TinyDTLS integrated to RIOT (without sockets) #5395

rfuentess · 2016-04-25T16:03:40Z

This is a tested example using RIOT and TinyDTLS ( V. 0.8.2 from Eclipse instead of sourceforge).

The TinyDTLS's code had been modified to support the GNRC architecture. This is why the clone of the pkg/tinydtls is with my fork of TinyDTLS.

There is too support for simple sockets, yet the examples are not finished, and are not included in this PR.

OlegHahm · 2016-04-26T09:39:30Z

Awesome! 👍

PeterKietzmann · 2016-04-26T15:52:39Z

I just felt like referencing this PR #4308 here

rfuentess · 2016-04-26T19:33:39Z

Hi @PeterKietzmann, I used that PR as a reference for the structure of the examples, about all the Makefile's. That is why I preserved the list of authors.

I was unable to run the examples from that PR, plus that in January, the TinyDTLS code suffered changes, after emigrating from Sourceforge to Eclipse.

miri64 · 2016-04-27T14:53:08Z

~~Wouldn't it make more sense to

a) still have it as a package, so changes from upstream can be pulled in easier (especially for security-related libs I would argue this still makes a lot of sense)?
b) implement it with conn so other stacks than GNRC could benefit from your work?~~

Sorry, I somehow misstook your examples as lib code...

miri64 · 2016-04-27T15:01:54Z

pkg/tinydtls/tinydtls

@@ -0,0 +1 @@
+Subproject commit 6c479ef140b09e83b79b7e26fad8d36c7bebe442


Please remove this file, it isn't needed

rfuentess · 2016-04-27T15:53:19Z

Hi, @authmillenon Thanks by noticing that my pkg/tinydtls/Makefile was obsolete (and was a good reminder that I need to update my fork).

b) implement it with conn so other stacks than GNRC could benefit from your work?~~

Could be, though to be honest, I selected the GNRC because was the one in the examples, additionally to the sockets. I'm not sure, which one is intended to be THE main stack in RIOT.

rfuentess · 2016-04-27T16:57:59Z

Hi @authmillenon This is rookie question with Git. I already find my error trying to link to a branch instead to a specific commit in the new Makefile. However, I want to concentrated all my work with TinyDTLS/RIOT in one specific branch from my fork (mostly because I need to work with Linux and contiki versions as well).

Is possible to indicated to the new makefile to check a specifically branch ? ( I was experimenting to squash all my commits in one single one in the same branch, but not success)

miri64 · 2016-04-27T17:11:51Z

Could be, though to be honest, I selected the GNRC because was the one in the examples, additionally to the sockets. I'm not sure, which one is intended to be THE main stack in RIOT.

GNRC is intended to be the main stack, but we don't want to bind users to that (conn isn't a stack though, it's just a lightweight application API to a stack [which is actually used to implement sockets]). But anyways, I think for now GNRC-only should suffice for the examples.

Hi @authmillenon This is rookie question with Git. I already find my error trying to link to a branch instead to a specific commit in the new Makefile. However, I want to concentrated all my work with TinyDTLS/RIOT in one specific branch from my fork (mostly because I need to work with Linux and contiki versions as well).

Is possible to indicated to the new makefile to check a specifically branch ? ( I was experimenting to squash all my commits in one single one in the same branch, but not success)

This should be possible, just give the branch name for the PKG_VERSION macro in the makefile (which you already do, as far as I can see, so I'm not sure I understood your problem)

miri64 · 2016-04-27T17:13:02Z

pkg/tinydtls/Makefile

+all: git-download
+	$(MAKE) -C $(PKG_BUILDDIR)
+
+include $(RIOTBASE)/pkg/pkg.mk


Please remove the superfluous newline

rfuentess · 2016-04-27T19:56:40Z

after updating my fork with fetch up I got the following problem.


