1: #!/usr/bin/perl
2: # The LearningOnline Network with CAPA
3: # lonc maintains the connections to remote computers
4: #
5: # $Id: loncnew,v 1.86 2007/06/13 03:05:54 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: chomp($message);
162: print $fh "$now:$message:$local\n";
163: }
164:
165: =pod
166:
167: =head2 Log
168:
169: Logs a message to the log file.
170: Parameters:
171:
172: =item severity
173:
174: One of CRITICAL, WARNING, INFO, SUCCESS used to select the
175: format string used to format the message. if the severity is
176: not a defined severity the Default format string is used.
177:
178: =item message
179:
180: The base message. In addtion to the format string, the message
181: will be appended to a string containing the name of our remote
182: host and the time will be formatted into the message.
183:
184: =cut
185:
186: sub Log {
187:
188: my ($severity, $message) = @_;
189:
190: if(!$LogFormats{$severity}) {
191: $severity = "DEFAULT";
192: }
193:
194: my $format = $LogFormats{$severity};
195:
196: # Put the window dressing in in front of the message format:
197:
198: my $now = time;
199: my $local = localtime($now);
200: my $finalformat = "$local ($$) [$RemoteHost] [$Status] ";
201: $finalformat = $finalformat.$format."\n";
202:
203: # open the file and put the result.
204:
205: my $execdir = $perlvar{'lonDaemons'};
206: my $fh = IO::File->new(">>$execdir/logs/lonc.log");
207: my $msg = sprintf($finalformat, $message);
208: $RecentLogEntry = $msg;
209: print $fh $msg;
210:
211:
212: }
213:
214:
215: =pod
216:
217: =head2 GetPeerName
218:
219: Returns the name of the host that a socket object is connected to.
220:
221: =cut
222:
223: sub GetPeername {
224:
225:
226: my ($connection, $AdrFamily) = @_;
227:
228: my $peer = $connection->peername();
229: my $peerport;
230: my $peerip;
231: if($AdrFamily == AF_INET) {
232: ($peerport, $peerip) = sockaddr_in($peer);
233: my $peername = gethostbyaddr($peerip, $AdrFamily);
234: return $peername;
235: } elsif ($AdrFamily == AF_UNIX) {
236: my $peerfile;
237: ($peerfile) = sockaddr_un($peer);
238: return $peerfile;
239: }
240: }
241: =pod
242:
243: =head2 Debug
244:
245: Invoked to issue a debug message.
246:
247: =cut
248:
249: sub Debug {
250:
251: my ($level, $message) = @_;
252:
253: if ($level <= $DebugLevel) {
254: Log("INFO", "-Debug- $message host = $RemoteHost");
255: }
256: }
257:
258: sub SocketDump {
259:
260: my ($level, $socket) = @_;
261:
262: if($level <= $DebugLevel) {
263: $socket->Dump(-1); # Ensure it will get dumped.
264: }
265: }
266:
267: =pod
268:
269: =head2 ShowStatus
270:
271: Place some text as our pid status.
272: and as what we return in a SIGUSR1
273:
274: =cut
275:
276: sub ShowStatus {
277: my $state = shift;
278: my $now = time;
279: my $local = localtime($now);
280: $Status = $local.": ".$state;
281: $0='lonc: '.$state.' '.$local;
282: }
283:
284: =pod
285:
286: =head2 SocketTimeout
287:
288: Called when an action on the socket times out. The socket is
289: destroyed and any active transaction is failed.
290:
291:
292: =cut
293:
294: sub SocketTimeout {
295: my $Socket = shift;
296: Log("WARNING", "A socket timeout was detected");
297: Debug(5, " SocketTimeout called: ");
298: $Socket->Dump(0);
299: if(exists($ActiveTransactions{$Socket})) {
300: FailTransaction($ActiveTransactions{$Socket});
301: }
302: KillSocket($Socket); # A transaction timeout also counts as
303: # a connection failure:
304: $ConnectionRetriesLeft--;
305: if($ConnectionRetriesLeft <= 0) {
306: Log("CRITICAL", "Host marked DEAD: ".GetServerHost());
307: $LondConnecting = 0;
308: }
309:
310: }
311:
312: #
313: # This function should be called by the child in all cases where it must
314: # exit. The child process must create a lock file for the AF_UNIX socket
315: # in order to prevent connection requests from lonnet in the time between
316: # process exit and the parent picking up the listen again.
317: #
318: # Parameters:
319: # exit_code - Exit status value, however see the next parameter.
320: # message - If this optional parameter is supplied, the exit
321: # is via a die with this message.
322: #
323: sub child_exit {
324: my ($exit_code, $message) = @_;
325:
326: # Regardless of how we exit, we may need to do the lock thing:
327:
328: #
329: # Create a lock file since there will be a time window
330: # between our exit and the parent's picking up the listen
331: # during which no listens will be done on the
332: # lonnet client socket.
333: #
334: my $lock_file = &GetLoncSocketPath().".lock";
335: open(LOCK,">$lock_file");
336: print LOCK "Contents not important";
337: close(LOCK);
338: unlink(&GetLoncSocketPath());
339:
340: if ($message) {
341: die($message);
342: } else {
343: exit($exit_code);
344: }
345: }
346: #----------------------------- Timer management ------------------------
347:
348: =pod
349:
350: =head2 Tick
351:
352: Invoked each timer tick.
353:
354: =cut
355:
356:
357: sub Tick {
358: my ($Event) = @_;
359: my $clock_watcher = $Event->w;
360:
361: my $client;
362: UpdateStatus();
363:
364: # Is it time to prune connection count:
365:
366:
367: if($IdleConnections->Count() &&
368: ($WorkQueue->Count() == 0)) { # Idle connections and nothing to do?
369: $IdleSeconds++;
370: if($IdleSeconds > $IdleTimeout) { # Prune a connection...
371: my $Socket = $IdleConnections->pop();
372: KillSocket($Socket);
373: $IdleSeconds = 0; # Otherwise all connections get trimmed to fast.
374: UpdateStatus();
375: if(($ConnectionCount == 0)) {
376: &child_exit(0);
377:
378: }
379: }
380: } else {
381: $IdleSeconds = 0; # Reset idle count if not idle.
382: }
383: #
384: # For each inflight transaction, tick down its timeout counter.
385: #
386:
387: foreach my $item (keys %ActiveConnections) {
388: my $State = $ActiveConnections{$item}->data->GetState();
389: if ($State ne 'Idle') {
390: Debug(5,"Ticking Socket $State $item");
391: $ActiveConnections{$item}->data->Tick();
392: }
393: }
394: # Do we have work in the queue, but no connections to service them?
395: # If so, try to make some new connections to get things going again.
396: #
397: # Note this code is dead now...
398: #
399: my $Requests = $WorkQueue->Count();
400: if (($ConnectionCount == 0) && ($Requests > 0) && (!$LondConnecting)) {
401: if ($ConnectionRetriesLeft > 0) {
402: Debug(5,"Work but no connections, Make a new one");
403: my $success;
404: $success = &MakeLondConnection;
405: if($success == 0) { # All connections failed:
406: Debug(5,"Work in queue failed to make any connectiouns\n");
407: EmptyQueue(); # Fail pending transactions with con_lost.
408: CloseAllLondConnections(); # Should all be closed but....
409: }
410: } else {
411: $LondConnecting = 0;
412: ShowStatus(GetServerHost()." >>> DEAD!!! <<<");
413: Debug(5,"Work in queue, but gave up on connections..flushing\n");
414: EmptyQueue(); # Connections can't be established.
415: CloseAllLondConnections(); # Should all already be closed but...
416: }
417:
418: }
419: if ($ConnectionCount == 0) {
420: $KeyMode = "";
421: $clock_watcher->cancel();
422: }
423: &UpdateStatus();
424: }
425:
426: =pod
427:
428: =head2 SetupTimer
429:
430: Sets up a 1 per sec recurring timer event. The event handler is used to:
431:
432: =item
433:
434: Trigger timeouts on communications along active sockets.
435:
436: =item
437:
438: Trigger disconnections of idle sockets.
439:
440: =cut
441:
442: sub SetupTimer {
443: Debug(6, "SetupTimer");
444: Event->timer(interval => 1, cb => \&Tick );
445: }
446:
447: =pod
448:
449: =head2 ServerToIdle
450:
451: This function is called when a connection to the server is
452: ready for more work.
