Show Table of Contents
A
Adjustable TTL limit for caching negative DNS responses in
IP now supports
Linux kernel now contains the
The
Fix to the
Installation of OVN now supports easily-configurable
Chapter 14. Networking
NetworkManager rebased to version 1.8
The NetworkManager package has been upgraded to upstream version 1.8, which provides a number of bug fixes and enhancements over the previous version. Notable changes include:
- Support for additional route options has been added.
- Managed state of device until reboot has been persisted.
- Devices that are externally managed are now correctly handled.
- Networked reliability on multihomed hosts has been enhanced.
- Hostname management is now more flexibly configured.
- Support for changing and enforcing
802-3 link propertieshas been added. (BZ#1414103)
NetworkManager now supports additional features for routes
With this update,
NetworkManager can set some advanced options: source_address (src, IPv4 only), from, type_of_service (tos), window, maximum_transmission_unit (mtu), congestion_window (cwnd), initial_congestion_window (initcwnd), and initial_receiver_window (initrwnd) for static IPv4 and IPv6 routes of connections. (BZ#1373698)
NetworkManager now better handles devices state
With this update,
NetworkManager now maintains the state of devices after the service restart and takes over interfaces which are set into managed mode during restart. In addition, NetworkManager can handle devices which are not explicitly set as unmanaged but controlled manually by the user or another network service. (BZ#1394579)
NetworkManager now supports MACsec (IEEE 802.1AE)
This update adds support for configuring Media Access Control Security (MACsec) encryption into
NetworkManager. (BZ#1337997)
NetworkManager now supports changing and enforcing 802-3 link properties
Previously,
NetworkManager only exposed 802-3 link properties : 802-3-ethernet.speed, 802-3 ethernet.duplex, and 802-3-ethernet.auto-negotiate. With this update, it is possible to change and enforce them. You can either do this automatically using auto-negotiate=yes, or manually using auto-negotiate=no, speed=<Mbit/s>, duplex=[half,full].
Note that if
auto-negotiate=no and either speed or duplex are not set, then the link negotiation is skipped and the auto-negotiate=no, speed=0, duplex=NULL default values are preserved.
Note also that the
auto-negotiate default value has been changed from yes to no to preserve backward compatibility. Previously, the property was ignored, but now an auto-negotiate value of yes can enforce link negotiation. Setting it to no with speed and/or duplex unset means that link negotiation is ignored. (BZ#1353612)
NetworkManager now supports ordering bond slaves based on device names
Previously, the existing order of activation for slave connections could cause problems determining the MAC address of the master interface. This update adds more predictable ordering based on device names. You can enable the new ordering using the
slaves-order=name setting in NetworkManager configuration.
Note that the new ordering is disabled by default and must be explicitly enabled. (BZ#1420708)
NetworkManager now supports VFs for SR-IOV devices
With this update, the
NetworkManager system service supports creating virtual functions (VFs) for Single Root I/O Virtualization (SR-IOV) PCI devices. The number of VFs can be specified using the sriov-num-vfs option in the device section of the NetworkManager configuration file. After VFs are created, NetworkManager can activate connection profiles on them.
Note that some properties of a VF interface, such as the Maximum Transmission Unit (MTU), can only be set to values compatible with those that are set on the physical interface. (BZ#1398934)
Kernel GRE rebased to version 4.8
Kernel Generic Routing Encapsulation (GRE) tunneling has been updated to upstream version 4.8, which provides a number of bug fixes and enhancements over the previous version. The most notable changes include:
- Code merge for transmit and receive paths for IPv4 GRE and IPv6 GRE
- Enhancements that allow link layer address changes without bringing the
gre(IPv4 GRE) orip6gre(IPv6 GRE) device down - Support for various offloads such as
checksum,scatter-gather,highdma,gso, orgro, for IPv6 GRE traffic - Automatic kernel module loading when adding
ip6gretapdevices - Miscellaneous tunneling fixes (such as error handling, MTU calculation, path MTU discovery) up to Linux kernel version 4.8 that affect GRE tunnels (BZ#1369158)
dnsmasq rebased to version 2.76
The dnsmasq packages have been upgraded to version 2.76, which provides a number of bug fixes and enhancements. Notable changes include the following:
- The
dhcp_release6utility is now supported. - The
ra-paramoption has been added. - Support for the
RFC-4242 information-refresh-timeoptions in the reply to the DHCPv6 information request has been added. - The
ra-advroutermode for RFC-3775-compliant mobile IPv6 support has been added. - The
script-arpscript has been added and two new functions for thedhcp-scriptscript have been included. - It is now possible to use random addresses for DHCPv6 temporary address allocations, instead of algorithmically determined stable addresses.