ruentess@ ~/Projects/RIOT/examples/dtls-echo $ make all 
+ local REMOTE=git://github.com/rfuentess/TinyDTLS.git
+ local SHA1=RIOT-OS
+ basename git://github.com/rfuentess/TinyDTLS.git
+ local REMOTE_NAME=TinyDTLS.git
+ local TARGET_PATH=~/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
+ _check_commit RIOT-OS
+ git clone git://github.com/rfuentess/TinyDTLS.git ~/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
Cloning into '~/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls'...
remote: Counting objects: 279, done.
remote: Total 279 (delta 0), reused 0 (delta 0), pack-reused 278
Receiving objects: 100% (279/279), 260.88 KiB | 133.00 KiB/s, done.
Resolving deltas: 100% (121/121), done.
Checking connectivity... done.
+ git -C ~/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls checkout RIOT-OS
Switched to a new branch 'RIOT-OS'
+ set +ex
~/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls/sha2/sha2.c:35:22: fatal error: tinydtls.h: No such file or directory
 #include "tinydtls.h"

The line + _check_commit RIOT-OS made me think that is not trying to download a branch but instead a specific commit

miri64 · 2016-04-27T20:19:01Z

The line + _check_commit RIOT-OS made me think that is not trying to download a branch but instead a specific commit

of course it checks out a specific commit. A branch is just a reference (in git internal terms: a file named after the branch, containing the object identity (sha-1 sum) of the most recent commit in the branch), so when you checkout a branch you checkout the most recent commit in that branch. What are you trying to do? Regarding your error: You need to adapt your include paths, since the checked out package no resides in another directory:

diff --git a/pkg/tinydtls/Makefile.include b/pkg/tinydtls/Makefile.include
index eb1f442..f56e059 100644
--- a/pkg/tinydtls/Makefile.include
+++ b/pkg/tinydtls/Makefile.include
@@ -1 +1 @@
-INCLUDES += -I$(RIOTBASE)/pkg/tinydtls -I$(RIOTBASE)/pkg/tinydtls/tinydtls
+INCLUDES += -I$(BINDIRBASE)/pkg/$(BOARD)/tinydtls

Also: please rebase if you use the new package scheme (but I guess you already did that just not pushed ;-))

miri64 · 2016-04-27T20:20:41Z

You will realize that a lot of the GNRC API changed ;-)

rfuentess · 2016-04-28T20:32:39Z

Hi @authmillenon,

I upgrade my lines with the new GNRC's API (is more friendly!) but when I'm trying to compile fore testing I got the following:

user@Machine ~/Projects/RIOT/examples/dtls-echo $ make all
Building application "dtls_echo" for "native" with MCU "native".

"make" -C /home/user/Projects/RIOT/pkg/tinydtls
rm -Rf /home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
mkdir -p /home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
/home/user/Projects/RIOT/dist/tools/git/git-cache clone "git://github.com/user/TinyDTLS.git" "RIOT-OS" "/home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls"
+ local REMOTE=git://github.com/user/TinyDTLS.git
+ local SHA1=RIOT-OS
+ basename git://github.com/user/TinyDTLS.git
+ local REMOTE_NAME=TinyDTLS.git
+ local TARGET_PATH=/home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
+ _check_commit RIOT-OS
+ git_cache cat-file -e RIOT-OS^{commit}
+ git -C /home/user/.gitcache cat-file -e RIOT-OS^{commit}
fatal: Cannot change to '/home/user/.gitcache': No such file or directory
+ git clone git://github.com/user/TinyDTLS.git /home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
Cloning into '/home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls'...
remote: Counting objects: 279, done.
remote: Total 279 (delta 0), reused 0 (delta 0), pack-reused 278
Receiving objects: 100% (279/279), 260.88 KiB | 98.00 KiB/s, done.
Resolving deltas: 100% (121/121), done.
Checking connectivity... done.
+ git -C /home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls checkout RIOT-OS
Branch RIOT-OS set up to track remote branch RIOT-OS from origin.
Switched to a new branch 'RIOT-OS'
+ set +ex
if test -d "/home/user/Projects/RIOT/pkg/tinydtls"/patches; then git -C "/home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls" am --ignore-whitespace "/home/user/Projects/RIOT/pkg/tinydtls"/patches/*.patch; fi
touch /home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls/.git-downloaded
make -C /home/user/Projects/RIOT/examples/dtls-echo/bin/pkg/native/tinydtls
..
...

