Skip to content

docker/10.5.5

@Jafaral Jafaral tagged this 25 Aug 15:31
Summary of Changes:

*:
    - Fix overflow in comparator functions used by sorted containers

bfdd:
    - Avoid close(-1) in bfd_dplane_finish_late

bgpd:
    - Allow GR-helper stale paths through the evaluate_paths peer-down skip
    - Backport ieee link-bw disable fixes to 10.5
    - Fix route-map use count for EVPN no advertise
    - Fix RD route_node refcount leak in bgp_safi_node_lookup()
    - Fix link-bandwidth AS truncation for extended encoding
    - Fix BAD_COPY_PASTE in bgp_path_info_cmp()
    - Check against NULL for bgp_evpn when handling VNI from Zebra
    - Make sure we do bgp_dest_unlock_node() after bgp_process()
    - Fix link-bandwidth extended community handling above ~34 Gbps
    - Fix crash for `no neighbor X path-attribute discard ...`
    - Don't emit "no neighbor X capability link-local" for unnumbered peers
    - Do not accept AIGP for OAD peers if not enabled
    - Fix bmp connect deletion with source-interface
    - Ignore parsing ORF route-refresh messages
    - Avoid memory leak via EVPN Type-4 nexthop churn
    - Avoid using dangling evpn pointr for bgp_debug_rdpfxpath2str()
    - Fix strlcat/strlcpy size parameter in NOTIFICATION send path
    - Don't advertise LLGR stale routes to non-LLGR peers
    - Only use 32-bit AS in aggregator attr when negotiated
    - Fix nexthop attr validation
    - Fix return NULL in bool function ecommunity_node_target_match
    - Add LLGR to capability length validation switch
    - Fix community string truncation for big community sets
    - Free community json object before regenerating
    - Improve handling of single index in ecom2str()
    - Avoid cluster list attribute truncation
    - Set extended flag for NHC attribute when re-encoding
    - Remove unneeded sort of communities in rmap delete
    - Start BFD hold timer ONLY if hold-time is configured (not zero)
    - Set strict mode flag when we configure BFD hold-timer
    - Fix infinite loop in MRT route dump for oversized paths
    - Skip route-map LPM optimisation for AF_FLOWSPEC
    - Fix use-after-free for ORF case
    - BGPd crash due to multiple bnc entry linked to same peer.
    - Cancel BFD strict hold timer on peer delete
    - Cancel LLGR stale timer on peer AF delete
    - Fix crash of bgp_bmp.c with a mirror situation
    - Validate EVPN Type-5 NLRI prefix length for IPv4
    - Check if sub TLV's length is sufficient for BGP_PREFIX_SID_SRV6_L3_SERVICE_SID_INFO
    - Skip route clearing for peers that were never established
    - Allow `no network ....` form for safi = EVPN or MPLS_VPN
    - Fix EVPN VRF auto RT deletion collision
    - Prevent assert on shutdown in evpn setup
    - Prevent crash when list is already NULL
    - Fix shutdown crash by restricting evpn cleanup to owner instance
    - Fix valgrind memory leaks on daemon shutdown
    - Define a maximum number of TLVs for ENCAP attribute
    - Replace assert() with a boundary check condition
    - Move rpki strict check to bgp_accept()

github:
    - Drop commitlint check

isisd:
    - Consume incomplete Router Capability subTLV header
    - Parse zero-length Router Capability subTLVs
    - Skip only Router Capability bytes on bad subTLV
    - Use subTLV header size in Router Capability parsing
    - Validate ASLA sub-sub-TLV length before consuming bytes
    - Reject duplicate SRv6 SID Structure Sub-Sub-TLV

ldpd:
    - Check fec elem length before accessing
    - Allow event_add_read to work properly on initial

lib:
    - Clear seg6local context when deleting seg6local state
    - Fix ZAPI SRv6 nexthop comparison
    - Fix wrong nexthop comparision for SRv6
    - Bound masklen values, don't assert
    - Test the right bytes in flowspec prefixes
    - Fix NULL deref in frr_signal_timer
    - Add missing hook_unregister_arg in mgmt_be_client_destroy

lib, zebra:
    - Bound SRv6 locator name length in ZAPI

nhrpd:
    - Validate strictly in nhrp_cie_pull() code path
    - Validate more in parse_ether_packet() path
    - Validate header sockunions

ospf6d:
    - Bypass MinLSArrival for self-originated MaxAge LSAs
    - Check length before accessing grace LSA TLVs

ospfd:
    - Validate PREFIX_SID subtlv len before accessing
    - Validate extended prefix TLV before accessing prefix SID
    - Remove unnecessary space

pimd:
    - Use RFC 3973 dm graft retry period for retransmission
    - Fix wrong endian convertion
    - Fix heap OOB write in BSM fragmenter
    - Apply candidate-BSR priority changes directly
    - Remove duplicate PIM_BSM_FRAG mtype declarations
    - Count BSM stats for unicast Candidate-RP messages
    - Free BSR Candidate-RP database on instance teardown
    - Handle BSR unicast socket creation failure
    - Fix candidate-BSR JSON address formatting
    - Fix Candidate-RP group bounds check for PIMv6
    - Fix inverted PIM socket receive failure counter
    - Refresh i_am_rp after BSR RP mapping updates
    - Fix mapping agent AutoRP discovery packet size
    - Guard NULL rp lookup in pim_bsm_clear upstream refresh
    - Guard NULL rp_all when instating BSM pending RP list
    - Guard NULL rp lookup in pim_rp_del upstream refresh
    - Fix crash when up->channel_oil is NULL
    - MLAG: skip pim_register_join on non-DR
    - Cap neighbors and reject PIM when iface has no PIM
    - Provide better ordering for calling pim_upstream_use_rpt
    - Harden BSM group/RP parsing paths
    - Cap PIM Hello secondary address list parsing
    - Reject truncated IP datagrams before IGMP/mtrace handling

tests:
    - Update IS-IS TLV fuzzing data
    - Edit bgp_evpn_rt5_addpath to pass on 10.5 branch
    - Cover EVPN advertise route-map counter cleanup
    - Fixup frr_load_config to be correct for 10.5 branch
    - Check if AIGP attribute is ignored at inbound if not enabled
    - Allow bfdd to be part of the unified config world
    - Add community unit tests
    - Expand PIM candidate BSR/RP topotest coverage
    - Add BSR RP daemon-kill failover topotests
    - Extend BSR fallback topotests to verify iAmRP
    - Add a bgp bmp mirror test

tools:
    - Add show neigh to support bundle generation

yang:
    - Allow match-metric value of zero

zebra:
    - Notify the client when an SRv6 SID allocation fails
    - Don't leave stale interface pointers in the VNI databases
    - Fix heap use after free link interface invalid when ifp freed
    - Fix SRv6 source-address config output
    - Fix neighbor entries ns_id
    - Tear down old L3VNI before adding new one on VNI value change

Signed-off-by: Jafar Al-Gharaibeh <jafar@atcorp.com>
Assets 2
Loading