Bug 446112

+++ This bug was initially created as a clone of Bug #205129 +++

Description of problem:
When tcp wrappers try to query a remote ident server, which is blocked (e.g. by
iptables), it leaves SIGALRM blocked. This is especially bad for sshd, because
then whole session then runs with SIGALRM blocked.


Version-Release number of selected component (if applicable):
tcp_wrappers-7.6-37.4


How reproducible:
100%


Steps to Reproduce:
1. on ssh client: "iptables -I INPUT -p tcp --dport ident -j DROP"
2. on ssh server: configure TCP wrappers to do an ident lookup (e.g. add "sshd:
ALL@ALL" line to /etc/hosts.allow)
3a. on ssh client: "ssh user@server 'ps xs|grep $$|grep -v grep'"
or 3b. on ssh client: "ssh user@server", and in the ssh session run something
like this:

perl -e '$SIG{ALRM}=sub{print"ALARM\n";}; alarm 1; sleep 5'


Actual results:
3a: the "BLOCKED" column of SSH output contains SIGALRM (BLOCKED & 0x2000
is 0x2000 on Linux/x86_64 and Linux/i386).
3b: no message is printed.


Expected results:
3a: BLOCKED & 0x2000 should be zero
3b: the "ALARM\n" message should be printed.


Additional info:
In the following message, Wietse Venema suggests that tcp_wrappers code is
correct and the bug is added by third parties:
http://www.gatago.com/mailing/unix/openssh-dev/4854382.html

Debian bug #354855 (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=354855)
apparently contains a patch for this problem. (META: this bugzilla does not have
 Debian bug tracking system available in "External Bug Reference" list).