lib: fix zclient crash when many peers reconnect after FRR restartReplace select() with poll() in zclient_read_sync_response() to avoid
FD_SETSIZE (1024) limit. When the zebra socket fd exceeds 1024, FD_SET()
triggers __fdelt_warn and aborts. poll() has no such limit.
Handle POLLNVAL, POLLHUP, and POLLERR explicitly to avoid busy-spin on
socket errors, matching the pattern from lib/event.c.
Backtrace before fix:
```
Mar 09 22:16:42 core frrinit.sh[898790]:...
tests: Add BGP-LS topotest suiteAdd pytest topology test for BGP Link-State covering producer and
consumer modes with ISIS IGP backend.
Tests BGP-LS capability negotiation, NLRI advertisement/reception
(Node, Link, Prefix), and dynamic topology change propagation
including route additions/removals and link state changes.
Verifies BGP-LS updates flow correctly from IGP through BGP-LS
producer to BGP-LS consumer.
Signed-off-...
bgpd: Set nexthop for BGP-LS based on peer address familyBGP-LS routes advertised to peers require a nexthop with the same
address family as the peer session (IPv4 peer needs IPv4 nexthop,
IPv6 peer needs IPv6 nexthop).
The MP_REACH_NLRI and MP_UNREACH_NLRI encoding code in bgp_attr does
not have direct knowledge of the peer's address family, making it
unable to select the appropriate nexthop address family for BGP-LS
advertisements.
Add peer flags...
bgpd: Parse BGP-LS NLRI received from peersAdd inbound BGP-LS message processing to receive Link-State topology
information from BGP peers.
When BGP receives a BGP UPDATE message from a peer, it calls
bgp_nlri_parse() to parse the NLRI. The bgp_nlri_parse() function
checks the SAFI (Subsequent Address Family Identifier) contained in
the UPDATE message and dispatches to the appropriate SAFI-specific
handler. Currently supported SAFIs in...
bgpd: Unregister from lsdb on terminationCall bgp_ls_unregister() in bgp_ls_terminate() to cleanly unregister
from the zebra link-state database when BGP instance terminates.
This ensures proper cleanup and stops receiving IGP topology updates
when the BGP-LS functionality is no longer needed.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Receive BGP-LS attributeAdd bgp_attr_ls() function to parse and intern BGP-LS attributes (Type 29)
from received BGP update messages. The function validates peer configuration,
parses Type 29 TLVs, and interns the resulting attribute for use in route
processing.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Encode BGP-LS NLRI and attribute when building UPDATE packetAdd BGP-LS encoding support to UPDATE message generation for
transmitting Link-State topology information to peers.
Implement bgp_packet_ls_attribute() to encode BGP-LS Attribute (Type 29)
by dispatching to type-specific encoders (node/link/prefix attributes).
Extend packet building functions to accept ls_nlri parameter:
- bgp_packet_attribute() - encode BGP-LS attribute when present
- bgp_pa...
bgpd: Add BGP-LS AFI/SAFI to negotiation functionsAdd BGP-LS capability checks to peer activity and address family
negotiation functions.
Update peer status and capability negotiation functions to include
BGP-LS AFI/SAFI:
- peer_active() - Check if peer has BGP-LS configured
- peer_active_nego() - Check if BGP-LS was negotiated with peer
- peer_group_af_configured() - Check if peer group has BGP-LS enabled
- bgp_open_option_parse() - Validat...
bgpd: Add BGP-LS capability display supportAdd display support for BGP-LS multi-protocol capability in BGP OPEN
message output.
Update bgp_capability_vty_out() to properly display BGP-LS capabilities:
- Display "AFI BGP-LS" for AFI_BGP_LS (16388)
- Display "SAFI BGP-LS" for SAFI_BGP_LS (71)
- Add JSON output support for BGP-LS capabilities
This allows users to verify BGP-LS capability negotiation between peers
using "show bgp neighbo...
bgpd: Register for linkstate db updates when peer is activatedRegister for Link State database updates when the first BGP-LS peer is
activated, and unregister when the last peer is deactivated.