- New optional DNS Security Extensions (DNSSEC) support has been disabled.
BIND changes the way it handles URI resource records, impacting also URI backward compatibility
With this update, the BIND suite no longer adds an additional length byte to a value field when using a URI resource record. This also means that BIND in Red Hat Enterprise Linux (RHEL) 7.4 communicates only in the format described in RFC 7553: https://tools.ietf.org/html/rfc7553.
Note that this update makes new URI records incompatible with records created using BIND in previous versions of RHEL. Namely, BIND in RHEL 7.4 cannot:
- Understand URI records provided by previous versions of BIND in RHEL.
- Serve URI records to clients using previous versions of BIND in RHEL.
However, BIND in RHEL 7.4 still can:
- Cache and receive records from both earlier and future versions of BIND in RHEL.
- Serve records in the old URI format encoded as Unknown DNS Resource Record. See RFC 3597 for details: https://tools.ietf.org/html/rfc3597.
After this update, you do not need to make any change to the DNS zone files. (BZ#1388534)
A DHCP client hook example added for DDNS for Microsoft Azure cloud
An example of the
DHCP client hook for Dynamic DNS (DDNS) for Microsoft Azure cloud has been added to the dhclient package. The administrator can now easily enable this hook, and register Red Hat Enterprise Linux clients with a DDNS server. (BZ#1374119)
dhcp_release6 now releases IPv6 addresses
With this update, the
dhcp_release6 utility can release Dynamic Host Configuration Protocol version 6 (DHCPv6) leases for IPv6 addresses on the local dnsmasq server. See the dhcp_release6(1) man page for more information about the dhcp_release6 command. (BZ#1375569)
Sendmail now supports ECDHE
This update adds the Elliptic Curve Diffie-Hellman Ephemeral Keys (ECDHE) support to Red Hat Enterprise Linux 7
Sendmail. ECDHE is a variant of the Diffie-Hellman protocol that uses elliptic curve cryptography. It is an anonymous key agreement protocol that allows two parties to establish a shared secret over an insecure channel. (BZ#1124827)
telnet now supports the -6 option
Adjustable TTL limit for caching negative DNS responses in Unbound
This update adds the
cache-max-negative-ttl configuration option for the Unbound service, which enables adjustment of the maximum TTL specifically for caching negative DNS responses. Previously, this limit was determined by the domain SOA record, or it was automatically the same as the maximum TTL limit for caching all DNS responses, if configured.
Note that if
Unbound is determining the TTL for DNS response caching, the value set for the cache-min-ttl option has precedence over the value specified by cache-max-negative-ttl. (BZ#1382383)
The scalability of UDP sockets has been improved
This update improves UDP forward memory accounting and reduces the lock contention of UDP sockets. As a result, the overall ingress throughput of UDP sockets receiving traffic from multiple peers is considerably increased without any outward functional changes. (BZ#1388467)
IP now supports IP_BIND_ADDRESS_NO_PORT in the kernel
This update adds the
IP_BIND_ADDRESS_NO_PORT socket option to the kernel. This allows the kernel to skip L4 tuple reservation when a bind() request is used to a port number of 0. As a result, many simultaneous connections to different destination hosts can be maintained. (BZ#1374498)
IPVS Source Hash scheduling now supports L4 hashing and SH fallback
With this update, the IP Virtual Server (IPVS) Source Hash scheduling algorithm includes:
- L4 hashing
- SH fallback of requests to the next active server in case the destination server has a weight of
0, which indicates that the destination server is inactive.
As a result, it is now possible to balance the load of requests from one source IP address based on port numbers. Requests to inactive servers no longer time out. (BZ#1365002)
iproute now supports changing bridge port options
With this update, changing bridge port options such as
state, priority, and cost have been added to the iproute package. As a result, iproute can be used as an alternative to the bridge-utils package. (BZ#1373971)
New options of Sockets API Extensions for SCTP (RFC 6458) implemented
This update implements options
SCTP_SNDINFO, SCTP_NXTINFO, SCTP_NXTINFO and SCTP_DEFAULT_SNDINFO to the Sockets API Extensions for the Stream Control Transmission Protocol (RFC 6458).
