1 LINUX KERNEL REQUIREMENTS
2 =========================
4 The linux kernel has had various major regressions, performance
5 issues and subtle bugs (especially in pmtu). Here is a short list
6 of some -stable kernels that have been tested (at least briefly)
7 and seem to be working well with Quagga/NHRP:
14 [1] But you need to apply the following two backported commits:
15 3cdaa5be9e ipv4: Don't increase PMTU with Datagram Too Big message
16 cb6ccf09d6 route: Use ipv4_mtu instead of raw rt_pmtu
18 See below for list of known issues in various kernel versions.
20 Kernels earlier than 3.12 need CONFIG_ARPD enabled in the configuration.
21 Many distributions do not enable it by default, and you may need to
22 compile your own kernel.
27 DMVPN and mGRE support in the kernel has been brittle. There are various
28 regressions in multiple kernel versions.
30 This list tries to collect them to one source of information:
32 - forward pmtu is disabled intentionally (but tunnel devices rely on it)
33 Broken since 3.14-rc1:
34 commit "ipv4: introduce ip_dst_mtu_maybe_forward and protect forwarding path against pmtu spoofing"
36 Set sysctl net.ipv4.ip_forward_use_pmtu=1
37 (Should fix kernel to have this by default on for tunnel devices)
39 - subtle path mtu mishandling issues
40 Broken since (uncertain)
42 commit "ipv4: Don't increase PMTU with Datagram Too Big message."
43 commit "route: Use ipv4_mtu instead of raw rt_pmtu"
45 - fragmentation of large packets inside tunnel not working
47 commit "ip_tunnels: Use skb-len to PMTU check."
48 Fixed in 3.14.54, 3.18.22, 4.1.9, 4.2-rc3
49 commit "ip_tunnel: fix ipv4 pmtu check to honor inner ip header df"
51 - ipsec will crash during xfrm gc
53 commit "flowcache: Make flow cache name space aware"
55 commit "flowcache: Fix kernel panic in flow_cache_flush_task"
57 - TSO on GRE tunnels failed, and resulted in very slow performance
58 Broke since 3.14.24, 3.18-rc3
59 commit "gre: Use inner mac length when computing tunnel length"
60 Fixed in 3.14.30, 3.18.4
61 commit "gre: fix the inner mac header in nbma tunnel xmit path"
62 commit "gre: Set inner mac header in gro complete"
64 - NAPI GRO handling was broken; causing immediate crash (32-bit only?)
66 commit "net: gro: allow to build full sized skb"
67 Fixed 3.14.5, 3.15-rc7
68 commit "net: gro: make sure skb->cb[] initial content has not to be zero"
70 - ip_gre dst caching broke NBMA GRE tunnels
72 Fixed in 3.14.5, 3.15-rc6
73 commit "ipv4: ip_tunnels: disable cache for nbma gre tunnels"
75 - Few packets can be lost when neighbor entry is in NUD_PROBE state,
76 and there is continuous traffic to it.
77 Broken since dawn of time
79 commit "neigh: probe application via netlink in NUD_PROBE"
81 - GRO was implemented for GRE, but the hw capabilities were not updated
82 correctly. In practice forwarding from non-GRE (physical) interface
83 to GRE interface with gro/gso/tx offloads enabled (also on the target
84 interface) does not work properly.
85 Broken around 3.9 to 3.11, need to check details.
87 - recvfrom() returned incorrect NBMA address, breaking NAT detection
89 commit "GRE: Refactor GRE tunneling code."
90 Fixed in 3.10.27, 3.12.8, 3.13-rc7
91 commit "ip_gre: fix msg_name parsing for recvfrom/recvmsg"
93 - sendto() was broken causing opennhrp not work at all
95 commit "GRE: Refactor GRE tunneling code."
96 Fixed in 3.10.12, 3.11-rc6
97 commit "ip_gre: fix ipgre_header to return correct offset"
99 - PMTU was broken due to GRE driver rewrite
100 Broken since 3.10-rc1
101 commit "GRE: Refactor GRE tunneling code."
103 commit "ip_tunnels: Use skb-len to PMTU check."
105 - PMTU was broken due to routing cache removal
107 commit "ipv4: Cache input routes in fib_info nexthops"
109 commit "ipv4: use next hop exceptions also for input routes"
111 Patches exist for 3.10, but they were not approved to 3.10-stable.
113 - Race condition during bootup: changing ARP flag did not flush
114 existing neighbor entries, causing problems if traffic was routed
115 to gre interface before opennhrp was running.
116 Broken since dawn of time
118 commit "arp: flush arp cache on IFF_NOARP change"
122 commit "xfrm: removes a superfluous check and add a statistic"
124 commit "xfrm: properly handle invalid states as an error"
126 - An incorrect ip_gre change broke NHRP traffic over GRE
128 commit "ip_gre: make ipgre_tunnel_xmit() not parse network header as IP unconditionally"
129 Fixed in 3.8.5, 3.9-rc4
130 commit "Revert "ip_gre: make ipgre_tunnel_xmit() not parse network header as IP unconditionally""
132 - Multicast traffic over mGRE was broken.
133 Broken since 2.6.34-rc2
134 commit "gre: fix hard header destination address checking"
136 commit "net: gre: provide multicast mappings for ipv4 and ipv6"
138 - Serious performance issues causing small throughput on medium to large DMVPN networks
139 Broken since dawn of time
141 multiple commits rewriting ipsec caching
143 - Even though around 2.6.24 is the first version where opennhrp started
144 to work, there has been various PMTU, performance, and functionality
145 bugs before 2.6.34. That's one of the first version I consider stable
146 wrt. to opennhrp functionality.