Modify peer_activate() and peer_deactivate() to manage Link State
database registration based on active BGP-LS peers:
- When first peer is activated for link-state address family, call
bgp_ls_register() to subscribe to IGP topology updates from zebra
- When las...
bgpd: Add show commands for BGP-LS NLRIsAdd commands to display BGP Link-State NLRIs from the BGP RIB.
New commands:
- show bgp link-state link-state [json]
Display all BGP-LS routes with node/link/prefix NLRIs
- show bgp link-state link-state NLRI [json]
Display specific BGP-LS route by NLRI identifier
NLRI format: [Type][ID][Protocol][Instance][AS][Area][Descriptors...]
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add BGP-LS configuration write supportExtend bgp_config_write() to serialize BGP Link-State address family
configuration.
This enables BGP-LS configuration to be:
- Displayed via show running-config
- Saved to startup-config (write memory)
- Written to configuration files
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
lib: Add CLI node for BGP Link-State address family configurationAdd BGP_LS_NODE to enum node_type in command.h to enable BGP Link-State
address family configuration mode.
This allows creating a dedicated configuration context for BGP-LS:
router bgp 65000
address-family link-state
<BGP-LS specific commands>
exit-address-family
Follows the same pattern as other BGP address families (BGP_EVPN_NODE,
BGP_SRV6_NODE).
Signed-off-by: Carmine Sca...
bgpd: Handle zebra opaque messagesAdd callback handler to process ZAPI Opaque messages from zebra containing
Link State updates from IGP producers.
Register bgp_zebra_opaque_msg_handler() as the callback for
ZEBRA_OPAQUE_MESSAGE type in the zebra client handler table. The handler:
- Decodes ZAPI Opaque message header using zclient_opaque_decode()
- Routes Link State messages (LINK_STATE_SYNC, LINK_STATE_UPDATE) to
bgp_ls_pr...
bgpd: Add functions to parse updates received from linkstate dbAdd callback functions to parse and process Link State updates received
from IGP producers via ZAPI Opaque messages.
When IGPs (IS-IS, OSPF) detect topology changes, they send Link State
Update messages through zebra to registered consumers. These functions
handle the received messages:
- bgp_ls_parse_update() - Parse ZAPI Opaque Link State messages
- bgp_ls_process_node() - Process Node upda...
bgpd: Add BGP-LS update and withdraw functionsImplement BGP-LS route installation and withdrawal to enable processing
of BGP-LS NLRIs received in BGP UPDATE messages.
In BGP-LS, UPDATE messages contain BGP-LS NLRIs (not traditional prefixes).
When a BGP UPDATE with BGP-LS NLRI is received, a bgp_dest is allocated
to represent that NLRI location in the RIB and stored in the BGP RIB tree.
The ls_nlri field in bgp_dest stores the BGP-LS NLRI...
bgpd: Add BGP-LS Attribute support to attr structureExtend the core BGP attribute structure to support BGP-LS Attributes
(RFC 9552 Section 5.3) by adding ls_attr field and updating all related
attribute processing functions.
Changes:
- Add ls_attr field to struct attr to store BGP-LS Attribute TLVs
- Update attrhash_key_make() to include ls_attr in attribute hash key
- Update attrhash_cmp() to compare ls_attr when present
- Extend bgp_attr_inte...
bgpd: Add helper functions to register linkstate db notificationsAdd helper functions to register and unregister BGP-LS for Link State
database notifications from IGP producers.
In FRR's Link State architecture, consumers must register with zebra to
receive ZAPI Opaque Link State Update messages from producers (IS-IS, OSPF).
These functions wrap the Link State API registration mechanism:
- bgp_ls_register() - Register to receive LS updates from IGPs
- bgp_...
bgpd: Add TED instance to BGP-LS data structureAdd Traffic Engineering Database (TED) instance to BGP-LS structure to
store link-state topology.