gcc: error: /home/user/Projects/RIOT/examples/dtls-echo/bin/native/gnrc_netif_default.a: No such file or directory
make: *** [all] Error 1

I tried adding the include "net/gnrc/netif.h" to the client and server but I still got that error. Plus, I'm seeing there are many netif in that compilation. Any suggestion to where to look for the problem?

miri64 · 2016-04-28T21:38:14Z

Including a header does not do much to include a module ;-). You have to add it to the USEMODULE list in your application's Makefile, but the module does not exist anymore: it was renamed to gnrc_netdev_default. Since your branch seems to be based on an older version, I think its best if you compare your Makefiles to the Makefile of the gnrc_networking application in current master.

rfuentess · 2016-04-29T17:53:30Z

Many thanks @authmillenon, I just forget the Makefile and didn't expected that after 3 months some packages could not be the same anymore.

Ok, let see if I didn't screwed anything after the rebase.

OlegHahm · 2016-04-29T20:23:14Z

Setting the CI label to have Travis and Murdock have a look.

biboc · 2016-10-24T14:15:37Z

Most of the board are black listed, shouldn't we fix bugs for boards before merging it?

miri64 · 2016-10-24T15:22:01Z

Most of the board are black listed, shouldn't we fix bugs for boards before merging it?

👍

rfuentess · 2016-10-25T07:42:49Z

Most of the board are black listed, shouldn't we fix bugs for boards before merging it?

For me is OK as that is a wise action for RIOT. Though, I'll be unable to work closely on them until finishing the thesis and the defense (March next year).

If this is action is preferred, then maybe is a good idea to close this PR and open a new one when those bugs are fixed.

miri64 · 2016-10-27T11:13:44Z

For now we need a rebase. Also: now that sock is in, consider porting to that (you will find it is a lot more comfortable than conn, also see #6005).

rfuentess · 2016-10-27T14:05:21Z

Sorry by the delay. I hope I had fixed correctly the chaos.

For now we need a rebase. Also: now that sock is in, consider porting to that (you will find it is a lot more comfortable than conn, also see #6005 ).

I was planning to do it after my thesis. But, I had though that conn was intended to be used instead of socket. Oh well, minor thing.

miri64 · 2016-10-27T15:53:38Z

@rfuentess (POSIX) sockets and sock are a different thing. sock is the new application layer API that is about to replace conn and is a lot more comfortable to implement/use.

miri64 · 2016-10-27T16:04:11Z

Here's some reference ;-).

miri64 · 2016-10-27T16:05:55Z

Sorry about the noise :( I now realize that the occurances of conn were only because of the missing rebase in the code :/

miri64 · 2016-10-27T16:18:22Z

@rfuentess Please consider rfuentess#2 to address my comments from April. Other than that I say we can merge.

rfuentess · 2016-10-29T08:23:02Z

@miri64 merged and rebase.

miri64

ACK & go when ~~travis~~ (murdock, I'm tired, okay?) is happy.

miri64

Needs some adaptations to current master

miri64 · 2016-10-29T18:18:21Z

pkg/tinydtls/Makefile

+PKG_NAME=tinydtls
+PKG_URL=git://github.com/rfuentess/TinyDTLS.git
+# PKG_VERSION=RIOT-OS
+PKG_VERSION=f824b5553a865c186a9b41236be03358f0c8feaf 


Trailing whitespace (as reported by Murdock)

miri64 · 2016-10-29T18:22:25Z

examples/dtls-echo/dtls-server.c

+    (void) _dtls_kernel_pid;
+
+    /* Only one instance of the server */
+    if (server.pid != KERNEL_PID_UNDEF) {


Since #5526 was merged this needs to be server.target.pid (and this PR needs a rebase to current master)

and this PR needs a rebase to current master

Just to avoid the conflicts from the previous week. This means to fetch my local repository with the changes from RIOT-OS and then rebase my branch with those changes, right ?

