tcp_bbr: reset full pipe detection on loss recovery undo
authorNeal Cardwell <ncardwell@google.com>
Thu, 7 Dec 2017 17:43:31 +0000 (12:43 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 2 Jan 2018 19:31:07 +0000 (20:31 +0100)
commiteb710b5f62ad3a18742bae70f91e8664ee23cbe3
treed635aeadc92c66e5fa3fe0b05ea6a7e3c2f39636
parentd3f3d4134eb700ad087691effbdf26fe8f3bfd90
tcp_bbr: reset full pipe detection on loss recovery undo

[ Upstream commit 2f6c498e4f15d27852c04ed46d804a39137ba364 ]

Fix BBR so that upon notification of a loss recovery undo BBR resets
the full pipe detection (STARTUP exit) state machine.

Under high reordering, reordering events can be interpreted as loss.
If the reordering and spurious loss estimates are high enough, this
could previously cause BBR to spuriously estimate that the pipe is
full.

Since spurious loss recovery means that our overall sending will have
slowed down spuriously, this commit gives a flow more time to probe
robustly for bandwidth and decide the pipe is really full.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/ipv4/tcp_bbr.c