Generate OSCORE object and use it for endpoint creation #736

rikard-sics · 2019-09-12T10:05:39Z

Code to implement the points listed here:
#718 (comment)

create the OSCORE object
use it in leshan client demo (keeping hardcoded value but create oscore object)
modify ServersInfoExtractor and ServerInfo to support oscore.
use ServerInfo in CaliforniumEndpointsManager to create endpoint with oscore support.

The OSCORE object is generated and filled in the Leshan client demo. It will then be used to create a ServerInfo object using the ServersInfoExtractor. Next the ServerInfo is used in the CaliforniumEndpointsManager to generate endpoints towards servers that OSCORE should be used with.

See also issue regarding OSCORE Leshan client integration:
#726

Code to implement the points listed here: eclipse-leshan#718 (comment) The OSCORE object is generated and filled in the Leshan client demo. It will then be used to create a ServerInfo object using the ServersInfoExtractor. Next the ServerInfo is used in the CaliforniumEndpointsManager to generate endpoints towards servers that OSCORE should be used with. See also issue regarding OSCORE client integration: eclipse-leshan#726 Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2019-09-12T10:08:11Z

One issue I had was how exactly to link the Security and Oscore objects together. I now see your suggestion here:
#726 (comment)

So having an instance ID in the Oscore object and then also in the Security object could bind them together?

rikard-sics · 2019-09-12T10:12:16Z

Some outstanding issues:
What is an appropriate identity when using OSCORE. Currently I left the following code.
serverIdentity = Identity.unsecure(serverInfo.getAddress()); //TODO: FIX?

Currently the HashMapCtxDB is used as a singleton:
HashMapCtxDB db = HashMapCtxDB.getInstance(); //TODO: Do not use singleton here but give it to endpoint builder (for Cf-M16)
When we update to Californium 2.0.0-M16 or higher this can be changed. Then the db is instead tied to a specific endpoint.

Lastly I see I forgot to apply the formatter. I can do this before final merge.

sbernard31 · 2019-09-17T12:48:13Z

leshan-client-core/pom.xml

+                <groupId>org.eclipse.californium</groupId>
+                <artifactId>cf-oscore</artifactId>
+                <version>${californium.version}</version>
+        </dependency>


leshan-client-code must not depend to californium. (see https://github.com/eclipse/leshan/wiki/Modules#leshan-client-core)

That means that you can not really used AlgorithmID in this module... :/
In a first time manipulating just the value should be OK and you will see if we find a better way.

I see, thanks for the link. I will amend the code so leshan-client-core does not have any dependencies from Californium. I can probably keep it as an int instead of AlgorithmID.

This should now be taken care of in commit 8fc7bfc
I needed the hex2ByteArray method from StringUtil in Californium. For now I extracted and took the code from there. If that is not fine I can re-implement it.

maybe you could find what you need in org.eclipse.leshan.util.Hex ?

maybe you could find what you need in org.eclipse.leshan.util.Hex ?

Yes, I will check that.

Update to use a method from the Leshan Hex class rather than the imported method from Californium in commit ebc728a

sbernard31

So having an instance ID in the Oscore object and then also in the Security object could bind them together?

Yep, see my comments below.

What is an appropriate identity when using OSCORE.

Good question !

in PSK this is the pskIdentity (login)
in RPK this is the public key
in X509 we use the CN of the certificate (which is debatable)

I suppose we should add a new attribute for OSCORE which could be used in addition of existing DTLS identity (in case of OSCORE over coaps) or without it (in case of OSCORE over coap).

In oscore what could be the value which could be used as an identifier ? I need to read the rfc again (intuitively I suppose that Sender ID or/and Recipient ID could be candidate?)

then the db is instead tied to a specific endpoint.

👍

I see I forgot to apply the formatter. I can do this before final merge.

ok

sbernard31 · 2019-09-17T13:07:36Z

leshan-client-demo/src/main/java/org/eclipse/leshan/client/demo/LeshanClientDemo.java

+            } else if (useOSCore) {
+                initializer.setInstancesForObject(SECURITY, secOSCore(serverURI, 123));
+                initializer.setInstancesForObject(OSCORE, Oscore.set("11223344", "AA", "BB")); //Hardcoded values
+                initializer.setInstancesForObject(SERVER, new Server(123, 30, BindingMode.U, false));


Regarding your remarks here :

One issue I had was how exactly to link the Security and Oscore objects together. I now see your suggestion here: #726 (comment)

To be able to do that, you will probably need to add a new XML version of security object.

