Bug #77539 for Net-BGP: Net-BGP-0.14 not able to restart a session with a peer after hold timer expires.

Module version: Net-BGP-0.14 Perl version: 5.10.1 OS version: RedHat 6.1 (Linux <my-hostname> 2.6.32-131.0.15.el6.x86_64 #1 SMP Tue May 10 15:42:40 EDT 2011 x86_64 x86_64 x86_64 GNU/Linux) Short Description: It appears it is possible for Net::BGP::Transport to get in a state where old messages from previously closed sessions remain at the front of the message queue (@{%this->{_message_queue}}), resulting in the inability to restart a peer. Below is a description of what I did to trigger the bug and a patch that seems to fix it. Long Description: I have Net::BGP talking to a Quagga bgpd process. My Net::BGP process includes the following code: $peer->add_timer( sub { my $peer = shift; my $peer_id = $peer->peer_id( ); logmsg "$peer_id timer callback\n"; # XXX - We seem to need this to get reconnects # to work. if ( not $peer->is_established( ) ) { logmsg "$peer_id: attempting restart\n"; $peer->start( ); return; } ... }, 60, ); If I send a SIGSTOP to the Quagga bgpd process and wait for the Hold Timer in my Net::BGP process to expire AND for the above timer callback to attempt restarting the connection multiple times before I send a SIGCONT to the Quagga bgpd the session never recovers. It alternates between "OPEN Message Error / Unsupported Version Number" errors and "UPDATE Message Error / Malformed Attribute List" errors. What appears to be happening is that messages are being left in @{$this->{_message_queue}} when Net::BGP::Transport::_close_session() is called. As I understand the code, it does not make sense for @{$this->{_message_queue}} to retain messages across a session close/session restart. Adding the line: $this->{_message_queue} = []; to Net::BGP::Transport::_close_session() fixes the problem, allowing the session to be restarted correctly. Here is a patch: diff -N -u -r Net-BGP-0.14.orig/lib/Net/BGP/Transport.pm Net-BGP-0.14/lib/Net/BGP/Transport.pm --- Net-BGP-0.14.orig/lib/Net/BGP/Transport.pm 2010-09-18 10:12:51.000000000 -0400 +++ Net-BGP-0.14/lib/Net/BGP/Transport.pm 2012-05-30 15:38:45.534481684 -0400 @@ -714,6 +714,7 @@ $this->{_hold_timer} = undef; $this->{_keep_alive_timer} = undef; $this->{_connect_retry_timer} = undef; + $this->{_message_queue} = []; return ( BGP_STATE_IDLE ); }