453:
454: If there is work in the Work queue the top element is dequeued
455: and the connection will start to work on it. If the work queue is
456: empty, the connection is pushed on the idle connection stack where
457: it will either get another work unit, or alternatively, if it sits there
458: long enough, it will be shut down and released.
459:
460: =cut
461:
462: sub ServerToIdle {
463: my $Socket = shift; # Get the socket.
464: $KeyMode = $Socket->{AuthenticationMode};
465: delete($ActiveTransactions{$Socket}); # Server has no transaction
466:
467: &Debug(5, "Server to idle");
468:
469: # If there's work to do, start the transaction:
470:
471: my $reqdata = $WorkQueue->dequeue(); # This is a LondTransaction
472: if ($reqdata ne undef) {
473: Debug(5, "Queue gave request data: ".$reqdata->getRequest());
474: &StartRequest($Socket, $reqdata);
475:
476: } else {
477:
478: # There's no work waiting, so push the server to idle list.
479: &Debug(5, "No new work requests, server connection going idle");
480: $IdleConnections->push($Socket);
481: }
482: }
483:
484: =pod
485:
486: =head2 ClientWritable
487:
488: Event callback for when a client socket is writable.
489:
490: This callback is established when a transaction reponse is
491: avaiable from lond. The response is forwarded to the unix socket
492: as it becomes writable in this sub.
493:
494: Parameters:
495:
496: =item Event
497:
498: The event that has been triggered. Event->w->data is
499: the data and Event->w->fd is the socket to write.
500:
501: =cut
502:
503: sub ClientWritable {
504: my $Event = shift;
505: my $Watcher = $Event->w;
506: if (!defined($Watcher)) {
507: &child_exit(-1,'No watcher for event in ClientWritable');
508: }
509: my $Data = $Watcher->data;
510: my $Socket = $Watcher->fd;
511:
512: # Try to send the data:
513:
514: &Debug(6, "ClientWritable writing".$Data);
515: &Debug(9, "Socket is: ".$Socket);
516:
517: if($Socket->connected) {
518: my $result = $Socket->send($Data, 0);
519:
520: # $result undefined: the write failed.
521: # otherwise $result is the number of bytes written.
522: # Remove that preceding string from the data.
523: # If the resulting data is empty, destroy the watcher
524: # and set up a read event handler to accept the next
525: # request.
526:
527: &Debug(9,"Send result is ".$result." Defined: ".defined($result));
528: if($result ne undef) {
529: &Debug(9, "send result was defined");
530: if($result == length($Data)) { # Entire string sent.
531: &Debug(9, "ClientWritable data all written");
532: $Watcher->cancel();
533: #
534: # Set up to read next request from socket:
535:
536: my $descr = sprintf("Connection to lonc client %d",
537: $ActiveClients{$Socket});
538: Event->io(cb => \&ClientRequest,
539: poll => 'r',
540: desc => $descr,
541: data => "",
542: fd => $Socket);
543:
544: } else { # Partial string sent.
545: $Watcher->data(substr($Data, $result));
546: if($result == 0) { # client hung up on us!!
547: # Log("INFO", "lonc pipe client hung up on us!");
548: $Watcher->cancel;
549: $Socket->shutdown(2);
550: $Socket->close();
551: }
552: }
553:
554: } else { # Error of some sort...
555:
556: # Some errnos are possible:
557: my $errno = $!;
558: if($errno == POSIX::EWOULDBLOCK ||
559: $errno == POSIX::EAGAIN ||
560: $errno == POSIX::EINTR) {
561: # No action taken?
562: } else { # Unanticipated errno.
563: &Debug(5,"ClientWritable error or peer shutdown".$RemoteHost);
564: $Watcher->cancel; # Stop the watcher.
565: $Socket->shutdown(2); # Kill connection
566: $Socket->close(); # Close the socket.
567: }
568:
569: }
570: } else {
571: $Watcher->cancel(); # A delayed request...just cancel.
572: return;
573: }
574: }
575:
576: =pod
577:
578: =head2 CompleteTransaction
579:
580: Called when the reply data has been received for a lond
581: transaction. The reply data must now be sent to the
582: ultimate client on the other end of the Unix socket. This is
583: done by setting up a writable event for the socket with the
584: data the reply data.
585:
586: Parameters:
587:
588: =item Socket
589:
590: Socket on which the lond transaction occured. This is a
591: LondConnection. The data received is in the TransactionReply member.
592:
593: =item Transaction
594:
595: The transaction that is being completed.
596:
597: =cut
598:
599: sub CompleteTransaction {
600: &Debug(5,"Complete transaction");
601:
602: my ($Socket, $Transaction) = @_;
603:
604: if (!$Transaction->isDeferred()) { # Normal transaction
605: my $data = $Socket->GetReply(); # Data to send.
606: if($LogTransactions) {
607: Log("SUCCESS", "Reply from lond: '$data'");
608: }
609: StartClientReply($Transaction, $data);
610: } else { # Delete deferred transaction file.
611: Log("SUCCESS", "A delayed transaction was completed");
612: LogPerm("S:".$Transaction->getClient().":".$Transaction->getRequest());
613: unlink($Transaction->getFile());
614: }
615: }
616:
617: =pod
618:
619: =head1 StartClientReply
620:
621: Initiates a reply to a client where the reply data is a parameter.
622:
623: =head2 parameters:
624:
625: =item Transaction
626:
627: The transaction for which we are responding to the client.
628:
629: =item data
630:
631: The data to send to apached client.
632:
633: =cut
634:
635: sub StartClientReply {
636:
637: my ($Transaction, $data) = @_;
638:
639: my $Client = $Transaction->getClient();
640:
641: &Debug(8," Reply was: ".$data);
642: my $Serial = $ActiveClients{$Client};
643: my $desc = sprintf("Connection to lonc client %d",
644: $Serial);
645: Event->io(fd => $Client,
646: poll => "w",
647: desc => $desc,
648: cb => \&ClientWritable,
649: data => $data);
650: }
651:
652: =pod
653:
654: =head2 FailTransaction
655:
656: Finishes a transaction with failure because the associated lond socket
657: disconnected. There are two possibilities:
658: - The transaction is deferred: in which case we just quietly
659: delete the transaction since there is no client connection.
660: - The transaction is 'live' in which case we initiate the sending
661: of "con_lost" to the client.
662:
663: Deleting the transaction means killing it from the %ActiveTransactions hash.
664:
665: Parameters:
666:
667: =item client
668:
669: The LondTransaction we are failing.
670:
671:
672: =cut
673:
674: sub FailTransaction {
675: my $transaction = shift;
676:
677: # If the socket is dead, that's already logged.
678:
679: if ($ConnectionRetriesLeft > 0) {
680: Log("WARNING", "Failing transaction "
681: .$transaction->getLoggableRequest());
682: }
683: Debug(1, "Failing transaction: ".$transaction->getLoggableRequest());
684: if (!$transaction->isDeferred()) { # If the transaction is deferred we'll get to it.
685: my $client = $transaction->getClient();
686: Debug(1," Replying con_lost to ".$transaction->getRequest());
687: StartClientReply($transaction, "con_lost\n");
688: }
689:
690: }
691:
692: =pod
693:
694: =head1 EmptyQueue
695:
696: Fails all items in the work queue with con_lost.
697: Note that each item in the work queue is a transaction.
698:
699: =cut
700:
701: sub EmptyQueue {
702: $ConnectionRetriesLeft--; # Counts as connection failure too.
703: while($WorkQueue->Count()) {
704: my $request = $WorkQueue->dequeue(); # This is a transaction
705: FailTransaction($request);
706: }
707: }
708:
709: =pod
710:
711: =head2 CloseAllLondConnections
712:
713: Close all connections open on lond prior to exit e.g.
714:
715: =cut
716:
717: sub CloseAllLondConnections {
718: foreach my $Socket (keys %ActiveConnections) {
719: if(exists($ActiveTransactions{$Socket})) {
720: FailTransaction($ActiveTransactions{$Socket});
721: }
722: KillSocket($Socket);
723: }
724: }
725:
726: =pod
727:
728: =head2 KillSocket
729:
730: Destroys a socket. This function can be called either when a socket
731: has died of 'natural' causes or because a socket needs to be pruned due to
732: idleness. If the socket has died naturally, if there are no longer any
733: live connections a new connection is created (in case there are transactions
734: in the queue). If the socket has been pruned, it is never re-created.