You should get it here

The security object xml has been updated and a link to the OSCORE object introduced in commit ecebdfa

sbernard31 · 2019-09-17T13:15:07Z

leshan-core/src/main/java/org/eclipse/leshan/SecurityMode.java

 *******************************************************************************/
 package org.eclipse.leshan;

 /**
 * The different DTLS security modes
 */
 public enum SecurityMode {
-    PSK(0), RPK(1), X509(2), NO_SEC(3);
+    PSK(0), RPK(1), X509(2), NO_SEC(3), OSCORE(4);


Regarding the specification, 4 is for "Certificate mode with EST" and there is nothing for OSCORE 🤔 .

(It is not excluded there is issue in the specification but for now, I don't understand all of this enough to say if this is an issue 😁)

I now removed the SecurityMode for OSCORE. Instead the OSCORE Security Mode link in the Security object will be used. Done in commit 3a4e777

sbernard31 · 2019-09-17T13:18:11Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/servers/ServersInfoExtractor.java

@@ -112,6 +122,16 @@ public static ServersInfo getInfo(Map<Integer, LwM2mObjectEnabler> objectEnabler
                        info.clientCertificate = getClientCertificate(security);
                        info.serverCertificate = getServerCertificate(security);
                        info.privateKey = getPrivateKey(security);
+                    } else if (info.secureMode == SecurityMode.OSCORE) {


as there is no OSCORE securityMode maybe we can just look if the resource OSCORE Security Mode(17) is set.

If this is set you get the right index for the FIXME below :)

Yes, I am no longer checking the security mode here to decide what to do but rather if the link from the Security object has been set. I am also using the link from the Security object to find the OSCORE object here. Done in commit 3e7db26

sbernard31 · 2019-09-17T13:23:18Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/object/Oscore.java

+        int defaultHmacAlgorithm = -10; //HKDF_HMAC_SHA_256
+
+        return new Oscore(masterSecret, senderId, recipientId, defaultAeadAlgorithm, defaultHmacAlgorithm, "");
+    }


What is a benefits of both static "constructor" above ?

I saw that the Security object has many static constructors for different combinations of security settings. I also wanted a constructor with good default values for the two algorithm parameters.

Do you mean that I should reduce it to only 1 static constructor or not have any at all, just the normal constructor?

You can create several constructors with different number of parameter.

(static constructor is a good way to name a constructor but the con is that you can not reused when you extend the class)

You can create several constructors with different number of parameter.

(static constructor is a good way to name a constructor but the con is that you can not reused when you extend the class)

Okay, so it would be better to have 1 or several normal constructors, but no static constructors for the Oscore class?

Yep, as it seems you don't really need to "name" your constructor, I think it's better to have several "normal" constructors.

Removed the static "constructors" in commit dc37bd9

sbernard31 · 2019-09-17T13:29:12Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/object/Security.java

+    /**
+     * Returns a new security instance (OSCORE) for a device management server.
+     */
+    public static Security secOSCore(String serverUri, int shortServerId) {


here this would be more something like :

oscore (or oscoreOnly)

and one day we will add something like this ?

pskWithOscore

rpkWithOscore

x509WithOscore

Yes, that makes sense. I can rename the current method to oscore/oscoreOnly. And then we proceed with adding methods later that combine DTLS+OSCORE.

Renamed the current method to oscoreOnly in commit dd58e43 . In the future further methods can be added.

sbernard31 · 2019-09-17T13:34:36Z

leshan-core-cf/src/main/java/org/eclipse/leshan/core/californium/EndpointFactory.java

@@ -28,6 +30,8 @@
 public interface EndpointFactory {

    CoapEndpoint createUnsecuredEndpoint(InetSocketAddress address, NetworkConfig coapConfig, ObservationStore store);
+
+    CoapEndpoint createOSCoreEndpoint(InetSocketAddress address, NetworkConfig coapConfig, ObservationStore store, HashMapCtxDB db);


for now my feeling is that there no need to add createOSCoreEndpoint maybe just add the db parameter to both other methods (but as the californium API is not yet finished, I maybe missed something)

Yes I removed this new method and added the db parameter to the others instead. Done in commit 495d501 (After updating to Californium 2.0.0-M16 this parameter can be used.)

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2019-09-17T14:51:32Z

I suppose we should add a new attribute for OSCORE which could be used in addition of existing DTLS identity (in case of OSCORE over coaps) or without it (in case of OSCORE over coap).

