Skip to content

Commit

Permalink
tcp: undo spurious timeout after SACK reneging
Browse files Browse the repository at this point in the history
On SACK reneging the sender immediately retransmits and forces a
timeout but disables Eifel (undo). If the (buggy) receiver does not
drop any packet this can trigger a false slow-start retransmit storm
driven by the ACKs of the original packets. This can be detected with
undo and TCP timestamps.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
  • Loading branch information
Yuchung Cheng authored and David S. Miller committed Mar 24, 2013
1 parent 8fe7f99 commit 7ebe183
Showing 1 changed file with 2 additions and 5 deletions.
7 changes: 2 additions & 5 deletions net/ipv4/tcp_input.c
Original file line number Diff line number Diff line change
Expand Up @@ -2059,11 +2059,8 @@ void tcp_enter_loss(struct sock *sk, int how)
if (tcp_is_reno(tp))
tcp_reset_reno_sack(tp);

if (!how) {
/* Push undo marker, if it was plain RTO and nothing
* was retransmitted. */
tp->undo_marker = tp->snd_una;
} else {
tp->undo_marker = tp->snd_una;
if (how) {
tp->sacked_out = 0;
tp->fackets_out = 0;
}
Expand Down

0 comments on commit 7ebe183

Please sign in to comment.