Infix v24.09.0
News: this release enhances the integration of all types of static routes with FRRouting (Frr), including all routes that can be set by DHCP and IPvLL (ZeroConf) clients. Due to this fundamental change, the system routing table is now primarily read from Frr, which increases the amount of relevant routing information available to the user. E.g., in the CLI exec command show ip route
and show ipv6 route
. Support for adjusting the administrative distance of all types of static routes has also been added to facilitate site specific adaptations. Please see the documentation for details.
Known Issues
- The CLI command
show interfaces
may for some terminal resolutions not display all interfaces (on systems with >20 interfaces). This problem is limited to the console port and only occurs for smaller terminals (30-50 rows height). Callingshow ifaces
from the shell, dumping/ietf-interfaces:interfaces
XPath usingsysrepocfg
, or using the CLI from an SSH session, is not affected. Issue #659
Changes
- Upgrade Buildroot to 2024.02.6 (LTS)
- Upgrade Linux kernel to 6.6.52 (LTS)
- Upgrade libyang to 3.4.2
- Upgrade sysrepo to 2.11.7
- Upgrade netopeer2 (NETCONF) to 2.2.31
- Updated
infix-routing.yang
to declare deviations for unsupported OSPF RPCs and Notifications inietf-ospf.yang
- The CLI admin-exec command
show dns
now also shows any configured name servers, not just ones acquired via DHCP. Issue #510 - Add support for IPv4 (autoconf)
request-address
. This instructs the ZeroConf client to start with the requested address. If this is not successful the client falls back to its default behavior. Issue #628 - Major speedup (10x) in operational data, in particular when querying interface status. Very noticeable in the CLI
show interfaces
command on devices with large port counts. Issue #651 - Silence
yanger
log warnings for failingmctl
command. Caused bymctl
reporting no multicast filtering enabled on bridge
Fixes
- Fix #357: EUI-64 based IPv6 autoconf address on bridges seem to be randomized. Problem caused by kernel setting a random MAC before any bridge port is added. Fixed by using the device's base MAC address on bridge interfaces. Possible to override using
phys-address
option - Fix #601: CLI regression in
show ospf
family of commands causing authorized users, likeadmin
, to not being able to query status of OSPF or BFD. Workaround by using the UNIX shellsudo vtysh
. Regression introduced in v24.08.0 - Fix #603: regression in GNS3 image, starts in test mode by default. Introduced in v24.08.
- Fix #613: CLI regression in tab completion of container commands, e.g.,
container shell <TAB>
. Regression introduced in v24.08.0 - Fix #616: Silent failure when selecting bash as login shell for non-admin user, this silent lock has been removed
- Fix #618: CLI command
show interfaces
does not show bridges and bridge ports, regression introduced in v24.08.0 -- only affects bridges without multicast snooping - Fix #623: CLI command
container upgrade NAME
does not work, regression introduced in v24.06.0 - Fix #625: initialize sysrepo startup datastore at boot. Improves usability when working directly against the sysrepo datastores from the shell with
sysrepocfg
andsysrepoctl
tools - Fix #635: OSPF: all router neighbors reported as neighbor on every interface
- Fix #638: Disabling IPv4LL (autoconf) on an interface does not clean up 169.254/16 addresses
- Fix #640: unable to set static default route due to priority inversion from DHCP or IPv4LL (ZeroConf) clients setting their routes directly in the kernel. This has resulted in a complete overhaul of route management, using FRRouting for all routes, including DHCP and IPv4LL routes, presentation in the CLI, and also support for custom route preference for static routes
- Fix #658: deleting VETH pairs does not work unless rebooting first. Creating a VETH pair, followed by at least one other reconfiguration before removing the pair, causes
confd
to fail when applying the interface changes (tries to delete both ends of the pair) - Spellcheck path to
/var/lib/containers
when unpacking OCI archives on container upgrade - cli: restore
tcpdump
permissions for administrator level users, regression introduced in v24.08.0 - The timeout before giving up on loading the
startup-config
at boot is now 1 minute, just like operations via other front-ends (NETCONF and RESTCONF). This was previously (incorrectly) set to 10 seconds