These new options replace the options
SCTP_SNDRCV, SCTP_EXTRCV and SCTP_DEFAULT_SEND_PARAM, which are now deprecated. See also the deprecated functionality section. (BZ#1339791)
ss now supports SCTP sockets list
Previously, the
netstat utility provided a list of Stream Control Transmission Protocol (SCTP) sockets. With this update, the ss utility is able to display the same list. (BZ#1063934)
wpa_supplicant rebased to version 2.6
The wpa_supplicant packages have been upgraded to upstream version 2.6, which provides a number of bug fixes and enhancements. Notably, the
wpa_supplicant utility now supports the Media Access Control Security (MACsec) encryption 802.1AE, which enables MACsec to be used in configuration by default. (BZ#1404793, BZ#1338005)
Linux kernel now contains the switchdev infrastructure and mlxsw
This update backports the following functionality into the Linux kernel:
- The Ethernet switch device driver model - the
switchdevinfrastructure; as a result, switch devices can now offload forwarding data plane from the kernel - The
mlxswdriver
Switch hardware supported by
mlxsw :
- Mellanox SwitchX-2 (slow path only)
- Mellanox SwitchIB and SwitchIB-2
- Mellanox Spectrum
Features supported by
mlxsw :
- Per port jumbo frames, speed setting, state setting, statistics
- Port splitting together with splitter cables
- Port mirroring
- QoS: 802.1p, Data Center Bridging (DCB)
- Access Control Lists (ACLs) using TC flower offloading have been introduced as a Technology Preview
Layer 2 features:
- VLANs
- Spanning Tree Protocol (STP)
- Link Aggregation (LAG) using team or bonding offloading
- Link Layer Discovery Protocol (LLDP)
Layer 3 features:
- Unicast routing
To configure all these features, use standard tools provided by the iproute package that has been updated as well. (BZ#1297841, BZ#1275772, BZ#1414400, BZ#1434587, BZ#1434591)
The Linux bridge code rebased to version 4.9
The Linux bridge code has been upgraded to upstream version 4.9, which provides a number of bug fixes and enhancements over the previous version. Notable changes include:
- Support for 802.1ad VLAN filtering and Tx VLAN acceleration
- Support for 802.11 Proxy Address Resolution Protocol (ARP)
- Support for switching offloading by using
switchdev - VLAN support for user
mdbentries - Support for extended attributes in
mdbentries - Support for temporary port router
- Support for per-VLAN statistics
- Support for Internet Group Management Protocol/Multicast Listener Discovery (IGMP/MLD) statistics
- All configuration settings supported by using
sysfsare now supported by netlink as well - Added per-port flag to control the unknown multicast flood (BZ#1352289)
bind-dyndb-ldap rebased to version 11.1
The bind-dyndb-ldap package has been upgraded to upstream version 11.1, which provides a number of bug fixes and enhancements over the previous version.
Notably, the
/etc/named.conf file now uses the new DynDB API. Updating the bind-dyndb-ldap package automatically converts the file to the new API style. (BZ#1393889)
DynDB API from the upstream version 9.11.0 of BIND added to Red Hat Enterprise Linux
This update backports the API for the
dyndb system plug-in, which was introduced in the bind package version 9.11.0 in upstream. As a result, the bind-dyndb-ldap plug-in in Red Hat Enterprise Linux now uses the new API. The downstream feature dynamic_db, which was used in previous releases of Red Hat Enterprise Linux, is no longer supported.
Because the upstream
dyndb uses a different configuration syntax than the downstream dynamic_db, the syntax also changes with this update. However, you do not need to make any manual configuration changes. (BZ#1393886)
tboot rebased to version 1.9.5
The tboot packages have been upgraded to upstream version 1.9.5, which provides a number of bug fixes and enhancements over the previous version. Notable changes include:
- This update adds the 2nd generation of the Link Control Protocol (LCP) creation utility for Trusted Platform Module (TPM) 2.0, as well as a user guide for the updated LCP creation utility.
- A workaround has been implemented to ensure the correct behavior of Intel Platform Trust Technology (PTT) and the Linux PTT driver.