735:
736: Parameters:
737:
738: =item Socket
739:
740: The socket to kill off.
741:
742: =item Restart
743:
744: nonzero if we are allowed to create a new connection.
745:
746: =cut
747:
748: sub KillSocket {
749: my $Socket = shift;
750:
751: Log("WARNING", "Shutting down a socket");
752: $Socket->Shutdown();
753:
754: # If the socket came from the active connection set,
755: # delete its transaction... note that FailTransaction should
756: # already have been called!!!
757: # otherwise it came from the idle set.
758: #
759:
760: if(exists($ActiveTransactions{$Socket})) {
761: delete ($ActiveTransactions{$Socket});
762: }
763: if(exists($ActiveConnections{$Socket})) {
764: delete($ActiveConnections{$Socket});
765: $ConnectionCount--;
766: if ($ConnectionCount < 0) { $ConnectionCount = 0; }
767: }
768: # If the connection count has gone to zero and there is work in the
769: # work queue, the work all gets failed with con_lost.
770: #
771: if($ConnectionCount == 0) {
772: EmptyQueue();
773: CloseAllLondConnections; # Should all already be closed but...
774: }
775: }
776:
777: =pod
778:
779: =head2 LondReadable
780:
781: This function is called whenever a lond connection
782: is readable. The action is state dependent:
783:
784: =head3 State=Initialized
785:
786: We''re waiting for the challenge, this is a no-op until the
787: state changes.
788:
789: =head3 State=Challenged
790:
791: The challenge has arrived we need to transition to Writable.
792: The connection must echo the challenge back.
793:
794: =head3 State=ChallengeReplied
795:
796: The challenge has been replied to. The we are receiveing the
797: 'ok' from the partner.
798:
799: =head3 State=ReadingVersionString
800:
801: We have requested the lond version and are reading the
802: version back. Upon completion, we'll store the version away
803: for future use(?).
804:
805: =head3 State=HostSet
806:
807: We have selected the domain name of our peer (multhomed hosts)
808: and are getting the reply (presumably ok) back.
809:
810: =head3 State=RequestingKey
811:
812: The ok has been received and we need to send the request for
813: an encryption key. Transition to writable for that.
814:
815: =head3 State=ReceivingKey
816:
817: The the key has been requested, now we are reading the new key.
818:
819: =head3 State=Idle
820:
821: The encryption key has been negotiated or we have finished
822: reading data from the a transaction. If the callback data has
823: a client as well as the socket iformation, then we are
824: doing a transaction and the data received is relayed to the client
825: before the socket is put on the idle list.
826:
827: =head3 State=SendingRequest
828:
829: I do not think this state can be received here, but if it is,
830: the appropriate thing to do is to transition to writable, and send
831: the request.
832:
833: =head3 State=ReceivingReply
834:
835: We finished sending the request to the server and now transition
836: to readable to receive the reply.
837:
838: The parameter to this function are:
839:
840: The event. Implicit in this is the watcher and its data. The data
841: contains at least the lond connection object and, if a
842: transaction is in progress, the socket attached to the local client.
843:
844: =cut
845:
846: sub LondReadable {
847:
848: my $Event = shift;
849: my $Watcher = $Event->w;
850: my $Socket = $Watcher->data;
851: my $client = undef;
852:
853: &Debug(6,"LondReadable called state = ".$Socket->GetState());
854:
855:
856: my $State = $Socket->GetState(); # All action depends on the state.
857:
858: SocketDump(6, $Socket);
859: my $status = $Socket->Readable();
860:
861: &Debug(2, "Socket->Readable returned: $status");
862:
863: if($status != 0) {
864: # bad return from socket read. Currently this means that
865: # The socket has become disconnected. We fail the transaction.
866:
867: Log("WARNING",
868: "Lond connection lost.");
869: if(exists($ActiveTransactions{$Socket})) {
870: FailTransaction($ActiveTransactions{$Socket});
871: } else {
872: # Socket is connecting and failed... need to mark
873: # no longer connecting.
874:
875: $LondConnecting = 0;
876: }
877: $Watcher->cancel();
878: KillSocket($Socket);
879: $ConnectionRetriesLeft--; # Counts as connection failure
880: return;
881: }
882: SocketDump(6,$Socket);
883:
884: $State = $Socket->GetState(); # Update in case of transition.
885: &Debug(6, "After read, state is ".$State);
886:
887: if($State eq "Initialized") {
888:
889:
890: } elsif ($State eq "ChallengeReceived") {
891: # The challenge must be echoed back; The state machine
892: # in the connection takes care of setting that up. Just
893: # need to transition to writable:
894:
895: $Watcher->cb(\&LondWritable);
896: $Watcher->poll("w");
897:
898: } elsif ($State eq "ChallengeReplied") {
899:
900: } elsif ($State eq "RequestingVersion") {
901: # Need to ask for the version... that is writiability:
902:
903: $Watcher->cb(\&LondWritable);
904: $Watcher->poll("w");
905:
906: } elsif ($State eq "ReadingVersionString") {
907: # Read the rest of the version string...
908: } elsif ($State eq "SetHost") {
909: # Need to request the actual domain get set...
910:
911: $Watcher->cb(\&LondWritable);
912: $Watcher->poll("w");
913: } elsif ($State eq "HostSet") {
914: # Reading the 'ok' from the peer.
915:
916: } elsif ($State eq "RequestingKey") {
917: # The ok was received. Now we need to request the key
918: # That requires us to be writable:
919:
920: $Watcher->cb(\&LondWritable);
921: $Watcher->poll("w");
922:
923: } elsif ($State eq "ReceivingKey") {
924:
925: } elsif ($State eq "Idle") {
926:
927: # This is as good a spot as any to get the peer version
928: # string:
929:
930: if($LondVersion eq "unknown") {
931: $LondVersion = $Socket->PeerVersion();
932: Log("INFO", "Connected to lond version: $LondVersion");
933: }
934: # If necessary, complete a transaction and then go into the
935: # idle queue.
936: # Note that a trasition to idle indicates a live lond
937: # on the other end so reset the connection retries.
938: #
939: $ConnectionRetriesLeft = $ConnectionRetries; # success resets the count
940: $Watcher->cancel();
941: if(exists($ActiveTransactions{$Socket})) {
942: Debug(5,"Completing transaction!!");
943: CompleteTransaction($Socket,
944: $ActiveTransactions{$Socket});
945: } else {
946: Log("SUCCESS", "Connection ".$ConnectionCount." to "
947: .$RemoteHost." now ready for action");
948: }
949: ServerToIdle($Socket); # Next work unit or idle.
950:
951: #
952: $LondConnecting = 0; # Best spot I can think of for this.
953: #
954:
955: } elsif ($State eq "SendingRequest") {
956: # We need to be writable for this and probably don't belong
957: # here inthe first place.
958:
959: Debug(6, "SendingRequest state encountered in readable");
960: $Watcher->poll("w");
961: $Watcher->cb(\&LondWritable);
962:
963: } elsif ($State eq "ReceivingReply") {
964:
965:
966: } else {
967: # Invalid state.
968: Debug(4, "Invalid state in LondReadable");
969: }
970: }
971:
972: =pod
973:
974: =head2 LondWritable
975:
976: This function is called whenever a lond connection
977: becomes writable while there is a writeable monitoring
978: event. The action taken is very state dependent:
979:
980: =head3 State = Connected
981:
982: The connection is in the process of sending the 'init' hailing to the
983: lond on the remote end. The connection object''s Writable member is
984: called. On error, ConnectionError is called to destroy the connection
985: and remove it from the ActiveConnections hash
986:
987: =head3 Initialized
988:
989: 'init' has been sent, writability monitoring is removed and
990: readability monitoring is started with LondReadable as the callback.
991:
992: =head3 ChallengeReceived
993:
994: The connection has received the who are you challenge from the remote
995: system, and is in the process of sending the challenge
996: response. Writable is called.
997:
998: =head3 ChallengeReplied
999:
1000: The connection has replied to the initial challenge The we switch to
1001: monitoring readability looking for the server to reply with 'ok'.
1002:
1003: =head3 RequestingKey
1004:
1005: The connection is in the process of requesting its encryption key.