Yes, but just to avoid any confusion here is what you should do

git remote add riot git://github.com/RIOT-OS/RIOT # you might already have done this git fetch riot git rebase -i riot/upstream # here remove any commits that do NOT belong to this PR git push -f

Since you don't have any conflicts in here this should run rather smoothly.

Question, It was riot/master, right ?

Ähh yes. Sorry.

miri64 · 2016-10-29T18:23:46Z

examples/dtls-echo/dtls-server.c

+    dtls_init();
+
+    /* We genarte  */
+    server.pid = thread_create(_server_stack, sizeof(_server_stack),


Use gnrc_netreg_entry_init_pid() to initialize server, please.

miri64 · 2016-10-29T18:23:59Z

examples/dtls-echo/dtls-server.c

+static void stop_server(void)
+{
+    /* check if server is running at all */
+    if (server.pid == KERNEL_PID_UNDEF) {


server.target.pid

miri64 · 2016-10-29T18:24:20Z

examples/dtls-echo/dtls-server.c

+
+    /* stop server */
+    gnrc_netreg_unregister(GNRC_NETTYPE_UDP, &server);
+    server.pid = KERNEL_PID_UNDEF;


server.target.pid

rfuentess · 2016-10-30T11:58:24Z

Ok, fetched riot, rebase with riot/master, added the changes requested, upgraded the client, tested correctly the client and server in native. BUT now the FIT IoT-LAB (greenoble and paris sites) are returning an empty list when ifconfig is used.

Not sure if it is just bad timing with the testbed or something we are forgetting to upgrade on the Makefile.

Edit: Seems is just bad timing with FIT IoT-Lab.

miri64 · 2016-10-30T12:22:55Z

Ok, fetched riot, rebase with riot/master, added the changes requested, upgraded the client, tested correctly the client and server in native. BUT now the FIT IoT-LAB (greenoble and paris sites) are returning an empty list when ifconfig is used.

Not sure if it is just bad timing with the testbed or something we are forgetting to upgrade on the Makefile.

Edit: Seems is just bad timing with FIT IoT-Lab.

Nothing to do with FIT IoT-Lab, nor timings, nor this PR. #5891 just messed up the dependencies for iotlab-m3 a little so the driver for the network device was not in included. Find fix at #6023.

rfuentess · 2016-10-30T13:15:48Z

Tested with available hardware. Already pushed.

Edit: Push again, after the rebase boards nucleo-f070 and nucleo-f030 do not have enough memory.

Support for TinyDTLS (0.8.6) is added together an example at 'examples/dtls-echo'.

miri64 · 2016-10-30T17:57:22Z

Confirmed it to be working. ACK and go as soon as Murdock agrees.

OlegHahm · 2016-10-30T18:20:04Z

Wow! Pretty cool! Nice job, @rfuentess!

kYc0o · 2016-10-31T23:09:53Z

Excellent! Congrats!

