Message ID | 20201005100040.1903854-1-matthieu.baerts@tessares.net |
---|---|
State | Accepted, archived |
Commit | c5d72b415ba25ff39479ca97bd8fae0bacf32f0d |
Delegated to: | Matthieu Baerts |
Headers | show |
Series | [mptcp-next] Squash-to: mptcp: keep track of advertised windows right edge | expand |
Matthieu Baerts <matthieu.baerts@tessares.net> wrote: > Move oow messages to debug: I guess we don't want to have this in > pr_info() :) > But maybe we don't want these info at all? Right, pr_info is a bug. I will continue to test a not-yet-sent patch that improves receive window tracking (i.e., what *WE* offered most recently to the peer), probably I can send it tomorrow afternoon. This will add two new variables to msk: the real mptcp-level receive window announced, and the ack sequence number that was used as the reference point. Right now the codes uses ack_seq plus current space, and that is sometimes not in line with what we told the peer. I think its useful to have a pr_debug (then modified to include that information too). This would then allow to detect/debug if peer is really out of window or if we announced a too large window.
Hi Florian, On 05/10/2020 14:28, Florian Westphal wrote: > Matthieu Baerts <matthieu.baerts@tessares.net> wrote: >> Move oow messages to debug: I guess we don't want to have this in >> pr_info() :) >> But maybe we don't want these info at all? > > Right, pr_info is a bug. I will continue to test a not-yet-sent > patch that improves receive window tracking (i.e., what *WE* offered > most recently to the peer), probably I can send it tomorrow afternoon. > > This will add two new variables to msk: the real mptcp-level receive > window announced, and the ack sequence number that was used as the > reference point. Right now the codes uses ack_seq plus current space, > and that is sometimes not in line with what we told the peer. Thank you for your reply! That would be good to be in line with what we told the peer! > I think its useful to have a pr_debug (then modified to include that > information too). As discussed at the last weekly meeting, we decided to apply the current patch and modify it later to include the new variable. - c5d72b415ba2: "squashed" in "mptcp: keep track of advertised windows right edge" - no co-dev/SOB added for that - Results: faedb495716f..5a5a571a738c Tests + export are in progress! Cheers, Matt
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 6b5a8f8d2175..7d81e9444dd3 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -180,7 +180,7 @@ static void mptcp_data_queue_ofo(struct mptcp_sock *msk, struct sk_buff *skb) if (after64(seq, max_seq)) { /* out of window */ mptcp_drop(sk, skb); - pr_info("oow by %ld\n", (unsigned long)seq - (unsigned long)max_seq); + pr_debug("oow by %ld", (unsigned long)seq - (unsigned long)max_seq); MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_NODSSWINDOW); return; }
Move oow messages to debug: I guess we don't want to have this in pr_info() :) But maybe we don't want these info at all? My CI saw a few of these messages. Fixes: 54b7d45eba95 ("mptcp: keep track of advertised windows right edge") Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net> --- net/mptcp/protocol.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)