1006: Writable is called.
1007:
1008: =head3 ReceivingKey
1009:
1010: The connection has sent the request for a key. Switch to readability
1011: monitoring to accept the key
1012:
1013: =head3 SendingRequest
1014:
1015: The connection is in the process of sending a request to the server.
1016: This request is part of a client transaction. All the states until
1017: now represent the client setup protocol. Writable is called.
1018:
1019: =head3 ReceivingReply
1020:
1021: The connection has sent a request. Now it must receive a reply.
1022: Readability monitoring is requested.
1023:
1024: This function is an event handler and therefore receives as
1025: a parameter the event that has fired. The data for the watcher
1026: of this event is a reference to a list of one or two elements,
1027: depending on state. The first (and possibly only) element is the
1028: socket. The second (present only if a request is in progress)
1029: is the socket on which to return a reply to the caller.
1030:
1031: =cut
1032:
1033: sub LondWritable {
1034: my $Event = shift;
1035: my $Watcher = $Event->w;
1036: my $Socket = $Watcher->data;
1037: my $State = $Socket->GetState();
1038:
1039: Debug(6,"LondWritable State = ".$State."\n");
1040:
1041:
1042: # Figure out what to do depending on the state of the socket:
1043:
1044:
1045:
1046:
1047: SocketDump(6,$Socket);
1048:
1049: # If the socket is writable, we must always write.
1050: # Only by writing will we undergo state transitions.
1051: # Old logic wrote in state specific code below, however
1052: # That forces us at least through another invocation of
1053: # this function after writability is possible again.
1054: # This logic also factors out common code for handling
1055: # write failures... in all cases, write failures
1056: # Kill the socket.
1057: # This logic makes the branches of the >big< if below
1058: # so that the writing states are actually NO-OPs.
1059:
1060: if ($Socket->Writable() != 0) {
1061: # The write resulted in an error.
1062: # We'll treat this as if the socket got disconnected:
1063: Log("WARNING", "Connection to ".$RemoteHost.
1064: " has been disconnected");
1065: if(exists($ActiveTransactions{$Socket})) {
1066: FailTransaction($ActiveTransactions{$Socket});
1067: } else {
1068: # In the process of conneting, so need to turn that off.
1069:
1070: $LondConnecting = 0;
1071: }
1072: $Watcher->cancel();
1073: KillSocket($Socket);
1074: return;
1075: }
1076:
1077:
1078:
1079: if ($State eq "Connected") {
1080:
1081: # "init" is being sent...
1082:
1083: } elsif ($State eq "Initialized") {
1084:
1085: # Now that init was sent, we switch
1086: # to watching for readability:
1087:
1088: $Watcher->cb(\&LondReadable);
1089: $Watcher->poll("r");
1090:
1091: } elsif ($State eq "ChallengeReceived") {
1092: # We received the challenge, now we
1093: # are echoing it back. This is a no-op,
1094: # we're waiting for the state to change
1095:
1096: } elsif ($State eq "ChallengeReplied") {
1097: # The echo was sent back, so we switch
1098: # to watching readability.
1099:
1100: $Watcher->cb(\&LondReadable);
1101: $Watcher->poll("r");
1102: } elsif ($State eq "RequestingVersion") {
1103: # Sending the peer a version request...
1104:
1105: } elsif ($State eq "ReadingVersionString") {
1106: # Transition to read since we have sent the
1107: # version command and now just need to read the
1108: # version string from the peer:
1109:
1110: $Watcher->cb(\&LondReadable);
1111: $Watcher->poll("r");
1112:
1113: } elsif ($State eq "SetHost") {
1114: # Setting the remote domain...
1115:
1116: } elsif ($State eq "HostSet") {
1117: # Back to readable to get the ok.
1118:
1119: $Watcher->cb(\&LondReadable);
1120: $Watcher->poll("r");
1121:
1122:
1123: } elsif ($State eq "RequestingKey") {
1124: # At this time we're requesting the key.
1125: # again, this is essentially a no-op.
1126:
1127: } elsif ($State eq "ReceivingKey") {
1128: # Now we need to wait for the key
1129: # to come back from the peer:
1130:
1131: $Watcher->cb(\&LondReadable);
1132: $Watcher->poll("r");
1133:
1134: } elsif ($State eq "SendingRequest") {
1135:
1136: # At this time we are sending a request to the
1137: # peer... write the next chunk:
1138:
1139:
1140: } elsif ($State eq "ReceivingReply") {
1141: # The send has completed. Wait for the
1142: # data to come in for a reply.
1143: Debug(8,"Writable sent request/receiving reply");
1144: $Watcher->cb(\&LondReadable);
1145: $Watcher->poll("r");
1146:
1147: } else {
1148: # Control only passes here on an error:
1149: # the socket state does not match any
1150: # of the known states... so an error
1151: # must be logged.
1152:
1153: &Debug(4, "Invalid socket state ".$State."\n");
1154: }
1155:
1156: }
1157:
1158: =pod
1159:
1160: =cut
1161:
1162:
1163: sub QueueDelayed {
1164: Debug(3,"QueueDelayed called");
1165:
1166: my $path = "$perlvar{'lonSockDir'}/delayed";
1167:
1168: Debug(4, "Delayed path: ".$path);
1169: opendir(DIRHANDLE, $path);
1170:
1171: my $host_id_re = '(?:'.join('|',map {quotemeta($_)} (@all_host_ids)).')';
1172: my @alldelayed = grep(/\.$host_id_re$/, readdir(DIRHANDLE));
1173: closedir(DIRHANDLE);
1174: foreach my $dfname (sort(@alldelayed)) {
1175: my $reqfile = "$path/$dfname";
1176: my ($host_id) = ($dfname =~ /\.([^.]*)$/);
1177: Debug(4, "queueing ".$reqfile." for $host_id");
1178: my $Handle = IO::File->new($reqfile);
1179: my $cmd = <$Handle>;
1180: chomp $cmd; # There may or may not be a newline...
1181: $cmd = $cmd."\n"; # now for sure there's exactly one newline.
1182: my $Transaction = LondTransaction->new("sethost:$host_id:$cmd");
1183: $Transaction->SetDeferred($reqfile);
1184: QueueTransaction($Transaction);
1185: }
1186:
1187: }
1188:
1189: =pod
1190:
1191: =head2 MakeLondConnection
1192:
1193: Create a new lond connection object, and start it towards its initial
1194: idleness. Once idle, it becomes elligible to receive transactions
1195: from the work queue. If the work queue is not empty when the
1196: connection is completed and becomes idle, it will dequeue an entry and
1197: start off on it.
1198:
1199: =cut
1200:
1201: sub MakeLondConnection {
1202: Debug(4,"MakeLondConnection to ".GetServerHost()." on port "
1203: .GetServerPort());
1204:
1205: my $Connection = LondConnection->new(&GetServerHost(),
1206: &GetServerPort(),
1207: &GetHostId());
1208:
1209: if($Connection eq undef) { # Needs to be more robust later.
1210: Log("CRITICAL","Failed to make a connection with lond.");
1211: $ConnectionRetriesLeft--;
1212: return 0; # Failure.
1213: } else {
1214:
1215: $LondConnecting = 1; # Connection in progress.
1216: # The connection needs to have writability
1217: # monitored in order to send the init sequence
1218: # that starts the whole authentication/key
1219: # exchange underway.
1220: #
1221: my $Socket = $Connection->GetSocket();
1222: if($Socket eq undef) {
1223: &child_exit(-1, "did not get a socket from the connection");
1224: } else {
1225: &Debug(9,"MakeLondConnection got socket: ".$Socket);
1226: }
1227:
1228: $Connection->SetTimeoutCallback(\&SocketTimeout);
1229:
1230: my $event = Event->io(fd => $Socket,
1231: poll => 'w',
1232: cb => \&LondWritable,
1233: data => $Connection,
1234: desc => 'Connection to lond server');
1235: $ActiveConnections{$Connection} = $event;
1236: if ($ConnectionCount == 0) {
1237: &SetupTimer; # Need to handle timeouts with connections...
1238: }
1239: $ConnectionCount++;
1240: Debug(4, "Connection count = ".$ConnectionCount);
1241: if($ConnectionCount == 1) { # First Connection:
1242: QueueDelayed;
1243: }
1244: Log("SUCESS", "Created connection ".$ConnectionCount
1245: ." to host ".GetServerHost());
1246: return 1; # Return success.
