Prevent underflows in tp->snd_wnd

[jtl]

Summary

We can underflow tp->snd_wnd if the remote side ACKs more than tp->snd_wnd. That can happen when the remote side responds to a window probe by ACKing the one byte it contains.

Obtained from: Juniper Networks (earlier version)
Sponsored by: Juniper Networks
MFC after: 2 weeks

Test Plan

Run a test script that duplicates this behavior; ensure it passes.

[hiren]

Though I generally agree with the overall idea, I am wondering whether we should discuss the larger question we talked about on email: Should the snd_wnd be decremented by number of bytes acked by the other end when we already get window information from the other end?

I'll start a discussion around this on transport@

[hiren]

You are right. Let's get this in while we wait for the better fix.

sys/netinet/tcp_input.c

2805 ↗

(On Diff #14290)

There are 4 sbavail(&so->so_snd) calls in this 'if' block. Do you think we can improve here by making one call and storing that value for later use?

[jtl]

sys/netinet/tcp_input.c

2805 ↗

(On Diff #14290)

sbavail() is an inline function call that is simply return (sb->sb_acc);. So, functionally, I think the compiler already treats these "calls" as simple variable accesses.

[hiren]

Seems okay to me but I'd let someone else also chime in.

sys/netinet/tcp_input.c

2805 ↗

(On Diff #14290)

fair enough if that is the case.

[jtl]

Any further comments/reviews? If I don't receive any, I'll plan to commit this on Monday.