[Bug 238478] TCP Cubic code bug in cubic_ack_received

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=238478

--- Comment #4 from Richard Scheffenegger <rscheff at freebsd.org> ---
Actually, concave and convex regions are defined with respect to the previous
cwnd (max_cwnd in the code), not the currently used cwnd.

The conditional in that space is there to prevent cwnd from shrinking, should
it have outgrown the currently calculated w_cubic_next value (e.g. during
slow-start).

So, after a close inspection, showing the cwnd to follow both convex and
concave trajectories properly, I believe the current code is correct. As there
are some style issues, and the comment could be expanded around this, will
commit a patch against this PR nevertheless.

-- 
You are receiving this mail because:
You are on the CC list for the bug.