1247: }
1248:
1249: }
1250:
1251: =pod
1252:
1253: =head2 StartRequest
1254:
1255: Starts a lond request going on a specified lond connection.
1256: parameters are:
1257:
1258: =item $Lond
1259:
1260: Connection to the lond that will send the transaction and receive the
1261: reply.
1262:
1263: =item $Client
1264:
1265: Connection to the client that is making this request We got the
1266: request from this socket, and when the request has been relayed to
1267: lond and we get a reply back from lond it will get sent to this
1268: socket.
1269:
1270: =item $Request
1271:
1272: The text of the request to send.
1273:
1274: =cut
1275:
1276: sub StartRequest {
1277:
1278: my ($Lond, $Request) = @_;
1279:
1280: Debug(6, "StartRequest: ".$Request->getRequest());
1281:
1282: my $Socket = $Lond->GetSocket();
1283:
1284: $Request->Activate($Lond);
1285: $ActiveTransactions{$Lond} = $Request;
1286:
1287: $Lond->InitiateTransaction($Request->getRequest());
1288: my $event = Event->io(fd => $Socket,
1289: poll => "w",
1290: cb => \&LondWritable,
1291: data => $Lond,
1292: desc => "lond transaction connection");
1293: $ActiveConnections{$Lond} = $event;
1294: Debug(8," Start Request made watcher data with ".$event->data."\n");
1295: }
1296:
1297: =pod
1298:
1299: =head2 QueueTransaction
1300:
1301: If there is an idle lond connection, it is put to work doing this
1302: transaction. Otherwise, the transaction is placed in the work queue.
1303: If placed in the work queue and the maximum number of connections has
1304: not yet been created, a new connection will be started. Our goal is
1305: to eventually have a sufficient number of connections that the work
1306: queue will typically be empty. parameters are:
1307:
1308: =item Socket
1309:
1310: open on the lonc client.
1311:
1312: =item Request
1313:
1314: data to send to the lond.
1315:
1316: =cut
1317:
1318: sub QueueTransaction {
1319:
1320: my $requestData = shift; # This is a LondTransaction.
1321: my $cmd = $requestData->getRequest();
1322:
1323: Debug(6,"QueueTransaction: ".$cmd);
1324:
1325: my $LondSocket = $IdleConnections->pop();
1326: if(!defined $LondSocket) { # Need to queue request.
1327: Debug(5,"Must queue...");
1328: $WorkQueue->enqueue($requestData);
1329: Debug(5, "Queue Transaction startnew $ConnectionCount $LondConnecting");
1330: if(($ConnectionCount < $MaxConnectionCount) && (! $LondConnecting)) {
1331:
1332: if($ConnectionRetriesLeft > 0) {
1333: Debug(5,"Starting additional lond connection");
1334: if(&MakeLondConnection() == 0) {
1335: EmptyQueue(); # Fail transactions, can't make connection.
1336: CloseAllLondConnections; # Should all be closed but...
1337: }
1338: } else {
1339: ShowStatus(GetServerHost()." >>> DEAD !!!! <<<");
1340: $LondConnecting = 0;
1341: EmptyQueue(); # It's worse than that ... he's dead Jim.
1342: CloseAllLondConnections; # Should all be closed but..
1343: }
1344: }
1345: } else { # Can start the request:
1346: Debug(8,"Can start...");
1347: StartRequest($LondSocket, $requestData);
1348: }
1349: }
1350:
1351: #-------------------------- Lonc UNIX socket handling ---------------------
1352:
1353: =pod
1354:
1355: =head2 ClientRequest
1356: Callback that is called when data can be read from the UNIX domain
1357: socket connecting us with an apache server process.
1358:
1359: =cut
1360:
1361: sub ClientRequest {
1362: Debug(6, "ClientRequest");
1363: my $event = shift;
1364: my $watcher = $event->w;
1365: my $socket = $watcher->fd;
1366: my $data = $watcher->data;
1367: my $thisread;
1368:
1369: Debug(9, " Watcher named: ".$watcher->desc);
1370:
1371: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1372: Debug(8, "rcv: data length = ".length($thisread)
1373: ." read =".$thisread);
1374: unless (defined $rv && length($thisread)) {
1375: # Likely eof on socket.
1376: Debug(5,"Client Socket closed on lonc for ".$RemoteHost);
1377: close($socket);
1378: $watcher->cancel();
1379: delete($ActiveClients{$socket});
1380: return;
1381: }
1382: Debug(8,"Data: ".$data." this read: ".$thisread);
1383: $data = $data.$thisread; # Append new data.
1384: $watcher->data($data);
1385: if($data =~ /\n$/) { # Request entirely read.
1386: if($data eq "close_connection_exit\n") {
1387: Log("CRITICAL",
1388: "Request Close Connection ... exiting");
1389: CloseAllLondConnections();
1390: exit;
1391: }
1392: Debug(8, "Complete transaction received: ".$data);
1393: if($LogTransactions) {
1394: Log("SUCCESS", "Transaction: '$data'"); # Transaction has \n.
1395: }
1396: my $Transaction = LondTransaction->new($data);
1397: $Transaction->SetClient($socket);
1398: QueueTransaction($Transaction);
1399: $watcher->cancel(); # Done looking for input data.
1400: }
1401:
1402: }
1403:
1404: #
1405: # Accept a connection request for a client (lonc child) and
1406: # start up an event watcher to keep an eye on input from that
1407: # Event. This can be called both from NewClient and from
1408: # ChildProcess.
1409: # Parameters:
1410: # $socket - The listener socket.
1411: # Returns:
1412: # NONE
1413: # Side Effects:
1414: # An event is made to watch the accepted connection.
1415: # Active clients hash is updated to reflect the new connection.
1416: # The client connection count is incremented.
1417: #
1418: sub accept_client {
1419: my ($socket) = @_;
1420:
1421: Debug(8, "Entering accept for lonc UNIX socket\n");
1422: my $connection = $socket->accept(); # Accept the client connection.
1423: Debug(8,"Connection request accepted from "
1424: .GetPeername($connection, AF_UNIX));
1425:
1426:
1427: my $description = sprintf("Connection to lonc client %d",
1428: $ClientConnection);
1429: Debug(9, "Creating event named: ".$description);
1430: Event->io(cb => \&ClientRequest,
1431: poll => 'r',
1432: desc => $description,
1433: data => "",
1434: fd => $connection);
1435: $ActiveClients{$connection} = $ClientConnection;
1436: $ClientConnection++;
1437: }
1438:
1439: =pod
1440:
1441: =head2 NewClient
1442:
1443: Callback that is called when a connection is received on the unix
1444: socket for a new client of lonc. The callback is parameterized by the
1445: event.. which is a-priori assumed to be an io event, and therefore has
1446: an fd member that is the Listener socket. We Accept the connection
1447: and register a new event on the readability of that socket:
1448:
1449: =cut
1450:
1451: sub NewClient {
1452: Debug(6, "NewClient");
1453: my $event = shift; # Get the event parameters.
1454: my $watcher = $event->w;
1455: my $socket = $watcher->fd; # Get the event' socket.
1456:
1457: &accept_client($socket);
1458: }
1459:
1460: =pod
1461:
1462: =head2 GetLoncSocketPath
1463:
1464: Returns the name of the UNIX socket on which to listen for client
1465: connections.
1466:
1467: =head2 Parameters:
1468:
1469: host (optional) - Name of the host socket to return.. defaults to
1470: the return from GetServerHost().
1471:
1472: =cut
1473:
1474: sub GetLoncSocketPath {
1475:
1476: my $host = GetServerHost(); # Default host.
1477: if (@_) {
1478: ($host) = @_; # Override if supplied.
1479: }
1480: return $UnixSocketDir."/".$host;
1481: }
1482:
1483: =pod
1484:
1485: =head2 GetServerHost
1486:
1487: Returns the host whose lond we talk with.
1488:
1489: =cut
1490:
1491: sub GetServerHost {
1492: return $RemoteHost; # Setup by the fork.
1493: }
1494:
1495: =pod
1496:
1497: =head2 GetServerId
1498:
1499: Returns the hostid whose lond we talk with.
1500:
1501: =cut
1502:
1503: sub GetHostId {
1504: return $RemoteHostId; # Setup by the fork.
