#28 contiki example dtls-client and dtls-server handshake

[prabhakaran kasinathan]

Hi,

I tried to set up a cooja simulation network to test dtls-handshake between dtls-client and server. I used "wismote" as the target platform and current development release of contiki and tinydtls.

By default the example provided in tinyDTLS doesn't start the handshake. I have to manually started the handshake from tinyDTLS-client by invoking
@~~~~~~
"try_send(dtls_context, &dst);" after a PROCESS_WAIT_EVENT_UNTIL(etimer_expired(&et1));
@~~~~~~
where et1 is set as 20 clock seconds to make sure rpl network is created. Please correct me, if I am doing some things wrongly.

I get following errors such as sequence number mismatches and the handshake is never completed:
@~~~~~~
The message sequence number is too small, expected 2, got: 1
..
retransmit buffer full
@~~~~~~
For more information, pls check the logs:
Cooja Setup: https://drive.google.com/file/d/0B3h2CPasVhruMVhVOUVvZ3ZHSDg/view?usp=sharing
Cooja Log: https://drive.google.com/file/d/0B3h2CPasVhruVl81TVZiX0hGWEU/view?usp=sharing
Cooja Pcap: https://drive.google.com/file/d/0B3h2CPasVhruM2g3ZUpLMEZreDQ/view?usp=sharing

I am not sure if this can be considered as a bug! but any help to solve this would be great.
Thank you!

[prabhakaran kasinathan]

--
update: contiki example DTLS client-server with PSK cipher suite only.
--
Setup1: DTLS-server (node1) / DTLS-client (node2) simulated within cooja.

I disabled DTLS_ECC in dtls_config.h and initiated the handshake from node2.

Result:
Handshake was not successful.

Logs:
Cooja Log and Pcap: https://drive.google.com/folderview?id=0B3h2CPasVhrufmJ6VUtjMzZNakRUV3htUXd6TEdGZldmbzJQMENkRFllbFpSSTlNN2MwcjA&usp=sharing

---

Setup2: Scandium-DTLS client example / DTLS-Server PSK only (node1) simulated within cooja.

Result:
Handshake was successful.

Logs:
Cooja, Scandium Log and Pcaps: https://drive.google.com/folderview?id=0B3h2CPasVhrufmluUk5TZVdlQ2pQRE5obGJJWE5qTXNDd29VYmw2UnpYTWpLaGxjejIyOU0&usp=sharing

[Achim Kraus]

My first impression is:
- The not successful case contains several "retransmission" but no "change cipher spec".
- Line 256 (cooja-tinydtls-pskonly.txt) indicates, that the client (at least) tries to send that "change cipher spec"

00:15.739   ID:2    DEBG send header: (%zu bytes):
00:15.746   ID:2    00000000 14 FE FD 00 00 00 00 00  00 00 04 00 01 
00:15.751   ID:2    DEBG send unencrypted: (%zu bytes):
00:15.753   ID:2    00000000 01

But then the message isn't shown in the pcap nor on the servers receive.
Usually the client should try to retransmit, but the "alert" stops it.

• The successful case doesn't contains "retransmission" but a "change cipher spec".

So my approach to get a session succesful would be to increase the retransmission timeout (dtls.c, dtls_send_multi, values set "n->t" and "n->timeout").

[prabhakaran kasinathan]

HI Achim Kraus,

Thanks for your advice, I did try to increase the retransmission timeout
as you have mentioned, but the results are the same. If you are successful
with DTLS client and server, it would be great to know some more details.

I will work on my free time to solve this issue and will report it if it's
solved.

Thank you for your time!

Best Regards,
Prabhakaran Kasinathan
On Mon, Jun 15, 2015 at 10:20 AM, Achim Kraus boaks@users.sf.net wrote:

My first impression is:
- The not successful case contains several "retransmission" but no
"change cipher spec".
- Line 256 (cooja-tinydtls-pskonly.txt) indicates, that the client (at
least) tries to send that "change cipher spec"

00:15.739 ID:2 DEBG send header: (%zu bytes):00:15.746 ID:2 00000000 14 FE FD 00 00 00 00 00 00 00 04 00 01 00:15.751 ID:2 DEBG send unencrypted: (%zu bytes):00:15.753 ID:2 00000000 01

But then the message isn't shown in the pcap nor on the servers receive.
Usually the client should try to retransmit, but the "alert" stops it.

• The successful case doesn't contains "retransmission" but a
  "change cipher spec".

So my approach to get a session succesful would be to increase the
retransmission timeout (dtls.c, dtls_send_multi, values set "n->t" and
"n->timeout").

---

• [tickets:#28] http://sourceforge.net/p/tinydtls/tickets/28 contiki
  example dtls-client and dtls-server handshake*

Status: open
Milestone: 1.0
Labels: dtls handskake sequence number mismatch
Created: Thu Jun 11, 2015 01:33 PM UTC by prabhakaran kasinathan
Last Updated: Fri Jun 12, 2015 04:04 PM UTC
Owner: Olaf Bergmann

Hi,

I tried to set up a cooja simulation network to test dtls-handshake
between dtls-client and server. I used "wismote" as the target platform and
current development release of contiki and tinydtls.

By default the example provided in tinyDTLS doesn't start the handshake. I
have to manually started the handshake from tinyDTLS-client by invoking
@~~~~~~
"try_send(dtls_context, &dst);" after a
PROCESS_WAIT_EVENT_UNTIL(etimer_expired(&et1));
@~~~~~~
where et1 is set as 20 clock seconds to make sure rpl network is created.
Please correct me, if I am doing some things wrongly.

I get following errors such as sequence number mismatches and the
handshake is never completed:
@~~~~~~
The message sequence number is too small, expected 2, got: 1
..
retransmit buffer full
@~~~~~~
For more information, pls check the logs:
Cooja Setup:
https://drive.google.com/file/d/0B3h2CPasVhruMVhVOUVvZ3ZHSDg/view?usp=sharing
Cooja Log:
https://drive.google.com/file/d/0B3h2CPasVhruVl81TVZiX0hGWEU/view?usp=sharing
Cooja Pcap:
https://drive.google.com/file/d/0B3h2CPasVhruM2g3ZUpLMEZreDQ/view?usp=sharing

I am not sure if this can be considered as a bug! but any help to solve
this would be great.
Thank you!

---

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/tinydtls/tickets/28/

To unsubscribe from further messages, please visit
https://sourceforge.net/auth/subscriptions/

[Achim Kraus]

I compile tinyDTLS using cgcc/ygwin under windows.
And sure, the client successfully connects the server when running under cygwin/posix/windows.

But, as mentioned above, your log indicate that a "very important" message (the cipher change spec) is missing. And that it was sent and never received.

So:
- to which value did you set the timeout? (code sniplet?)
- can you provide us the logs?