Yes, that makes sense.

In oscore what could be the value which could be used as an identifier ? I need to read the rfc again (intuitively I suppose that Sender ID or/and Recipient ID could be candidate?)

A combination of the sender ID, recipient ID and ID context should identify a pair of endpoints communicating using OSCORE.

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 · 2019-09-18T12:33:11Z

leshan-client-demo/src/main/java/org/eclipse/leshan/client/demo/LeshanClientDemo.java

-                initializer.setInstancesForObject(SECURITY, oscoreOnly(serverURI, 123));
-                initializer.setInstancesForObject(OSCORE, new Oscore("11223344", "AA", "BB")); //Hardcoded values
+            	Oscore oscoreObject = new Oscore("11223344", "AA", "BB"); //Hardcoded values
+            	oscoreObject.setId(12345);


if you want you can create an Oscore constructor which take an instanceID in parameter

Yes, I will do that instead.

Now done in commit ae73ab2

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 · 2019-09-18T12:47:35Z

You should have a look at §5.5.5 - Endpoint Client Name about you "identity" concern.

About that is sender ID / receiver ID available in californium endpoint context ?

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 · 2019-09-18T12:49:47Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/object/Oscore.java

-    public Oscore(String masterSecret, String senderId, String recipientId, int aeadAlgorithm, int hkdfAlgorithm, String masterSalt) {
-        this.masterSecret = masterSecret;
+    public Oscore(int instanceId, String masterSecret, String senderId, String recipientId, int aeadAlgorithm, int hkdfAlgorithm, String masterSalt) {
+        this.setId(instanceId);


clearly not a big deal but you can use super(id)

True, I will change that as part of the next commit. Done in commit 6617c21

sbernard31 · 2019-09-18T12:50:57Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/servers/ServersInfoExtractor.java

+                    boolean useOscore = false;
+                    if(oscoreObjectId == OSCORE) {
+                    	useOscore = true;
+                    }


boolean useOscore = oscoreObjectId == OSCORE ?

Yeah, looks more compact and nicer. I will change it. Done in commit 6617c21

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 · 2019-09-18T13:05:27Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/servers/ServersInfoExtractor.java

+                    if(useOscore) {
+                    	// search corresponding oscore object
+                    	LwM2mObjectInstance oscoreInstance = null;
+                        for (LwM2mObjectInstance oscore : oscores.getInstances().values()) {


Not need to iterate, there is a getter by Id => oscores.getInstance(id)

Yes, that is much nicer. Fixed in commit a255bd6

rikard-sics · 2019-09-18T13:06:53Z

About that is sender ID / receiver ID available in californium endpoint context ?

Yes it is available but only from Californium 2.0.0-M17. Since the context DB is a singleton in the Californium we are using now the values could also be gotten from there (for now).

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2019-09-18T14:06:25Z

You should have a look at §5.5.5 - Endpoint Client Name about you "identity" concern.

I now updated the code so that the Sender ID of the client is equal to its name (endpoint). So if I start a client with name "c1" its sender ID will be 0x6331, and on the server side in the web interface the sender ID must be entered as "6331" (hexadecimal). Done in commit 401db2b

Of course this restricts the configuration options for the administrator somewhat since the sender ID and client name cannot be chosen independently.

rikard-sics · 2019-09-18T14:26:39Z

For further work on the identity I would perhaps wait until after updating to Californium 2.0.0-M17. There the sender ID, recipient ID and ID context are conveniently available in the endpoint context.

sbernard31 · 2019-09-18T14:48:16Z

...cf/src/main/java/org/eclipse/leshan/client/californium/impl/CaliforniumEndpointsManager.java

-                System.out.println("Adding OSCORE CTX " + serverInfo.getFullUri().toASCIIString());
+            // oscore only mode
+            if(serverInfo.useOscore) {
+            	System.out.println("Adding OSCORE CTX " + serverInfo.getFullUri().toASCIIString());


Please use LOG instead of sysout ? :)

Yes, that was an old debug statement I should have fixed. Now done in commit a83918e

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2019-09-18T15:21:36Z

I had to do one more commit regarding the registration, Identity and SecurityChecker. Previously I could not register an OSCORE client using the name I configured it with in the web interface of the server. Since it counted as not secure but yet had a non-null securityInfo it wanted me to use DTLS. I fixed this in commit 40213f6

The SecurityChecker should be expanded with checks to the identity when using OSCORE. Also the Identity class expanded to support OSCORE-related identities. SecurityInfo already has a useOSCore flag.

sbernard31 · 2019-09-18T15:21:40Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/servers/ServerInfo.java

@@ -65,7 +66,7 @@ public URI getFullUri() {
    }

    public boolean isSecure() {
-        return secureMode != SecurityMode.NO_SEC;
+        return secureMode != SecurityMode.NO_SEC || useOscore == true;


I know this is a bit strange but here I will not really change this.