1505: }
1506:
1507: =pod
1508:
1509: =head2 GetServerPort
1510:
1511: Returns the lond port number.
1512:
1513: =cut
1514:
1515: sub GetServerPort {
1516: return $perlvar{londPort};
1517: }
1518:
1519: =pod
1520:
1521: =head2 SetupLoncListener
1522:
1523: Setup a lonc listener event. The event is called when the socket
1524: becomes readable.. that corresponds to the receipt of a new
1525: connection. The event handler established will accept the connection
1526: (creating a communcations channel), that int turn will establish
1527: another event handler to subess requests.
1528:
1529: =head2 Parameters:
1530:
1531: host (optional) Name of the host to set up a unix socket to.
1532:
1533: =cut
1534:
1535: sub SetupLoncListener {
1536: my ($host,$SocketName) = @_;
1537: if (!$host) { $host = &GetServerHost(); }
1538: if (!$SocketName) { $SocketName = &GetLoncSocketPath($host); }
1539:
1540:
1541: unlink($SocketName);
1542:
1543: my $socket;
1544: unless ($socket =IO::Socket::UNIX->new(Local => $SocketName,
1545: Listen => 250,
1546: Type => SOCK_STREAM)) {
1547: if($I_am_child) {
1548: &child_exit(-1, "Failed to create a lonc listener socket");
1549: } else {
1550: die "Failed to create a lonc listner socket";
1551: }
1552: }
1553: return $socket;
1554: }
1555:
1556: #
1557: # Toggle transaction logging.
1558: # Implicit inputs:
1559: # LogTransactions
1560: # Implicit Outputs:
1561: # LogTransactions
1562: sub ToggleTransactionLogging {
1563: print STDERR "Toggle transaction logging...\n";
1564: if(!$LogTransactions) {
1565: $LogTransactions = 1;
1566: } else {
1567: $LogTransactions = 0;
1568: }
1569:
1570:
1571: Log("SUCCESS", "Toggled transaction logging: $LogTransactions \n");
1572: }
1573:
1574: =pod
1575:
1576: =head2 ChildStatus
1577:
1578: Child USR1 signal handler to report the most recent status
1579: into the status file.
1580:
1581: We also use this to reset the retries count in order to allow the
1582: client to retry connections with a previously dead server.
1583:
1584: =cut
1585:
1586: sub ChildStatus {
1587: my $event = shift;
1588: my $watcher = $event->w;
1589:
1590: Debug(2, "Reporting child status because : ".$watcher->data);
1591: my $docdir = $perlvar{'lonDocRoot'};
1592:
1593: open(LOG,">>$docdir/lon-status/loncstatus.txt");
1594: flock(LOG,LOCK_EX);
1595: print LOG $$."\t".$RemoteHost."\t".$Status."\t".
1596: $RecentLogEntry."\n";
1597: #
1598: # Write out information about each of the connections:
1599: #
1600: if ($DebugLevel > 2) {
1601: print LOG "Active connection statuses: \n";
1602: my $i = 1;
1603: print STDERR "================================= Socket Status Dump:\n";
1604: foreach my $item (keys %ActiveConnections) {
1605: my $Socket = $ActiveConnections{$item}->data;
1606: my $state = $Socket->GetState();
1607: print LOG "Connection $i State: $state\n";
1608: print STDERR "---------------------- Connection $i \n";
1609: $Socket->Dump(-1); # Ensure it gets dumped..
1610: $i++;
1611: }
1612: }
1613: flock(LOG,LOCK_UN);
1614: close(LOG);
1615: $ConnectionRetriesLeft = $ConnectionRetries;
1616: UpdateStatus();
1617: }
1618:
1619: =pod
1620:
1621: =head2 SignalledToDeath
1622:
1623: Called in response to a signal that causes a chid process to die.
1624:
1625: =cut
1626:
1627:
1628: sub SignalledToDeath {
1629: my $event = shift;
1630: my $watcher= $event->w;
1631:
1632: Debug(2,"Signalled to death! via ".$watcher->data);
1633: my ($signal) = $watcher->data;
1634: chomp($signal);
1635: Log("CRITICAL", "Abnormal exit. Child $$ for $RemoteHost "
1636: ."died through "."\"$signal\"");
1637: #LogPerm("F:lonc: $$ on $RemoteHost signalled to death: "
1638: # ."\"$signal\"");
1639: exit 0;
1640:
1641: }
1642:
1643: =pod
1644:
1645: =head2 ToggleDebug
1646:
1647: This sub toggles trace debugging on and off.
1648:
1649: =cut
1650:
1651: sub ToggleDebug {
1652: my $Current = $DebugLevel;
1653: $DebugLevel = $NextDebugLevel;
1654: $NextDebugLevel = $Current;
1655:
1656: Log("SUCCESS", "New debugging level for $RemoteHost now $DebugLevel");
1657:
1658: }
1659:
1660: =pod
1661:
1662: =head2 ChildProcess
1663:
1664: This sub implements a child process for a single lonc daemon.
1665: Optional parameter:
1666: $socket - if provided, this is a socket already open for listen
1667: on the client socket. Otherwise, a new listen is set up.
1668:
1669: =cut
1670:
1671: sub ChildProcess {
1672: # We've inherited all the
1673: # events of our parent and those have to be cancelled or else
1674: # all holy bloody chaos will result.. trust me, I already made
1675: # >that< mistake.
1676:
1677: my $host = GetServerHost();
1678: foreach my $listener (keys %parent_dispatchers) {
1679: my $watcher = $parent_dispatchers{$listener};
1680: my $s = $watcher->fd;
1681: if ($listener ne $host) { # Close everyone but me.
1682: Debug(5, "Closing listen socket for $listener");
1683: $s->close();
1684: }
1685: Debug(5, "Killing watcher for $listener");
1686:
1687: $watcher->cancel();
1688: delete($parent_dispatchers{$listener});
1689:
1690: }
1691:
1692: # kill off the parent's signal handlers too!
1693: #
1694:
1695: for my $handler (keys %parent_handlers) {
1696: my $watcher = $parent_handlers{$handler};
1697: $watcher->cancel();
1698: delete($parent_handlers{$handler});
1699: }
1700:
1701: $I_am_child = 1; # Seems like in spite of it all I may still getting
1702: # parent event dispatches.. flag I'm a child.
1703:
1704:
1705: #
1706: # Signals must be handled by the Event framework...
1707: #
1708:
1709: Event->signal(signal => "QUIT",
1710: cb => \&SignalledToDeath,
1711: data => "QUIT");
1712: Event->signal(signal => "HUP",
1713: cb => \&ChildStatus,
1714: data => "HUP");
1715: Event->signal(signal => "USR1",
1716: cb => \&ChildStatus,
1717: data => "USR1");
1718: Event->signal(signal => "USR2",
1719: cb => \&ToggleTransactionLogging);
1720: Event->signal(signal => "INT",
1721: cb => \&ToggleDebug,
1722: data => "INT");
1723:
1724: # Figure out if we got passed a socket or need to open one to listen for
1725: # client requests.
1726:
1727: my ($socket) = @_;
1728: if (!$socket) {
1729:
1730: $socket = SetupLoncListener();
1731: }
1732: # Establish an event to listen for client connection requests.
1733:
1734:
1735: Event->io(cb => \&NewClient,
1736: poll => 'r',
1737: desc => 'Lonc Listener Unix Socket',
1738: fd => $socket);
1739:
1740: $Event::DebugLevel = $DebugLevel;
1741:
1742: Debug(9, "Making initial lond connection for ".$RemoteHost);
1743:
1744: # Setup the initial server connection:
1745:
1746: # &MakeLondConnection(); // let first work request do it.
1747:
1748: # need to accept the connection since the event may not fire.
1749:
1750: &accept_client($socket);
1751:
1752: Debug(9,"Entering event loop");
1753: my $ret = Event::loop(); # Start the main event loop.
1754:
1755:
1756: &child_exit (-1,"Main event loop exited!!!");
1757: }
1758:
1759: # Create a new child for host passed in:
1760:
1761: sub CreateChild {
1762: my ($host, $hostid) = @_;
1763:
1764: my $sigset = POSIX::SigSet->new(SIGINT);
1765: sigprocmask(SIG_BLOCK, $sigset);
1766: $RemoteHost = $host;
1767: Log("CRITICAL", "Forking server for ".$host);
1768: my $pid = fork;
1769: if($pid) { # Parent
1770: $RemoteHost = "Parent";
1771: $ChildPid{$pid} = $host;
1772: sigprocmask(SIG_UNBLOCK, $sigset);
1773: undef(@all_host_ids);
1774: } else { # child.
