Use struct sockaddr_in6 for all addresses #1081
Conversation
krizhanovsky
requested review from
vankoven
and removed request for
krizhanovsky
| * Valid HMAC for 24 bytes (first 2 bytes with AF_INET value, 14 zero bytes of | ||
| * IP address and 8 bytes with timestamp in little-endian order to match HMAC | ||
| * generated in TempestaFW code on x86): | ||
| * Valid HMAC for 36 bytes (2 bytes for AF_INET6, 26 zero bytes for the reset of |
There was a problem hiding this comment.
'for the reset' -> 'for the rest'
tempesta_fw/sock_clnt.c
Outdated
| @@ -404,7 +404,7 @@ tfw_listen_sock_start(TfwListenSock *ls) | |||
|
|
|||
| inet_sk(sk)->freebind = 1; | |||
| sk->sk_reuse = 1; | |||
| r = ss_bind(sk, &addr->sa, tfw_addr_sa_len(addr)); | |||
| r = ss_bind(sk, tfw_addr_sa(addr), tfw_addr_sa_len(addr)); | |||
There was a problem hiding this comment.
The ss_bind() has different actions for AF_INET and AF_INET6 sockets. May be it worth to pass TfwAddr * as parameter for the ss_bind()? Same apply to ss_connect().
There was a problem hiding this comment.
Changed ss_connect() and ss_bind() to accept const TfwAddr *.
ss_bind() now uses only inet6_bind(), but that should be fine, since inet6_bind() recognizes IPv4-mapped addresses and sets IPv4 source addresses just fine.
tempesta_fw/http_limits.c
Outdated
| @@ -265,7 +265,7 @@ int prio0, prio1, prio3; | |||
| #define frang_dbg(fmt_msg, addr, ...) \ | |||
| do { \ | |||
| char abuf[TFW_ADDR_STR_BUF_SIZE] = {0}; \ | |||
| tfw_addr_fmt_v6(&(addr)->v6.sin6_addr, 0, abuf); \ | |||
| tfw_addr_fmt_v6(&(addr)->sin6_addr, 0, abuf); \ | |||
There was a problem hiding this comment.
The macros is very close to the TFW_WARN_MOD_ADDR6() used few lines above excluding TFW_WARN/TFW_DBG difference. It seems pretty generic, shouldn't we put it to the log.h?
There was a problem hiding this comment.
There is no obvious deduplication, I haven't found any other similar places. So I'd rather avoid moving it into log.h, at least in this PR.
tempesta_fw/addr.c
Outdated
| } | ||
| return 0; | ||
|
|
||
| if (tfw_addr_port(listener) != tfw_addr_port(server)) |
There was a problem hiding this comment.
It seems, that the same check with exit is repeated here and in the v4mapped block above. Perhaps it would be better to leave only one this check - before the v4 verification.
There was a problem hiding this comment.
Moved the check to the beginning of the function.
i-rinat
force-pushed
the
ri-775-always-use-ipv6
branch
2 times, most recently
from
cba92f4
to
b75daae
Compare
Drops everything from the TfwAddr union, making it essentially an alias for struct sockaddr_in6. IPv4 addresses are converted to IPv4-mapped addresses at the parsing stage. Macros for log reporting with addresses of both IPv4 and IPv6 are unified. They got a new parameter controlling whenever ports should be reported too.
i-rinat
force-pushed
the
ri-775-always-use-ipv6
branch
from
b75daae
to
1ba513b
Compare
|
Addressed (almost all of) comments. That lead to further changes:
|
IPv4 addresses are represented as IPv4-mapped IPv6.