Hi Jeff,

It appears you are hitting a known "maxAlarms" deadlock on LocalEndpoint 
(ASACORE-2650<https://jira.allseenalliance.org/browse/ASACORE-2650>), which 
unfortunately is still an active ticket. We had hit this deadlock regularly in 
the past, particularly when testing various scenarios in our corporate network. 
The good news is there is a workaround put in place in 16.04. Please try one of 
the following:

1. Consider using the AllJoyn C APIs exported from msajapi.dll. The export 
library (msajapi.lib) is shipped as part of the Windows 10 SDK.
2. If you must use the C++ binding, try building AllJoyn and your service with 
LOCAL_ENDPOINT_MAXALARMS=0 defined. See 
ASACORE-2810<https://jira.allseenalliance.org/browse/ASACORE-2810> for more 
details.

Thanks,
Way

From: [email protected] 
[mailto:[email protected]] On Behalf Of Blayney, 
Jeff
Sent: Monday, June 13, 2016 3:02 PM
To: [email protected]
Cc: Poole, Gary <[email protected]>; Cazier, Robb P (CEPS SW Development) 
<[email protected]>; [email protected]; Singh, Baljit ((PPS-ASG)) 
<[email protected]>; Truong, Binh <[email protected]>
Subject: Re: [Allseen-core] Help on Alljoyn issue

+ Binh and Babak

From: Blayney, Jeff
Sent: Monday, June 13, 2016 4:00 PM
To: '[email protected]' 
<[email protected]<mailto:[email protected]>>
Cc: Singh, Baljit ((PPS-ASG)) 
<[email protected]<mailto:[email protected]>>; Poole, Gary 
<[email protected]<mailto:[email protected]>>; Cazier, Robb P (CEPS SW 
Development) <[email protected]<mailto:[email protected]>>
Subject: FW: Help on Alljoyn issue

Hello,

I am developer on the team using AllJoyn. We are experiencing an intermittent 
deadlock on the PC client, and it seems to be related to authentication. We had 
days when everything works great and days when the app keeps getting stuck in 
deadlock. The deadlock usually happens on first method call after joining a 
session. Frequency is higher when we have more than 2 devices when they join 
session with peer again.

Here is the information about our AllJoyn setup and how we are using it.

*         We initialize our Bus attachment with 64 threads max Concurrency.

*         We are using non-multipoint sessions because there was issue with 
rejoins multipoint sessions.

*         We are using IP transport layer.

*         On our service start-up we clear our keystore. This at minimum 
required on iOS because a defect.

*         On our app start up, we join session with all discovered devices that 
have our AppId and then we call a method on remote peer to authenticate with 
them.

*         We have made sure that all methods in BusObject, SessionListener, 
SessionPortListener, AboutListener and Authlistener are not doing too much 
overhead or long running task. If we have long running task, we spawn new 
thread for it.

*         In all our listeners and BusObject methods we call 
EnableConcurrentCallback at beginning of the method. We often experience an 
error ER_TIMER_NOT_ALLOWED. Even though we are following best practices for 
Enabling concurrent callbacks.

*         Our interface secure and we are using PSK for authorization.

*         iOS is using 15.04 while Android is using 15.09a and windows is using 
15.09

*         We are setting SetLinkTimeout on all hosted and joined sessions.

*         We occasionally get following error in securityviolation auth callback

o   QStatus::ER_BUS_MESSAGE_DECRYPTION_FAILED

o   QStatus::ER_BUS_DESTINATION_NOT_AUTHENTICATED

o   QStatus::ER_AUTH_FAIL
Usually those error get resolved if we make a method call on remote peer and 
that seems to force authentication.


Thanks for any direction or assistance,

-Baljit and Jeff

From: Truong, Binh
Sent: Friday, June 10, 2016 12:29 PM
To: Shakib, Babak <[email protected]<mailto:[email protected]>>
Cc: Cazier, Robb P (CEPS SW Development) 
<[email protected]<mailto:[email protected]>>; Poole, Gary 
<[email protected]<mailto:[email protected]>>; Singh, Baljit 
((PPS-ASG)) <[email protected]<mailto:[email protected]>>; Blayney, Jeff 
<[email protected]<mailto:[email protected]>>
Subject: Help on Alljoyn issue

Hi Babak,

Thank you for offering to help find us some Alljoyn expert, our problem is the 
following:

Some background, the AllJoyn solution the team has develop involves an AllJoyn 
application on windows 10 running as a service using windows 10 AllJoyn Router, 
the UI is a win32 application, this application communicates to a similar 
AllJoyn android application with bundled router and potentially other mobile 
platforms also like iOS. This is using Alljoyn Core 15.09.

The issue we are seeing is that on windows 10 when the session starts there's 
an authentication happening but sometimes it's takes a very long time (20 secs) 
and other times it seems like its deadlocked.  The belief here is this process 
is all taken care off by AllJoyn core and the only thing we pass in the 
keystore.  Have you seen this issue before and are there pointers you can help 
us debug it further, we seem to see this only on windows.

We available for a call to explain more and get some advice.

Thanks in advance,
-Binh

<deadlock.png>
<newDeadlock.png>
_______________________________________________
Allseen-core mailing list
[email protected]
https://lists.allseenalliance.org/mailman/listinfo/allseen-core

Reply via email to