1775: $RemoteHostId = $hostid;
1776: ShowStatus("Connected to ".$RemoteHost);
1777: $SIG{INT} = 'DEFAULT';
1778: sigprocmask(SIG_UNBLOCK, $sigset);
1779: &ChildProcess(); # Does not return.
1780: }
1781: }
1782:
1783: # parent_client_connection:
1784: # Event handler that processes client connections for the parent process.
1785: # This sub is called when the parent is listening on a socket and
1786: # a connection request arrives. We must:
1787: # Start a child process to accept the connection request.
1788: # Kill our listen on the socket.
1789: # Parameter:
1790: # event - The event object that was created to monitor this socket.
1791: # event->w->fd is the socket.
1792: # Returns:
1793: # NONE
1794: #
1795: sub parent_client_connection {
1796: if ($I_am_child) {
1797: # Should not get here, but seem to anyway:
1798: &Debug(5," Child caught parent client connection event!!");
1799: my ($event) = @_;
1800: my $watcher = $event->w;
1801: $watcher->cancel(); # Try to kill it off again!!
1802: } else {
1803: &Debug(9, "parent_client_connection");
1804: my ($event) = @_;
1805: my $watcher = $event->w;
1806: my $socket = $watcher->fd;
1807: my $connection = $socket->accept(); # Accept the client connection.
1808: Event->io(cb => \&get_remote_hostname,
1809: poll => 'r',
1810: data => "",
1811: fd => $connection);
1812: }
1813: }
1814:
1815: sub get_remote_hostname {
1816: my ($event) = @_;
1817: my $watcher = $event->w;
1818: my $socket = $watcher->fd;
1819:
1820: my $thisread;
1821: my $rv = $socket->recv($thisread, POSIX::BUFSIZ, 0);
1822: Debug(8, "rcv: data length = ".length($thisread)." read =".$thisread);
1823: if (!defined($rv) || length($thisread) == 0) {
1824: # Likely eof on socket.
1825: Debug(5,"Client Socket closed on lonc for p_c_c");
1826: close($socket);
1827: $watcher->cancel();
1828: return;
1829: }
1830:
1831: my $data = $watcher->data().$thisread;
1832: $watcher->data($data);
1833: if($data =~ /\n$/) { # Request entirely read.
1834: chomp($data);
1835: } else {
1836: return;
1837: }
1838:
1839: &Debug(5,"Creating child for $data (parent_client_connection)");
1840: (my $hostname,my $lonid,@all_host_ids) = split(':',$data);
1841: $ChildHost{$hostname}++;
1842: if ($ChildHost{$hostname} == 1) {
1843: &CreateChild($hostname,$lonid);
1844: } else {
1845: &Log('WARNING',"Request for a second child on $hostname");
1846: }
1847: # Clean up the listen since now the child takes over until it exits.
1848: $watcher->cancel(); # Nolonger listening to this event
1849: $socket->send("done\n");
1850: $socket->close();
1851: }
1852:
1853: # parent_listen:
1854: # Opens a socket and starts a listen for the parent process on a client UNIX
1855: # domain socket.
1856: #
1857: # This involves:
1858: # Creating a socket for listen.
1859: # Removing any socket lock file
1860: # Adding an event handler for this socket becoming readable
1861: # To the parent's event dispatcher.
1862: # Parameters:
1863: # loncapa_host - LonCAPA cluster name of the host represented by the client
1864: # socket.
1865: # Returns:
1866: # NONE
1867: #
1868: sub parent_listen {
1869: my ($loncapa_host) = @_;
1870: Debug(5, "parent_listen: $loncapa_host");
1871:
1872: my ($socket,$file);
1873: if (!$loncapa_host) {
1874: $loncapa_host = 'common_parent';
1875: $file = $perlvar{'lonSockCreate'};
1876: } else {
1877: $file = &GetLoncSocketPath($loncapa_host);
1878: }
1879: $socket = &SetupLoncListener($loncapa_host,$file);
1880:
1881: $listening_to{$socket} = $loncapa_host;
1882: if (!$socket) {
1883: die "Unable to create a listen socket for $loncapa_host";
1884: }
1885:
1886: my $lock_file = $file.".lock";
1887: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1888:
1889: my $watcher =
1890: Event->io(cb => \&parent_client_connection,
1891: poll => 'r',
1892: desc => "Parent listener unix socket ($loncapa_host)",
1893: data => "",
1894: fd => $socket);
1895: $parent_dispatchers{$loncapa_host} = $watcher;
1896:
1897: }
1898:
1899: sub parent_clean_up {
1900: my ($loncapa_host) = @_;
1901: Debug(-1, "parent_clean_up: $loncapa_host");
1902:
1903: my $socket_file = &GetLoncSocketPath($loncapa_host);
1904: unlink($socket_file); # No problem if it doesn't exist yet [startup e.g.]
1905: my $lock_file = $socket_file.".lock";
1906: unlink($lock_file); # No problem if it doesn't exist yet [startup e.g.]
1907: }
1908:
1909:
1910:
1911: # This sub initiates a listen on the common unix domain lonc client socket.
1912: # loncnew starts up with no children, and only spawns off children when a
1913: # connection request occurs on the common client unix socket. The spawned
1914: # child continues to run until it has been idle a while at which point it
1915: # eventually exits and once more the parent picks up the listen.
1916: #
1917: # Parameters:
1918: # NONE
1919: # Implicit Inputs:
1920: # The configuration file that has been read in by LondConnection.
1921: # Returns:
1922: # NONE
1923: #
1924: sub listen_on_common_socket {
1925: Debug(5, "listen_on_common_socket");
1926: &parent_listen();
1927: }
1928:
1929: # server_died is called whenever a child process exits.
1930: # Since this is dispatched via a signal, we must process all
1931: # dead children until there are no more left. The action
1932: # is to:
1933: # - Remove the child from the bookeeping hashes
1934: # - Re-establish a listen on the unix domain socket associated
1935: # with that host.
1936: # Parameters:
1937: # The event, but we don't actually care about it.
1938: sub server_died {
1939: &Debug(9, "server_died called...");
1940:
1941: while(1) { # Loop until waitpid nowait fails.
1942: my $pid = waitpid(-1, WNOHANG);
1943: if($pid <= 0) {
1944: return; # Nothing left to wait for.
1945: }
1946: # need the host to restart:
1947:
1948: my $host = $ChildPid{$pid};
1949: if($host) { # It's for real...
1950: &Debug(9, "Caught sigchild for $host");
1951: delete($ChildPid{$pid});
1952: delete($ChildHost{$host});
1953: &parent_clean_up($host);
1954:
1955: } else {
1956: &Debug(5, "Caught sigchild for pid not in hosts hash: $pid");
1957: }
1958: }
1959:
1960: }
1961:
1962: #
1963: # Parent process logic pass 1:
1964: # For each entry in the hosts table, we will
1965: # fork off an instance of ChildProcess to service the transactions
1966: # to that host. Each pid will be entered in a global hash
1967: # with the value of the key, the host.
1968: # The parent will then enter a loop to wait for process exits.
1969: # Each exit gets logged and the child gets restarted.
1970: #
1971:
1972: #
1973: # Fork and start in new session so hang-up isn't going to
1974: # happen without intent.
1975: #
1976:
1977:
1978:
1979:
1980:
1981:
1982: ShowStatus("Forming new session");
1983: my $childpid = fork;
1984: if ($childpid != 0) {
1985: sleep 4; # Give child a chacne to break to
1986: exit 0; # a new sesion.
1987: }
1988: #
1989: # Write my pid into the pid file so I can be located
1990: #
1991:
1992: ShowStatus("Parent writing pid file:");
1993: my $execdir = $perlvar{'lonDaemons'};
1994: open (PIDSAVE, ">$execdir/logs/lonc.pid");
1995: print PIDSAVE "$$\n";
1996: close(PIDSAVE);
1997:
1998:
1999:
2000: if (POSIX::setsid() < 0) {
2001: print "Could not create new session\n";
2002: exit -1;
2003: }
2004:
2005: ShowStatus("Forking node servers");
2006:
2007: Log("CRITICAL", "--------------- Starting children ---------------");
2008:
2009: LondConnection::ReadConfig; # Read standard config files.
