tcp: apply rate limits to challenge ACKs

[tuexen]

Summary

When sending challenge ACKs from the SYN-cache, apply the same rate limiting as in other states.

[cc]

sys/netinet/tcp_syncache.c

1658

I think the two new members of struct syncache shall be initialized around here.

2070–2072

Can remove the else clause by initialize the error first. Like this:

static int
syncache_send_challenge_ack(struct syncache *sc, struct mbuf *m)
{
int error = 0;

if (...) { ... }

return (error);
}

Just my two cents if it is just a coding style.

[tuexen]

sys/netinet/tcp_syncache.c

1658

I think the two new members of struct syncache shall be initialized around here.

sc is initialized to zero either using uma_zalloc(..., ...| M_ZERO) or by bzero(). So we don't need to explicitly zero fields. Both new fields are therefore initialized to 0 as intended.

2070–2072

Can remove the else clause by initialize the error first. Like this:

static int
syncache_send_challenge_ack(struct syncache *sc, struct mbuf *m)
{
int error = 0;

if (...) { ... }

return (error);
}

Just my two cents if it is just a coding style.

I actually prefer the style I used. I someone adds another branch, the person has to write an explicit initialization of error. If this is forgotten, it is detected by the compiler or a static code analyzer. If you unconditionally initialize it to 0, this bug is not catched (assuming you want a non-zero error value).
But I am flexible on this, since this is a style question.