- New fields have been added in the Linux kernel header struct declaration, in order to accommodate for new capabilities of the Linux kernel. (BZ#1384210)
Packages related to rdma consolidated by rebase into rdma-core version 13
The packages related to the rdma package have been upgraded and consolidated into a single source package, rdma-core version 13. The packages are:
- rdma
- iwpmd
- libibverbs
- librdmacm
- ibacm
- libibumad
- libocrdma
- libmlx4
- libmlx5
- libhfi1verbs
- libi40iw
- srp_daemon (formerly srptools)
- libmthca
- libcxgb3
- libcxgb4
- libnes
- libipathverbs
- librxe
- rdma-ndd
The following, previously not included, packages have been added as part of the new package rdma-core:
- libqedr
- libhns
- libvmw_pvrdma
All
ibverbs hardware-specific provider libraries are now bundled in the libibverbs sub-package, streamlining installation and preventing possible versioning mismatches. (BZ#1404035)
OVN IP address management support added for static MAC addresses
This update adds support for dynamic IP address assignment with user-specified static MAC addresses. As a result, Open Virtual Network (OVN) users can now create configurations with dynamic IP that are associated with static MAC addresses. (BZ#1368043)
Enhanced networked reliability on multihomed hosts
On interfaces with a route that is already present on another interface, the NetworkManager utility now automatically switches the reverse path filtering method from
Strict to Loose. This enhances network reliability on multihomed host machines. (BZ#1394344)
Offloading of GENEVE, VXLAN, and GRE tunnels is now supported
With this update, the infrastructure to support offloading of GENEVE, VXLAN, and GRE tunnels has been added. In addition, various bugs have been fixed in the
GENEVE tunnel implementation. (BZ#1326309)
LCO for tunnel traffic is now supported
With this update, the
Local Checksum Offloading (LCO) technique has been added to enable certain network cards to utilize checksum offloading for tunnel traffic. This enhancement improves the performance of VXLAN, GRE, and other tunnels. (BZ#1326318)
Improved tunnel performance on NICs
With this update, tunnel performance on some Network Interface Cards (NICs) that do not support tunnel offloads by default has been enhanced. As a result, users can now take advantage of existing hardware offloads on these NICs. (BZ#1326353)
NPT is now supported in the kernel
With this update, the
IPv6-to-IPv6 Network Prefix Translation (NPTv6) function defined in RFC 6296 has been added in the Netfilter framework. As a result, it is now possible to enable NPT for stateless translation between IPv6 prefixes. (BZ#1432897)
DNS configuration is now supported through the D-Bus API
Previously, external applications could not easily retrieve the
DNS parameters used by NetworkManager. With this update, DNS configuration has been supported through the D-Bus API. As a result, all DNS-related information, including name servers and domains, is available to client applications through the D-Bus API of NetworkManager. An example of such application is the nmcli tool, which can now display DNS configuration. (BZ#1404594)
PPP support is now moved into a separate package
With this update, the Point-to-Point Protocol (PPP) support is moved into a separate, optional NetworkManager-ppp package. As a result, the dependency chain of
NetworkManager is smaller and it is possible to limit the number of installed packages.
Note that to configure PPP settings, you must make sure that the NetworkManager-ppp package is installed. (BZ#1404598)
The tc utility now supports flower
The
tc utility has been enhanced to use the kernel flower traffic control classifier. With this update, a user can add, modify, or delete flower classifier rules from an interface. (BZ#1422629)
Fix to the CRC32c value computation in SCTP forwarding path
Previously, the kernel incorrectly computed the
CRC32c value of Stream Control Transmission Protocol (SCTP) packets with offloaded checksum when the kernel forwarded them to an interface that did not support offloading. This update fixes the computation of CRC32c in the forwarding path. As a result, SCTP packets are now correctly transmitted in the described situation. (BZ#1072503)
New packages: iperf3
This update adds the iperf3 packages version 3.1.7 to Red Hat Enterprise Linux 7. The
iperf3 utility enables active measuring of the maximum achievable bandwidth on IP networks. (BZ#913329)
Installation of OVN now supports easily-configurable firewalld rules
This feature adds
firewalld configuration rules for Open Virtual Network (OVN) to the openvswitch packages. As a result, the user can install easier OVN with firewalld enabled, instead of needing to create firewalld configuration manually. (BZ#1390938)
netlink now supports bridge master attributes
With this update, whenever bridge attributes are changed, a notification is sent out to listeners. This includes changes triggered by sysfs, rtnl, ioctl, or user applications, such as
NetworkManager. (BZ#950243)

Where did the comment section go?
Red Hat's documentation publication system recently went through an upgrade to enable speedier, more mobile-friendly content. We decided to re-evaluate our commenting platform to ensure that it meets your expectations and serves as an optimal feedback mechanism. During this redesign, we invite your input on providing feedback on Red Hat documentation via the discussion platform.