Bug 1590641

Summary: Unbound sometimes crashes with a buffer overflow
Product: [Fedora] Fedora Reporter: Christian Stadelmann <fedora>
Component: unboundAssignee: Paul Wouters <pwouters>
Status: CLOSED DUPLICATE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 28CC: dominik, pemensik, pj.pandit, pwouters, theo148, thozza
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2018-10-08 17:43:52 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
From syslog none

Description Christian Stadelmann 2018-06-13 06:11:00 UTC
Created attachment 1450794 [details]
From syslog

Description of problem:
Sometimes I can see my network is not available. When I have a look at syslog, I see that unbound has crashed.

Version-Release number of selected component (if applicable):
unbound-1.7.1-1.fc28.x86_64

How reproducible:
very rarely

Steps to Reproduce:
1. unclear

Actual results:
Resolving DNS addresses no longer works. Pinging well-known public IP addresses still works fine. See syslog attached for details.

Expected results:
Resolving DNS addresses should work fine

Additional info:
There is no crash reported to coredumpctl so I cannot use that one to get you a backtrace, sorry. Is this caused by unbound itself? If yes, it would be good to have unbound behave as any other application and leave us a coredump to hunt down bugs.

Comment 1 Dominik 'Rathann' Mierzejewski 2018-06-20 11:36:13 UTC
Same here with unbound-1.7.2-1.fc28.x86_64:
Jun 20 13:16:16 systemd[1]: Starting Unbound recursive Domain Name Server...
Jun 20 13:16:16 unbound-checkconf[1218]: unbound-checkconf: no errors in /etc/unbound/unbound.conf
Jun 20 13:16:16 systemd[1]: Started Unbound recursive Domain Name Server.
Jun 20 13:16:16 unbound[1248]: [1248:0] notice: init module 0: ipsecmod
Jun 20 13:16:16 unbound[1248]: [1248:0] notice: init module 1: validator
Jun 20 13:16:16 unbound[1248]: [1248:0] notice: init module 2: iterator
Jun 20 13:16:16 unbound[1248]: [1248:0] info: start of service (unbound 1.7.2).
Jun 20 13:16:16 unbound[1248]: [1248:0] error: .: failed lookup, cannot probe to master k.root-servers.net
Jun 20 13:16:16 unbound[1248]: [1248:0] error: .: failed lookup, cannot probe to master g.root-servers.net
Jun 20 13:16:16 unbound[1248]: [1248:0] error: .: failed lookup, cannot probe to master f.root-servers.net
Jun 20 13:16:16 unbound[1248]: [1248:0] error: .: failed lookup, cannot probe to master e.root-servers.net
Jun 20 13:16:16 unbound[1248]: [1248:0] error: .: failed lookup, cannot probe to master c.root-servers.net
Jun 20 13:16:16 unbound[1248]: [1248:0] error: .: failed lookup, cannot probe to master b.root-servers.net
Jun 20 13:16:23 unbound[1248]: [1248:3] info: generate keytag query _ta-4a5c-4f66. NULL IN
Jun 20 13:16:24 unbound[1248]: *** buffer overflow detected ***: /usr/sbin/unbound terminated
Jun 20 13:16:24 systemd[1]: unbound.service: Main process exited, code=killed, status=6/ABRT
Jun 20 13:16:24 systemd[1]: unbound.service: Failed with result 'signal'.

Comment 2 Dominik 'Rathann' Mierzejewski 2018-07-02 08:09:09 UTC
Still reproducible with unbound-1.7.3-1.fc28.x86_64 . I'm not sure why abrt isn't catching this. Arguably this is a serious issue because there's no visual indication to the user why DNS (or, for the layman, network connectivity) isn't working. Maybe adding an auto-restart option to the systemd unit would be a good idea as a work-around for the time being.

Comment 3 Tomáš Hozza 2018-07-19 13:40:58 UTC
Happens to me from time to time with unbound-1.7.3-3.fc28.x86_64

Comment 4 Petr Menšík 2018-10-08 17:43:52 UTC
I think this is already reported in bug #1562594. It is hard to distinguish it from the former, but I think this is still the same. Closing as duplicate.

*** This bug has been marked as a duplicate of bug 1562594 ***