2010:
2011: $RemoteHost = "[parent]";
2012: &listen_on_common_socket();
2013:
2014: $RemoteHost = "Parent Server";
2015:
2016: # Maintain the population:
2017:
2018: ShowStatus("Parent keeping the flock");
2019:
2020:
2021: # We need to setup a SIGChild event to handle the exit (natural or otherwise)
2022: # of the children.
2023:
2024: Event->signal(cb => \&server_died,
2025: desc => "Child exit handler",
2026: signal => "CHLD");
2027:
2028:
2029: # Set up all the other signals we set up.
2030:
2031: $parent_handlers{INT} = Event->signal(cb => \&Terminate,
2032: desc => "Parent INT handler",
2033: signal => "INT");
2034: $parent_handlers{TERM} = Event->signal(cb => \&Terminate,
2035: desc => "Parent TERM handler",
2036: signal => "TERM");
2037: $parent_handlers{HUP} = Event->signal(cb => \&KillThemAll,
2038: desc => "Parent HUP handler.",
2039: signal => "HUP");
2040: $parent_handlers{USR1} = Event->signal(cb => \&CheckKids,
2041: desc => "Parent USR1 handler",
2042: signal => "USR1");
2043: $parent_handlers{USR2} = Event->signal(cb => \&UpdateKids,
2044: desc => "Parent USR2 handler.",
2045: signal => "USR2");
2046:
2047: # Start procdesing events.
2048:
2049: $Event::DebugLevel = $DebugLevel;
2050: Debug(9, "Parent entering event loop");
2051: my $ret = Event::loop();
2052: die "Main Event loop exited: $ret";
2053:
2054: =pod
2055:
2056: =head1 CheckKids
2057:
2058: Since kids do not die as easily in this implementation
2059: as the previous one, there is no need to restart the
2060: dead ones (all dead kids get restarted when they die!!)
2061: The only thing this function does is to pass USR1 to the
2062: kids so that they report their status.
2063:
2064: =cut
2065:
2066: sub CheckKids {
2067: Debug(2, "Checking status of children");
2068: my $docdir = $perlvar{'lonDocRoot'};
2069: my $fh = IO::File->new(">$docdir/lon-status/loncstatus.txt");
2070: my $now=time;
2071: my $local=localtime($now);
2072: print $fh "LONC status $local - parent $$ \n\n";
2073: foreach my $host (keys %parent_dispatchers) {
2074: print $fh "LONC Parent process listening for $host\n";
2075: }
2076: foreach my $pid (keys %ChildPid) {
2077: Debug(2, "Sending USR1 -> $pid");
2078: kill 'USR1' => $pid; # Tell Child to report status.
2079: }
2080:
2081: }
2082:
2083: =pod
2084:
2085: =head1 UpdateKids
2086:
2087: parent's SIGUSR2 handler. This handler:
2088:
2089: =item
2090:
2091: Rereads the hosts file.
2092:
2093: =item
2094:
2095: Kills off (via sigint) children for hosts that have disappeared.
2096:
2097: =item
2098:
2099: QUITs children for hosts that already exist (this just forces a status display
2100: and resets the connection retry count for that host.
2101:
2102: =item
2103:
2104: Starts new children for hosts that have been added to the hosts.tab file since
2105: the start of the master program and maintains them.
2106:
2107: =cut
2108:
2109: sub UpdateKids {
2110:
2111: Log("INFO", "Updating connections via SIGUSR2");
2112:
2113: # I'm not sure what I was thinking in the first implementation.
2114: # someone will have to work hard to convince me the effect is any
2115: # different than Restart, especially now that we don't start up
2116: # per host servers automatically, may as well just restart.
2117: # The down side is transactions that are in flight will get timed out
2118: # (lost unless they are critical).
2119:
2120: &KillThemAll();
2121: }
2122:
2123:
2124: =pod
2125:
2126: =head1 Restart
2127:
2128: Signal handler for HUP... all children are killed and
2129: we self restart. This is an el-cheapo way to re read
2130: the config file.
2131:
2132: =cut
2133:
2134: sub Restart {
2135: &KillThemAll; # First kill all the children.
2136: Log("CRITICAL", "Restarting");
2137: my $execdir = $perlvar{'lonDaemons'};
2138: unlink("$execdir/logs/lonc.pid");
2139: exec("$executable");
2140: }
2141:
2142: =pod
2143:
2144: =head1 KillThemAll
2145:
2146: Signal handler that kills all children by sending them a
2147: SIGHUP. Responds to sigint and sigterm.
2148:
2149: =cut
2150:
2151: sub KillThemAll {
2152: Debug(2, "Kill them all!!");
2153:
2154: #local($SIG{CHLD}) = 'IGNORE';
2155: # Our children >will< die.
2156: # but we need to catch their death and cleanup after them in case this is
2157: # a restart set of kills
2158: my @allpids = keys(%ChildPid);
2159: foreach my $pid (@allpids) {
2160: my $serving = $ChildPid{$pid};
2161: ShowStatus("Nicely Killing lonc for $serving pid = $pid");
2162: Log("CRITICAL", "Nicely Killing lonc for $serving pid = $pid");
2163: kill 'QUIT' => $pid;
2164: }
2165: ShowStatus("Finished killing child processes off.");
2166: }
2167:
2168:
2169: #
2170: # Kill all children via KILL. Just in case the
2171: # first shot didn't get them.
2172:
2173: sub really_kill_them_all_dammit
2174: {
2175: Debug(2, "Kill them all Dammit");
2176: local($SIG{CHLD} = 'IGNORE'); # In case some purist reenabled them.
2177: foreach my $pid (keys %ChildPid) {
2178: my $serving = $ChildPid{$pid};
2179: &ShowStatus("Nastily killing lonc for $serving pid = $pid");
2180: Log("CRITICAL", "Nastily killing lonc for $serving pid = $pid");
2181: kill 'KILL' => $pid;
2182: delete($ChildPid{$pid});
2183: my $execdir = $perlvar{'lonDaemons'};
2184: unlink("$execdir/logs/lonc.pid");
2185: }
2186: }
2187:
2188: =pod
2189:
2190: =head1 Terminate
2191:
2192: Terminate the system.
2193:
2194: =cut
2195:
2196: sub Terminate {
2197: &Log("CRITICAL", "Asked to kill children.. first be nice...");
2198: &KillThemAll;
2199: #
2200: # By now they really should all be dead.. but just in case
2201: # send them all SIGKILL's after a bit of waiting:
2202:
2203: sleep(4);
2204: &Log("CRITICAL", "Now kill children nasty");
2205: &really_kill_them_all_dammit;
2206: Log("CRITICAL","Master process exiting");
2207: exit 0;
2208:
2209: }
2210:
2211: sub my_hostname {
2212: use Sys::Hostname;
2213: my $name = &hostname();
2214: &Debug(9,"Name is $name");
2215: return $name;
2216: }
2217:
2218: =pod
2219:
2220: =head1 Theory
2221:
2222: The event class is used to build this as a single process with an
2223: event driven model. The following events are handled:
2224:
2225: =item UNIX Socket connection Received
2226:
2227: =item Request data arrives on UNIX data transfer socket.
2228:
2229: =item lond connection becomes writable.
2230:
2231: =item timer fires at 1 second intervals.
2232:
2233: All sockets are run in non-blocking mode. Timeouts managed by the timer
2234: handler prevents hung connections.
2235:
2236: Key data structures:
2237:
2238: =item RequestQueue
2239:
2240: A queue of requests received from UNIX sockets that are
2241: waiting for a chance to be forwarded on a lond connection socket.
2242:
2243: =item ActiveConnections
2244:
2245: A hash of lond connections that have transactions in process that are
2246: available to be timed out.
2247:
2248: =item ActiveTransactions
2249:
2250: A hash indexed by lond connections that contain the client reply
2251: socket for each connection that has an active transaction on it.
2252:
2253: =item IdleConnections
2254:
2255: A hash of lond connections that have no work to do. These connections
2256: can be closed if they are idle for a long enough time.
2257:
2258: =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