2086: Rereads the hosts file.
2087:
2088: =item
2089:
2090: Kills off (via sigint) children for hosts that have disappeared.
2091:
2092: =item
2093:
2094: QUITs children for hosts that already exist (this just forces a status display
2095: and resets the connection retry count for that host.
2096:
2097: =item
2098:
2099: Starts new children for hosts that have been added to the hosts.tab file since
2100: the start of the master program and maintains them.
2101:
2102: =cut
2103:
2104: sub UpdateKids {
2105:
2106: Log("INFO", "Updating connections via SIGUSR2");
2107:
2108: # I'm not sure what I was thinking in the first implementation.
2109: # someone will have to work hard to convince me the effect is any
2110: # different than Restart, especially now that we don't start up
2111: # per host servers automatically, may as well just restart.
2112: # The down side is transactions that are in flight will get timed out
2113: # (lost unless they are critical).
2114:
2115: &KillThemAll();
2116: }
2117:
2118:
2119: =pod
2120:
2121: =head1 Restart
2122:
2123: Signal handler for HUP... all children are killed and
2124: we self restart. This is an el-cheapo way to re read
2125: the config file.
2126:
2127: =cut
2128:
2129: sub Restart {
2130: &KillThemAll; # First kill all the children.
2131: Log("CRITICAL", "Restarting");
2132: my $execdir = $perlvar{'lonDaemons'};
2133: unlink("$execdir/logs/lonc.pid");
2134: exec("$executable");
2135: }
2136:
2137: =pod
2138:
2139: =head1 KillThemAll
2140:
2141: Signal handler that kills all children by sending them a
2142: SIGHUP. Responds to sigint and sigterm.
2143:
2144: =cut
2145:
2146: sub KillThemAll {
2147: Debug(2, "Kill them all!!");
2148: local($SIG{CHLD}) = 'IGNORE'; # Our children >will< die.
2149: foreach my $pid (keys %ChildPid) {
2150: my $serving = $ChildPid{$pid};
2151: ShowStatus("Nicely Killing lonc for $serving pid = $pid");
2152: Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
2153: kill 'QUIT' => $pid;
2154: }
2155: }
2156:
2157:
2158: #
2159: # Kill all children via KILL. Just in case the
2160: # first shot didn't get them.
2161:
2162: sub really_kill_them_all_dammit
2163: {
2164: Debug(2, "Kill them all Dammit");
2165: local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
2166: foreach my $pid (keys %ChildPid) {
2167: my $serving = $ChildPid{$pid};
2168: &ShowStatus("Nastily killing lonc for $serving pid = $pid");
2169: Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
2170: kill 'KILL' => $pid;
2171: delete($ChildPid{$pid});
2172: my $execdir = $perlvar{'lonDaemons'};
2173: unlink("$execdir/logs/lonc.pid");
2174: }
2175: }
2176:
2177: =pod
2178:
2179: =head1 Terminate
2180:
2181: Terminate the system.
2182:
2183: =cut
2184:
2185: sub Terminate {
2186: &Log("CRITICAL", "Asked to kill children.. first be nice...");
2187: &KillThemAll;
2188: #
2189: # By now they really should all be dead.. but just in case
2190: # send them all SIGKILL's after a bit of waiting:
2191:
2192: sleep(4);
2193: &Log("CRITICAL", "Now kill children nasty");
2194: &really_kill_them_all_dammit;
2195: Log("CRITICAL","Master process exiting");
2196: exit 0;
2197:
2198: }
2199:
2200: sub my_hostname {
2201: use Sys::Hostname;
2202: my $name = &hostname();
2203: &Debug(9,"Name is $name");
2204: return $name;
2205: }
2206:
2207: =pod
2208:
2209: =head1 Theory
2210:
2211: The event class is used to build this as a single process with an
2212: event driven model. The following events are handled:
2213:
2214: =item UNIX Socket connection Received
2215:
2216: =item Request data arrives on UNIX data transfer socket.
2217:
2218: =item lond connection becomes writable.
2219:
2220: =item timer fires at 1 second intervals.
2221:
2222: All sockets are run in non-blocking mode. Timeouts managed by the timer
2223: handler prevents hung connections.
2224:
2225: Key data structures:
2226:
2227: =item RequestQueue
2228:
2229: A queue of requests received from UNIX sockets that are
2230: waiting for a chance to be forwarded on a lond connection socket.
2231:
2232: =item ActiveConnections
2233:
2234: A hash of lond connections that have transactions in process that are
2235: available to be timed out.
2236:
2237: =item ActiveTransactions
2238:
2239: A hash indexed by lond connections that contain the client reply
2240: socket for each connection that has an active transaction on it.
2241:
2242: =item IdleConnections
2243:
2244: A hash of lond connections that have no work to do. These connections
2245: can be closed if they are idle for a long enough time.
2246:
2247: =cut
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/back.gif){kind=icon}
Return to loncnew CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [LON-CAPA] / loncom