1 | GVPE NEWS |
1 | GVPE NEWS |
2 | |
2 | |
3 | TODO: bridge mode, finally? |
3 | - update embedded liblzf. |
4 | TODO: gcm mode? |
4 | - use liblzf with static hash table: lowers stack usage. |
5 | TODO: replace ripemd160 as the only authentication hash. |
5 | - do not initialise liblzf hash table for faster operation. |
6 | TODO: increase rsa size. |
6 | - use C++ variadic templates for callback.h, get rid of callback.pl. |
7 | TODO: replace transport bits by transport endpoint structs? |
|
|
8 | TODO: ecdh to avoid session replay attacks |
|
|
9 | TODO: http://incog-izick.blogspot.de/2011/08/using-openssl-aes-gcm.html |
|
|
10 | TODO: http://stackoverflow.com/questions/12153009/openssl-c-example-of-aes-gcm-using-evp-interfaces |
|
|
11 | |
7 | |
12 | proposed: 3 types, req, resreq, res |
8 | 3.1 Sat Oct 27 07:50:43 CEST 2018 |
|
|
9 | - port to openssl 1.1 BUT SEE WARNING IN configure.ac. Do not use! |
|
|
10 | - tinc cruft removal: remove getopt.[ch], it's less portable to |
|
|
11 | have it then to not have it. |
|
|
12 | - tinc cruft removal: use pkg-config to detect libressl/openssl. |
|
|
13 | - minor cleanups. |
13 | |
14 | |
14 | req (hmac1) rsa(seqno1 hmac1 aes1 seqno2 hmac2 aes2 auth) ecdh1 |
15 | 3.0 Thu Nov 10 15:39:58 CET 2016 |
15 | res (hmac1) hash(rsa-contents) ecdh2 |
|
|
16 | |
|
|
17 | req hmac0(*) rsa(seqno hmac0 hmac aes auth) hkdf-salt ecdh1 |
|
|
18 | res hmac0(rsa-contents ecdh2) |
|
|
19 | hmac_key = hkdf(hkdf-salt, hmac | ecdh) |
|
|
20 | aes_key = hkdf(hkdf-salt, aes | ecdh) |
|
|
21 | |
|
|
22 | TODO: protocol magic fixen(!!!) |
|
|
23 | TODO: "global" |
|
|
24 | TODO: verify |
|
|
25 | TODO: make sense of overhead calculation |
|
|
26 | TODO: if-up &c should not be scripts? |
|
|
27 | TODO: low-power |
|
|
28 | TODO: ipv6 |
|
|
29 | - INCOMPATIBLE CHANGE: core protocol version 1.0. |
16 | - INCOMPATIBLE CHANGE: core protocol version 1.0. |
30 | - INCOMPATIBLE CHANGE: node sextions are now introduced |
17 | - INCOMPATIBLE CHANGE: node sections are now introduced |
31 | with "node nodename", not "node = nodename". |
18 | with "node nodename", not "node = nodename". |
|
|
19 | - INCOMPATIBLE CHANGE: gvpectrl -g will now generate a single |
|
|
20 | keypair, while -G will try to generate all keypairs as before. |
|
|
21 | - openssl 1.0.2 is the latest supported openssl release, |
|
|
22 | openssl 1.1.0 is not supported at the moment as the work to |
|
|
23 | make it compatible to both versions is just too much. a switch |
|
|
24 | to openssl 1.1 or another library will be done in a future release. |
|
|
25 | - update examples to not generate keys centrally, but locally on each |
|
|
26 | node. |
|
|
27 | - add workaround for temporary/rare ENOBUFS condition. |
32 | - while individual packets couldn't be replayed, a whole session |
28 | - while individual packets couldn't be replayed, a whole session |
33 | could be replayed - this has been fixed by an extra key exchange. |
29 | could be replayed - this has been fixed by an extra key exchange. |
|
|
30 | - fix a delete vs. delete [] mismatch in the central logging function. |
34 | - in addition to rsa key exchange and authentication, the handshake now |
31 | - in addition to rsa key exchange and authentication, the handshake now |
35 | adds a diffie-hellman key exchange (using curve25119) for perfect |
32 | adds a diffie-hellman key exchange (using curve25119) for perfect |
36 | forward secrecy. mac and cipher keys are derived using HKDF. |
33 | forward secrecy. mac and cipher keys are derived using HKDF. |
37 | - rsa key sizes are now configurable and larger (default is 3072). |
34 | - rsa key sizes are now configurable and larger (default is 3072). |
38 | correspondingly, the minimum mtu is no longer 296 but 576. |
35 | correspondingly, the minimum mtu is no longer 296 but 576. |
… | |
… | |
41 | - router reconnects could cause excessive rekeying on other connections. |
38 | - router reconnects could cause excessive rekeying on other connections. |
42 | - gvpectrl no longer generates all missing public keys, but |
