[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [tinydtls-dev] Feedback trying to use tinyDTLS with Contiki 2.7
|
Is the bug related to my fixes or is it the crash that was present before ?
Regards,
Gaëtan
On 12/07/2016 02:19 PM, Laurent Deru wrote:
It seems I shouted victory a bit too early... On Cooja platform, the
node crashes if the size of DTLS_MAX_BUF is less than 1200, more
precisely sendbuf in dtls_send_multi(). The messages sent and received
are always under 200 bytes however. It looks like a nasty stack
corruption as I can't get a valid stack trace in the debugger, and
when I try to add debug code the behaviour changes completely.
Kind regards,
Laurent.
Le 7/12/16 à 11:37, Gaëtan Harter a écrit :
Hi Laurent,
I'm glad to here that.
It would crash if a message had trouble being delivered and was in
fact retransmitted by 'dtls_retransmit_process'.
By "luck" my border router was mis-configured and not routing big
packets, so the crash was reproducible which helped narrow down the
issue.
Regards,
Gaëtan
On 12/06/2016 02:51 PM, Laurent Deru wrote:
Hi,
Your second fix solves a unexplained random crash we had since a
long time, now all the dtls connections seem stable !
Kind regards,
Laurent.
Le 5/12/16 à 14:32, Gaëtan Harter a écrit :
Hi all developers,
I posted the changes I did on gerrit one commit at a time.
The most important changes are I think:
* The fix for "peer.h" https://git.eclipse.org/r/86360
* The fix for contiki "dtls_retransmit_process"
https://git.eclipse.org/r/86351
These are real issues that makes the code crash without knowing why.
The others are easily fixable when using tinydtls in your
application but prevent it to work out of the box without warning.
Please give me feedback on these changes.
I can rebase any moment if needed.
Regards,
Gaëtan Harter
On 12/01/2016 04:45 PM, Gaëtan Harter wrote:
Hi Raul, developers,
On 11/29/2016 04:13 PM, Raul Fuentes wrote:
Hi Gaëtan,
Now after some time, the client and server get stuck I will
try to
see why.
I have noticed this also with my implementation over RIOT. There
is a point where the server is unable to accept new DTLS sessions
(around the sixth time a node tries to establish a DTLS session
with PSK). My theory is that the DTLS session are not cleaning
properly the memory released from previous DTLS sessions.
For the "getting stuck" status, it was that the first ClientHello
was received by the server but not the second.
The second is a bit bigger than the first one and it was a size
limit.
It was only a problem of my border-router UIP_CONF_BUFFER_SIZE
that was not set and so was not big enough.
Setting it to 1500 fixed everything.
Communication with both PSK and ECC is now working with one client
and one server.
However with ECC, connection is really slow on our hardware so
maybe not really usable in practice.
I did not tested long time usage for the moment however.
For the problems of include I mentioned before, I choose to
include dtls_config.h in tinydtls.h and include tinydtls.h in
almost all headers.
Now it's both working with utash and without.
I will give more feedback if I find new issues or have good news
while using it more.
Regards,
Gaëtan Harter
Sincerely, Raul Fuentes
_______________________________________________
tinydtls-dev mailing list
tinydtls-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or
unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tinydtls-dev
_______________________________________________
tinydtls-dev mailing list
tinydtls-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or
unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/tinydtls-dev