Historically this was used to know if we should use unsecured(coap) endpoint or secured(coaps) endpoint.

So maybe the right name is "isCoaps" ?
But anyway I read the code again and maybe this does not make so much sense now, so it would be more wise to not touch it for now.

Yes, fair enough, I do not need to change the functionality of isSecure. I see your point that it is more about coap vs. coaps (and OSCORE could apply to both cases). I undid this in commit 17098a7

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 · 2019-09-18T15:37:35Z

About CaliforniumEndpointsManager, maybe I missed something but the idea is to begin by supporting Oscore without DTLS right ?

So I think the oscore stuff is not at "The Good Place".

For further work on the identity I would perhaps wait until after updating to Californium 2.0.0-M17. There the sender ID, recipient ID and ID context are conveniently available in the endpoint context.

makes sense !

rikard-sics · 2019-09-18T15:40:02Z

About CaliforniumEndpointsManager, maybe I missed something but the idea is to begin by supporting Oscore without DTLS right ?

Yes, my intention was to start by supporting OSCORE without DTLS. So the good place would be outside of the if(serverInfo.isSecure()), like down by row 220?

sbernard31 · 2019-09-18T15:46:04Z

So the good place would be outside of the if(serverInfo.isSecure()), like down by row 220?

yep :)

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2019-09-18T15:53:00Z

yep :)

Okay, sounds good :)

I have now done this in commit 7116077

Perhaps at some later stage it could be extracted to a separate method since we may want to combine the OSCORE configuration of an endpoint with different DTLS modes in a convenient way and minimize code reuse.

sbernard31 · 2019-09-18T16:08:20Z

Perhaps at some later stage it could be extracted to a separate method since we may want to combine the OSCORE configuration of an endpoint with different DTLS modes in a convenient way and minimize code reuse.

Correct, we will see when we will add this.

I think we are good, do you plan to add more stuff in this PR ? (maybe just fix formatting issue?)

rikard-sics · 2019-09-18T16:59:42Z

I think we are good, do you plan to add more stuff in this PR ? (maybe just fix formatting issue?)

Yes, fixing the formatting was the last thing I had planned. I will proceed with that.

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2019-09-19T06:21:16Z

I think we are good, do you plan to add more stuff in this PR ? (maybe just fix formatting issue?)

I have now applied the Leshan Eclipse formatter to the parts of the code I have changed. Done in commit 0138768

sbernard31 · 2019-09-19T13:14:02Z

I squashed all your commits in one (e310127) which is now integrated in oscore branch.

I tried a "mvn clean install" and that failed so I fixed issues, see commit cc6136a
Do not hesitate to launch this often, tests is here to help you ;)

About formatting I find several more issues, see commit e2ae4bc.

Are you using eclipse ? Did you follow this guide. In theory "format on save" is activated per project, so it should be very difficult to write bad formatted code 🤔.

What do you plan as next steps ? :)

rikard-sics · 2019-09-19T13:25:37Z

I squashed all your commits in one (e310127) which is now integrated in oscore branch.

Thanks!

I tried a "mvn clean install" and that failed so I fixed issues, see commit cc6136a
Do not hesitate to launch this often, tests is here to help you ;)

Yes, I was using Eclipse and not running the maven tests from the command line like that. I should have ran tests more frequently and will try to do so in the future.

About formatting I find several more issues, see commit e2ae4bc.

Are you using eclipse ? Did you follow this guide. In theory "format on save" is activated per project, so it should be very difficult to write bad formatted code .

Yes, I am using Eclipse. Unfortunately up until now I have not been using the "format on save" feature. But rather formatting manually with CTRL+SHIFT+F. I will start using "format on save" as you suggest.

