Chapter 22. General Updates

Some network devices have unacceptably long names. This is due to certain firmware reporting meaningless data, such as the device's onboard index value, which the kernel passes to user-space.

Previously, this resulted in problems with maximum name length, especially with VLANs. With this update, systemd rejects unacceptably long names and falls back to a different naming scheme. As a result, long network device names will no longer appear.

IMPORTANT: This also means that names on existing installations might change, and the affected network devices will not go online.

The change in name will happen on network cards with names enoX where X is more than 16383. This will mostly affect vmware machines, because their firmware has the described problem. (BZ#1230210)

Due to an endianness problem, the version of systemd in Red Hat Enterprise Linux 7.2 read the device identification bytes in a wrong order, causing the dev/disk/by-id/wwn-* symbolic links to be generated incorrectly. A patch has been applied to put the device identification bytes in the correct order and the symbolic links are now generated correctly. Any reference that depends on the value obtained from /dev/disk/by-id/wwn-* needs to be modified to work correctly in Red Hat Enterprise Linux 7.3 and later. (BZ#1308795)

Previously, the default value of the net.unix.max_dgram_qlen kernel option was 16. As a consequence, when the network traffic was too high, certain services could terminate unexpectedly. This update sets the value to 512, thus preventing this problem. Users need to reboot the machine to apply this change. (BZ#1267707)

Red Hat Enterprise Linux 7.2 introduced ldconfig.service, which is run at an early stage of the boot process, before non-root file systems are mounted. Before this update, when ldconfig.service was run, links in the /lib/ and /lib64/ directories were removed if they pointed to file systems which were not yet mounted. In Red Hat Enterprise Linux 7.3, ldconfig.service has been removed, and the problem no longer occurs. (BZ#1301990)

Previously, an inefficient algorithm for reaping processes caused the systemd service to become unresponsive when a large number of processes terminated in a short interval. With this update, the algorithm has been improved, and systemd is now able to reap the processes more quickly, which prevents the described systemd hang from occurring. (BZ#1360160)

When both the 32-bit and 64-bit packages of the gnome-dictionary multilib packages were installed, upgrading from Red Hat Enterprise Linux 7.2 to Red Hat Enterprise Linux 7.3 failed. To fix this problem, the 32-bit package has been removed from Red Hat Enterprise Linux 7.3. As a result, upgrading in this situation works as expected. (BZ#1360338)