Description of problem:
When I start an application like OpenSSL, it did start by listening on IPv6 "any" address, now it's listening on the IPv4 specific "any" address.
Version-Release number of selected component (if applicable):
Not sure, included in the RHEL-8.4.0-20201209.n.0 compose
Steps to Reproduce:
1. start openssl s_server
2. ss -ptna | grep openssl
LISTEN 0 128 0.0.0.0:4433 0.0.0.0:* users:(("openssl",pid=7325,fd=3))
LISTEN 0 128 *:4433 *:* users:(("openssl",pid=846593,fd=3))
The bug is s390x specific
Would you please share your /etc/nsswitch.conf file? Is the system host name listed in /etc/hosts or resolvable over DNS? Thanks.
I assume the expected behavior occurs under Red Hat Enterprise Linux 8.3.0.
Actual reproducer commands:
openssl req -nodes -x509 -new -out server.pem -keyout server.key -days 365 -subj "/C=XX/ST=mystate/L=mytown/O=myorganisation/OU=myou/CN=myname/emailAddress=myemail/"
openssl s_server -www -key server.key -cert server.pem
ss -ptna | grep openssl
My test system does not have any link-local IPv6 addresses, so the glibc behavior is as documented. OpenSSL uses AI_PASSIVE | AI_ADDRCONFIG.
ah, so the problem is that openssl is asking from wrong address and the host configuration exposed it
thanks for debugging it, I'm switching the bug to openssl then
bz1908334 is a sister bug for gnutls (same issue, also caused by AI_ADDRCONFIG application).
(In reply to Florian Weimer from comment #3)
> My test system does not have any link-local IPv6 addresses, so the glibc
> behavior is as documented. OpenSSL uses AI_PASSIVE | AI_ADDRCONFIG.
Can you please share more details on the networking for this system?
Which interfaces it has, how are they configured?
(In reply to Hubert Kario from comment #7)
> they don't have IPv6 addresses on public interface, not even link-local
That much I understood. IPv6 public addresses AFAIK are assigned by IT and the
systems will fetch it with dhcpv6. But link-local addresses can be auto-generated
and doesn't depend on anything, so we may have two problems at hand here then.
So it would be helpful if you could describe how the network is configured on these
systems. Is it using networkmanager? Is it, somehow, disabling ipv6 maybe (say, via
(In reply to Hubert Kario from comment #0)
> Actual results:
> LISTEN 0 128 0.0.0.0:4433 0.0.0.0:*
> Expected results:
> LISTEN 0 128 *:4433 *:*
For this, it shouldn't matter specifics of cabled interfaces, as you can have ipv6
enabled only on loopback and it should be enough. Can you please share the output of
'ip addr list' ?
I don't know how they are configured, I didn't do anything special, just asked beaker for s390x machine. Clone my task if you want to see what environment you end up with.
Beniamino asked for a NetworkManager bug regarding the underlying issue, so I filed 1909213.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory (Moderate: openssl security and bug fix update), and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.