What do you plan as next steps ? :)

Some thoughts for next steps:

Update code to use Californium 2.0.0-M17
Implement an Identity and related things in SecurityChecker for OSCORE
Work on the server side issues according to discussion in OSCORE over coap at server side #727
Add possibility of entering OSCORE context information as command line parameters for client

What do you think makes most sense to prioritize?

sbernard31 · 2019-09-19T13:29:02Z

This order makes sense to me :)

rikard-sics · 2019-09-19T13:30:21Z

This order makes sense to me :)

Okay, sounds good. I will proceed accordingly. :)

sbernard31 reviewed Sep 17, 2019

View reviewed changes

Removed Californium dependencies from leshan-client-core

8fc7bfc

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics added 4 commits September 17, 2019 20:48

Removed static initializers from Oscore class, instead use constructors

dc37bd9

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Use hex string manipulation method from Leshan rather than adding new

ebc728a

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Renamed initialization method for OSCORE only in Security object

dd58e43

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Updated Security object xml and introduced link to OSCORE object

ecebdfa

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 reviewed Sep 18, 2019

View reviewed changes

rikard-sics added 2 commits September 18, 2019 14:35

Use link from Security to find OSCORE object when creating ServerInfo

3e7db26

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Modified OSCORE object to take an instance id in its constructor

ae73ab2

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Removed OSCORE endpoint create method and added db parameter to others

495d501

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 reviewed Sep 18, 2019

View reviewed changes

Two small fixes

6617c21

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 reviewed Sep 18, 2019

View reviewed changes

rikard-sics added 3 commits September 18, 2019 15:08

Do not iterate to find OSCORE object in ServersInfoExtractor

a255bd6

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Removed SecurityMode for OSCORE. Instead use OSCORE Security Mode link

3a4e777

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Set client name as client OSCORE Sender ID

401db2b

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 reviewed Sep 18, 2019

View reviewed changes

rikard-sics added 2 commits September 18, 2019 16:53

Updated print statement to use LOG instead

a83918e

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Fix for clients to register using name configured in server interface

40213f6

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 reviewed Sep 18, 2019

View reviewed changes

Undid the change to isSecure method in ServerInfo

17098a7

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Moved creation of OSCORE endpoint to a better place

7116077

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Applying Leshan Eclipse formatter to code and small comment fixes

0138768

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 closed this Sep 19, 2019

sbernard31 added this to the 2.0.0 milestone Oct 24, 2019

Generate OSCORE object and use it for endpoint creation #736

Generate OSCORE object and use it for endpoint creation #736

Conversation

rikard-sics commented Sep 12, 2019 • edited Loading

rikard-sics commented Sep 12, 2019

rikard-sics commented Sep 12, 2019

sbernard31 Sep 17, 2019 • edited Loading

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

sbernard31 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

rikard-sics Sep 17, 2019 • edited Loading

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

rikard-sics Sep 18, 2019 • edited Loading

Choose a reason for hiding this comment

rikard-sics commented Sep 17, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sbernard31 commented Sep 18, 2019

Choose a reason for hiding this comment

rikard-sics Sep 18, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rikard-sics Sep 18, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rikard-sics commented Sep 18, 2019

rikard-sics commented Sep 18, 2019 • edited Loading

rikard-sics commented Sep 18, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rikard-sics commented Sep 18, 2019 • edited Loading

Choose a reason for hiding this comment

rikard-sics Sep 18, 2019 • edited Loading

Choose a reason for hiding this comment

sbernard31 commented Sep 18, 2019

rikard-sics commented Sep 18, 2019

sbernard31 commented Sep 18, 2019

rikard-sics commented Sep 18, 2019

sbernard31 commented Sep 18, 2019

rikard-sics commented Sep 18, 2019

rikard-sics commented Sep 19, 2019

sbernard31 commented Sep 19, 2019

rikard-sics commented Sep 19, 2019 • edited Loading

sbernard31 commented Sep 19, 2019

rikard-sics commented Sep 19, 2019

rikard-sics commented Sep 12, 2019 •

edited

Loading

sbernard31 Sep 17, 2019 •

edited

Loading

rikard-sics Sep 17, 2019 •

edited

Loading

rikard-sics Sep 18, 2019 •

edited

Loading

rikard-sics Sep 18, 2019 •

edited

Loading

rikard-sics Sep 18, 2019 •

edited

Loading

rikard-sics commented Sep 18, 2019 •

edited

Loading

rikard-sics commented Sep 18, 2019 •

edited

Loading

rikard-sics Sep 18, 2019 •

edited

Loading

rikard-sics commented Sep 19, 2019 •

edited

Loading