abandoned turn causes processes to sleep forever when handling retransmitted publishes

[etamme]

This exists on latest master 055f4b1

Scenario: 2 PUBLISHES with the same E-Tag enter https://github.com/OpenSIPS/opensips/blob/master/modules/presence/presentity.c#L546 as a result of a retransmission.

The first publish finds the presentity, increments last turn to say 5 for the sake of argument, then it enters the while loop, finds the presentity and continues processing updating it with a new e-tag and saving it the db releases locks and finishes.

The second publish enters because its turn matches 5, it finds the presentity and increments last turn to 6, it then enters the while loop, but it fails to find the presentity because the E-Tag has been changed by the previous publish. It continues processing, but it never calls next_turn_phtable before exiting.

This will cause the process to lockup forever, waiting for its turn, on the next publish received.

[wakamoleguy]

It continues processing, but it never calls next_turn_phtable before exiting.

It technically does call next_turn_phtable, but on a new presentity in the phtable. It does not increment the current_turn on the same presentity that it updated last_turn on. This causes a mismatch on that presentity going forward.

This will cause the process to lockup forever, waiting for its turn, on the next publish received.

Specifically, both of the processes in this scenario will continue on their way. The next publish to the original presentity will lock, because current_turn is less than last_turn, and nothing exists to increment it.

[bogdan-iancu]

Thank you for investigating this - now that is clear what is happens, I will work on a way to fix it.

[wakamoleguy]

Here's a preliminary patch we put together:

https://gist.github.com/wakamoleguy/dc9d694e64364ba6f7d2

We haven't tested it yet, or even really compiled it. The general idea is that if a process detects that it is about to throw away an old presentity, it still waits its turn and calls next_turn_phtable on the old one before moving on.

[wakamoleguy]

I could make the patch into an actual pull request, if you'd prefer, although for several reasons (including vulgarity of comments, sorry!) it would be more for discussion than an actual pull. Let me know what you'd prefer.

[stale]

Any updates here? No progress has been made in the last 15 days, marking as stale. Will close this issue if no further updates are made in the next 30 days.

[bogdan-iancu]

OK, this needs to be revisited, the problem is still there.