In FRR's Link State architecture, IGP protocols (IS-IS, OSPF) act as
producers that build link-state databases from protocol exchanges and
send updates via ZAPI Opaque messages. BGP-LS acts as a consumer that
synchronizes with IGPs and listens for updates, maintaining a TED with
the topology infor...
bgpd: Add BGP-LS NLRI encode functionsAdd functions to serialize BGP-LS NLRIs to wire format per RFC 9552.
The encode functions convert in-memory NLRI structures to the TLV-based
wire format used in MP_REACH_NLRI and MP_UNREACH_NLRI path attributes.
- bgp_ls_encode_node_descriptor() - Encode Node Descriptors
- bgp_ls_encode_link_descriptor() - Encode Link Descriptors
- bgp_ls_encode_prefix_descriptor() - Encode Prefix Descriptors...
bgpd: Add error code for BGP-LS packet errorsAdd EC_BGP_LS_PACKET error code for BGP-LS NLRI and attribute packet
parsing/encoding errors.
This error code is used by BGP-LS decode functions to log warnings when
malformed packets are received from peers, such as invalid TLV lengths,
missing mandatory fields, or unknown NLRI types.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add hash table to store BGP-LS NLRIsAdd hash table infrastructure to store BGP-LS NLRIs originated by the
local node and received from peers.
- Add nlri_hash to bgp_ls structure for NLRI storage
- Add hash functions to compute hash keys and compare NLRIs
- Add bgp_ls_nlri_intern() to insert NLRI and increment reference count
- Add bgp_ls_nlri_unintern() to remove NLRI and decrement reference count
- Add bgp_ls_entry_lookup() to ...
bgpd: Add BGP-LS NLRI deep copy functionAdd function to create deep copy of BGP-LS NLRI structures.
Used in subsequent commits for NLRI storage and management.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add BGP-LS NLRI size calculation helperAdd function to calculate wire format size of BGP-LS NLRI structures.
Used in subsequent commits for NLRI encoding.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add BGP-LS NLRI validation functionsAdd validation functions to verify BGP-LS NLRI is valid per RFC 9552.
Used in subsequent commits to validate NLRIs received from peers.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add BGP-LS NLRI comparison functionsAdd functions to determine if two BGP-LS NLRIs are equal.
These functions will be used in subsequent commits for NLRI hash table
lookups and duplicate handling.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add BGP-LS instance structure and initializationIntroduce core BGP-LS module infrastructure.
- Add `struct bgp_ls` to hold BGP-LS state
- Add `bgp_ls_init()` to initialize BGP-LS module during BGP instance
creation
- Add `bgp_ls_terminate()` to cleanup BGP-LS module during BGP instance
destruction
Subsequent commits will extend this structure with NLRI and TED support.
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
bgpd: Add debugging support for BGP-LSAdd debugging support for BGP Link-State operations to aid development
and troubleshooting.
- Add `debug bgp link-state` command
- Add `no debug bgp link-state` command
- Extend `no debug bgp` to also disable link-state debugging
- Show link-state debugging status in `show debugging bgp`
Signed-off-by: Carmine Scarpitta <cscarpit@cisco.com>
*: Add BGP-LS AFI/SAFI constantsAdd BGP Link-State (BGP-LS) Address Family infrastructure to prepare
for full protocol implementation in subsequent commits (RFC 9552).
Introduce AFI_BGP_LS (16388) and SAFI_BGP_LS (71) and update all
AFI/SAFI-aware code paths to recognize the new address family.
Protocol-specific logic is marked with TODO placeholders to be
implemented in subsequent commits.
Signed-off-by: Carmine Scarpitta...
bgpd: avoid premature memory allocation in subgroup_announce_check()The subgroup_announce_check() function had memory allocation operations
placed before several early return checks. This could cause memory leaks
when the function returned early after memory had already been allocated.
Reorganized the code to ensure all validation checks that might return
early are performed before any memory-allocating operations.
Signed-off-by: Enke Chen <enchen@paloaltonet...
Jafar Al-Gharaibeh
Nick Bouliane
Carmine Scarpitta
Chirag Shah
Enke Chen