mptcp: Initialize map_seq upon subflow establishment
Subsystems (2)
Christoph Paasch
RFC 1/1 mptcp: Initialize map_seq upon subflow establishment
When the other MPTCP-peer uses 32-bit data-sequence numbers, we rely on map_seq to indicate how to expand to a 64-bit data-sequence number in expand_seq() when receiving data.
For new subflows, this field is not initialized, thus results in an "invalid" mapping being discarded.
Fix this by initializing map_seq upon subflow establishment time.
--- net/mptcp/protocol.c | 2 ++ 1 file changed, 2 insertions(+)
net/mptcp/protocol.c +2 -0
@@ -1635,6 +1635,8 @@ bool mptcp_finish_join(struct sock *sk)list_add_tail (&subflow->node, &msk->join_list);spin_unlock_bh (&msk->join_list_lock); } + + subflow->map_seq = msk->ack_seq;return ret; }
Signature · 1 line
2.23.0
Paolo Abeni
Quoted 20 lines · Christoph Paasch
When the other MPTCP-peer uses 32-bit data-sequence numbers, we rely on
map_seqto indicate how to expand to a 64-bit data-sequence number inexpand_seq()when receiving data.For new subflows, this field is not initialized, thus results in an "invalid" mapping being discarded.
Fix this by initializing
map_sequpon subflow establishment time.Fixes f296234c98a8 ("mptcp: Add handling of incoming MP_JOIN requests")Signed-off-by Christoph Paasch <cpaasch(a)apple.com>--- net/mptcp/protocol.c | 2 ++ 1 file changed, 2 insertions(+)net/mptcp/protocol.c +0
-0
net/mptcp/protocol.c bool mptcp_finish_join(struct sock *sk) · line 1642
Quoted 8 lines
net/mptcp/protocol.c +2
-0@@ -1635,6 +1635,8 @@ bool mptcp_finish_join(struct sock *sk)list_add_tail (&subflow->node, &msk->join_list);spin_unlock_bh (&msk->join_list_lock); } + + subflow->map_seq = msk->ack_seq;return ret; }
The patch LGTM, thanks!
Perhaps we can move the assignment under the 'if (ret) {' conditional, just for clarity?
May I guess the above is the root cause for the retransmissions we get while interopertating with mptcp.org? Is that all, or there are still retransmissions?
Thanks
Paolo
Christoph Paasch
Quoted 32 lines · Paolo Abeni
Quoted 20 lines · Christoph Paasch
When the other MPTCP-peer uses 32-bit data-sequence numbers, we rely on
map_seqto indicate how to expand to a 64-bit data-sequence number inexpand_seq()when receiving data.For new subflows, this field is not initialized, thus results in an "invalid" mapping being discarded.
Fix this by initializing
map_sequpon subflow establishment time.Fixes f296234c98a8 ("mptcp: Add handling of incoming MP_JOIN requests")Signed-off-by Christoph Paasch <cpaasch(a)apple.com>--- net/mptcp/protocol.c | 2 ++ 1 file changed, 2 insertions(+)net/mptcp/protocol.c +0
-0net/mptcp/protocol.c bool mptcp_finish_join(struct sock *sk) · line 1642
Quoted 8 lines
net/mptcp/protocol.c +2
-0@@ -1635,6 +1635,8 @@ bool mptcp_finish_join(struct sock *sk)list_add_tail (&subflow->node, &msk->join_list);spin_unlock_bh (&msk->join_list_lock); } + + subflow->map_seq = msk->ack_seq;return ret; }The patch LGTM, thanks!
Quoted 2 lines
Perhaps we can move the assignment under the 'if (ret) {' conditional, just for clarity?
Yes, I can move it there.
Quoted 3 lines
May I guess the above is the root cause for the retransmissions we get while interopertating with mptcp.org? Is that all, or there are still retransmissions?
Yes, that was all there is. Because data didn't get acknowledged on the other subflows, thus MPTCP retransmitted until it tried the initial subflow, where it finally was being acknowledged.
My next problem now is ADD_ADDR-format that is incorrect in out-of-tree.
On a different note - while digging through the code I found the naming a bit surprising. msk->ack_seq is basically a "TCP rcv_nxt" variable. Same for the subflow's map_seq is the subflow's view of "MPTCP-level rcv_nxt". Right?
Would the code be easier to understand if it would be called msk->rcv_nxt and subflow->rcv_nxt_dsn ?
Not really important or urgent - just wondering if it makes sense to make MPTCP's data-processing more look TCP-like to make it easier to follow as the concepts of Sequence-numbers and ACK-numbers are the same in MPTCP and TCP.
Christoph