Fault Chain

In bgpd, the process accepts a Tunnel Encapsulation attribute (type 23) where the outer TLV (Type-Length-Value) length does not match the actual attribute length. This mismatch causes the reference count of the attribute's interned sub-TLV chain to become unbalanced, leading to the chain being freed while routes in the Adj-RIB-In entries and RIB paths still reference it. When these routes are replaced, withdrawn, or cleared, bgpd accesses the freed memory and attempts to free it a second time. This double-free condition can lead to an assertion failure in memory accounting, causing bgpd to abort and drop all BGP sessions. The issue requires an established BGP session and can be triggered by any neighbor, including eBGP neighbors, unless the attribute is explicitly discarded.

In FRRouting (FRR) 8.3 through 10.5.1, bgp notify decapsulate hard reset() in bgpd/bgp packet.c does not validate the length of the data field of a received BGP NOTIFICATION Cease/Hard Reset message (RFC 8538). When the data field is exactly one byte long, the function reads one byte past the end of a heap buffer and the 16-bit length of the encapsulated data underflows to 65535, causing a 65535-byte out-of-bounds heap read followed by an assertion failure. A remote BGP peer that has established a session and negotiated the Graceful Restart capability with the Notification (N) bit can send such a message to crash the bgpd daemon, terminating all BGP sessions (denial of service). This is fixed in 10.0.5, 10.1.5, 10.2.6, 10.3.4, 10.4.3, 10.5.2 and 10.6.0.