39 | - gvpectrl no longer generates all missing public keys, but |
43 | only missing private keys. private keys are also put |
40 | only missing private keys. private keys are also put |
44 | into the configured location. |
41 | into the configured location. |
45 | - the pid-file now accepts %s as nodename as elsewhere. |
42 | - the pid-file now accepts %s as nodename as elsewhere. |
|
|
43 | - switch to counter mode (only aes supported at the moment in |
|
|
44 | openssl). this gets rid of the need to generate a random iv, |
|
|
45 | is likely more secure (and, as a side effect, gets rid of |
|
|
46 | slow randomness generation. counter mode is often faster |
|
|
47 | then cbc mode as well, and packets are smaller). |
46 | - no longer use RAND_bytes to generate session keys - you NEED |
48 | - no longer use RAND_bytes to generate session keys - you NEED |
47 | a real source of entropy now (e.g. egd or /dev/random - see the |
49 | a real source of entropy now (e.g. egd or /dev/random - see the |
48 | openssl documentation). |
50 | openssl documentation). |
49 | - multiple node statements for the same node are now supported |
51 | - multiple node statements for the same node are now supported |
50 | and will be merged. |
52 | and will be merged. |
51 | - a new directive "global" switches back to the global section |
53 | - a new directive "global" switches back to the global section |
52 | of the config file. |
54 | of the config file. |
53 | - 12 random prefix bytes are now properly supported, leading to |
|
|
54 | a fully random IV. |
|
|
55 | - use aes with a random key in counter-mode to generate IVs, |
|
|
56 | for speed reasons (generating 12 random bytes with openssl can |
|
|
57 | take longer than encrypting and mac'ing a 1.5kb packet, and IVs |
|
|
58 | do not need to be cryptographically strong random numbers |
|
|
59 | (and in fact, shouldn't be)). |
|
|
60 | - if-up scripts can now be specified with absolute paths. |
55 | - if-up scripts can now be specified with absolute paths. |
61 | - new global option: serial, to detect configuration mismatches. |
56 | - new global option: serial, to detect configuration mismatches. |
62 | - use HKDF as authentication proof, not HMAC or a plain hash |
57 | - use HKDF as authentication proof, not HMAC or a plain hash |
63 | (hint by Ilmari Karonen). |
58 | (hint by Ilmari Karonen). |
64 | - during rekeying or conenction establishments, hmac authentication |
59 | - during rekeying or connection establishments, hmac authentication |
65 | errors could occur and reset the connection. Transient hmac |
60 | errors could occur and reset the connection. Transient hmac |
66 | authentication errors are now being ignored for 3 seconds. |
61 | authentication errors are now being ignored for 3 seconds. |
67 | - log the reason for a conneciton loss. |
62 | - log the reason for a conneciton loss. |
68 | - use a (hopefully) constant time memcmp to compare internal secrets. |
63 | - use a (hopefully) constant time memcmp to compare internal secrets. |
|
|
64 | - fix a (harmless) errornous out of bounds stack read that would trigger |
|
|
65 | gcc's -fsanitize=address. |
|
|
66 | - bump old packet window size from 512 to 65536. |
|
|
67 | - update for big changes in openssl 1.1 API, wrap primitives |
|
|
68 | to make further changes easier. |
|
|
69 | - correctly check return values for openssl 1.0.0 and later. |
|
|
70 | - check for both public and private key file when deciding whether |
|
|
71 | to skip generating a key to avoid accidental overwrites. |
69 | |
72 | |
70 | 2.25 Sat Jul 13 06:42:33 CEST 2013 |
73 | 2.25 Sat Jul 13 06:42:33 CEST 2013 |
71 | - INCOMPATIBLE CHANGE: no longer enable udp protocol if no other |
74 | - INCOMPATIBLE CHANGE: no longer enable udp protocol if no other |
72 | protocols are enabled - this is necessary when you have nodes with |
75 | protocols are enabled - this is necessary when you have nodes with |
73 | completely unknown protocols, to force mediated connection requests. |
76 | completely unknown protocols, to force mediated connection requests. |