[Bug 824141] New: appending 127.0.0.2 line to /etc/hosts causes postfix to go berseck
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c0 Summary: appending 127.0.0.2 line to /etc/hosts causes postfix to go berseck Classification: openSUSE Product: openSUSE 12.3 Version: Final Platform: Other OS/Version: openSUSE 12.3 Status: NEW Severity: Major Priority: P5 - None Component: YaST2 AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: carlos.e.r@opensuse.org QAContact: jsrain@suse.com Found By: --- Blocker: --- There is a feature request here: https://features.opensuse.org/308824 closed because 11.3 was done. But 12.3 is active and problem remains. The Bugzillas named there are not accessible to me (access denied). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c1 --- Comment #1 from Carlos Robinson <carlos.e.r@opensuse.org> 2013-06-12 11:45:27 UTC --- Some people say that it is postfix fault. This is the answer from the postfix people: +++···································· Date: Wed, 12 Jun 2013 12:36:55 -0400 (EDT) Reply-To: Postfix users <postfix-users@postfix.org> From: wietse@... (Wietse Venema) To: Postfix users <postfix-users@postfix.org> Carlos E. R.:
Apararently, my previous reply has been lost. I resend.
On 2013-06-12 14:40, Wietse Venema wrote:
Carlos E. R.:
Does the machine have a network interface with IP address 127.0.0.2?
Dunno. I guess not, because it is not listed in ifconfig output.
Then, 127.0.0.2 should not be specified in inet_interfaces. Wietse ····································++- Thus, they reject this being a postfix problem. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c2 --- Comment #2 from Aaron Burgemeister <aburgemeister@gmail.com> 2013-06-12 12:10:05 UTC --- Any idea why Postfix is looking in /etc/hosts in the first place? Would binding 127.0.0.2 explicitly fix things? The test about "Does the machine have a network interface with IP address 127.0.0.2" seems to miss the entire point that every machine, per RFC, binds the entire class A 127.x.x.x network to the loopback (lo) address by default, as shown here: Code: ---------- me@mybox:~> ip addr 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 brd 127.255.255.255 scope host lo inet 127.0.0.2/8 brd 127.255.255.255 scope host secondary lo inet6 ::1/128 scope host valid_lft forever preferred_lft forever ---------- The /8 on the end of the IP address means that this answers for everything matching ONLY the first eight bits of the address. As a test, ping any IPv4 address starting with 127 and it will respond immediately and the statistics for 'lo' ('ip -s link') will increase accordingly. Because of this, as well as the proper function of 99% of applications out there, this feels like a bug in Postfix. I still agree that, like in SLE, the default to write the hostname to the loopback address is a bug because it was implemented to work around other products' bugs, but in the end fixing this is missing the point. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c3 --- Comment #3 from Carlos Robinson <carlos.e.r@opensuse.org> 2013-06-12 12:30:41 UTC --- (In reply to comment #2)
Any idea why Postfix is looking in /etc/hosts in the first place? Would binding 127.0.0.2 explicitly fix things?
The test about "Does the machine have a network interface with IP address 127.0.0.2" seems to miss the entire point that every machine, per RFC, binds the entire class A 127.x.x.x network to the loopback (lo) address by default, as shown here:
Well, that surpasses my capacities, you would have to take it with the postfix people (local maintainers or upstream) :-) What I know is that the line: 127.0.0.2 some_host.some_domain some_host in hosts, combined with the line: inet_interfaces = $myhostname, localhost in postfix/main.cf causes this error in postfix: fatal: parameter inet_interfaces: no local interface found for 127.0.0.2 My reading is that "$myhostname" gets translated to some_host.some_domain, and this to 127.0.0.2, and this back to "localhost" (the entire 127 range).
Because of this, as well as the proper function of 99% of applications out there, this feels like a bug in Postfix. I still agree that, like in SLE, the default to write the hostname to the loopback address is a bug because it was implemented to work around other products' bugs, but in the end fixing this is missing the point.
I have read that it also breaks ntp, samba, ldap... the reports that should explain this are closed to me, I have no permissions to read them, so I can't argue those points (per the feature request initial post). On the other hand, I read that sendmail needs this line. What I would propose is a compromise: as there is a tick box to enable/disable this in YaST network configuration module, that tick box should display a text explaining what it does, and what breaks either way, so that if we get hit by the problems we know where they come from and we don't need to google it ;-) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c4 --- Comment #4 from Carlos Robinson <carlos.e.r@opensuse.org> 2013-06-13 14:01:03 UTC --- A good explanation of Postfix behaviour: http://lists.opensuse.org/opensuse/2013-06/msg00550.html -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c5 --- Comment #5 from Aaron Burgemeister <aburgemeister@gmail.com> 2013-06-13 14:57:04 UTC --- I'm afraid I do not understand Postfix's problem yet. Am I correct in assuming the problem comes because Postfix requests the address assigned to the current hostname, and does that so that it can send pings out to destinations with a source IP address specified? If so, why it is: a) Sending pings... I can think of some mediocre reasons, like unreliably determining if a remote system will accept SMTP packets or something, but that should really be using something more applicable than ICMP, like TCP b) Sending with a source IP specified? Why would the system's routing tables not be allowed to do their job in finding a source IP to use when sending something out? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c6 Per Jessen <per@computer.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |per@computer.org --- Comment #6 from Per Jessen <per@computer.org> 2013-06-14 01:02:17 UTC --- (In reply to comment #5)
b) Sending with a source IP specified? Why would the system's routing tables not be allowed to do their job in finding a source IP to use when sending something out?
Because the system's routing tables cannot always be relied upon to do the job correctly :-) For those situations postfix has "smtp_bind_address" and "smtp_bind_address6".
I'm afraid I do not understand Postfix's problem yet. Am I correct in assuming the problem comes because Postfix requests the address assigned to the current hostname, and does that so that it can send pings out to destinations with a source IP address specified?
afaict, postfix tries to identify the interface belonging to an address it has been configured to listen on. (I don't know why). I'm not sure, but it appears that postfix does not take the special properties of the loop-back interface into account. It probably could, but the situation is a bit contrived, imho. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=824141 https://bugzilla.novell.com/show_bug.cgi?id=824141#c FeiXiang Zhang <fxzhang@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |pgajdos@suse.com AssignedTo|bnc-team-screening@forge.pr |varkoly@suse.com |ovo.novell.com | -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com