OlegHahm assigned kYc0o Apr 26, 2016

OlegHahm added this to the Release 2016.07 milestone Apr 26, 2016

OlegHahm added Area: network Area: Networking Type: new feature The issue requests / The PR implemements a new feature for RIOT labels Apr 26, 2016

OlegHahm mentioned this pull request Apr 27, 2016

Initial import of tinydtls as a package #4308

Closed

miri64 reviewed Apr 27, 2016
View reviewed changes

rfuentess force-pushed the TinyDTLS-GNRC branch 2 times, most recently from b73a90c to b739d21 Compare April 29, 2016 17:52

OlegHahm added the CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR label Apr 29, 2016

rfuentess force-pushed the TinyDTLS-GNRC branch 2 times, most recently from 589b969 to 34329dd Compare May 3, 2016 12:18

kaspar030 added the CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR label Oct 24, 2016

kYc0o removed the CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR label Oct 26, 2016

rfuentess force-pushed the TinyDTLS-GNRC branch from 30a5650 to 106d95e Compare October 27, 2016 14:00

miri64 mentioned this pull request Oct 27, 2016

fix whitespaces rfuentess/RIOT#2

Merged

rfuentess force-pushed the TinyDTLS-GNRC branch from eba6046 to 285f5ca Compare October 29, 2016 08:21

miri64 approved these changes Oct 29, 2016

View reviewed changes

miri64 added the CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR label Oct 29, 2016

miri64 requested changes Oct 29, 2016

View reviewed changes

rfuentess force-pushed the TinyDTLS-GNRC branch from 285f5ca to 8a3ade3 Compare October 30, 2016 13:14

TinyDTLS: integration into RIOT (without sockets)

cf64aba

Support for TinyDTLS (0.8.6) is added together an example at 'examples/dtls-echo'.

rfuentess force-pushed the TinyDTLS-GNRC branch from 8a3ade3 to cf64aba Compare October 30, 2016 17:05

miri64 approved these changes Oct 30, 2016

View reviewed changes

miri64 merged commit 1de2175 into RIOT-OS:master Oct 30, 2016

rfuentess deleted the TinyDTLS-GNRC branch May 19, 2017 09:41

		@@ -0,0 +1 @@
		Subproject commit 6c479ef140b09e83b79b7e26fad8d36c7bebe442

TinyDTLS integrated to RIOT (without sockets) #5395

TinyDTLS integrated to RIOT (without sockets) #5395

Conversation

rfuentess commented Apr 25, 2016 • edited Loading

OlegHahm commented Apr 26, 2016

PeterKietzmann commented Apr 26, 2016

rfuentess commented Apr 26, 2016

miri64 commented Apr 27, 2016 • edited Loading

Choose a reason for hiding this comment

rfuentess commented Apr 27, 2016

rfuentess commented Apr 27, 2016

miri64 commented Apr 27, 2016

Choose a reason for hiding this comment

rfuentess commented Apr 27, 2016

miri64 commented Apr 27, 2016

miri64 commented Apr 27, 2016

rfuentess commented Apr 28, 2016 • edited Loading

miri64 commented Apr 28, 2016

rfuentess commented Apr 29, 2016

OlegHahm commented Apr 29, 2016

biboc commented Oct 24, 2016

miri64 commented Oct 24, 2016

rfuentess commented Oct 25, 2016

miri64 commented Oct 27, 2016

rfuentess commented Oct 27, 2016

miri64 commented Oct 27, 2016

miri64 commented Oct 27, 2016

miri64 commented Oct 27, 2016

miri64 commented Oct 27, 2016

rfuentess commented Oct 29, 2016 • edited Loading

miri64 left a comment • edited Loading

Choose a reason for hiding this comment

miri64 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rfuentess commented Oct 30, 2016 • edited Loading

miri64 commented Oct 30, 2016

rfuentess commented Oct 30, 2016 • edited Loading

miri64 commented Oct 30, 2016

OlegHahm commented Oct 30, 2016

kYc0o commented Oct 31, 2016

rfuentess commented Apr 25, 2016 •

edited

Loading

miri64 commented Apr 27, 2016 •

edited

Loading

rfuentess commented Apr 28, 2016 •

edited

Loading

rfuentess commented Oct 29, 2016 •

edited

Loading

miri64 left a comment •

edited

Loading

rfuentess commented Oct 30, 2016 •

edited

Loading

rfuentess commented Oct 30, 2016 •

edited

Loading