[OpenSER-Devel] need advice for avp lock

[Di-Shi Sun]

Hi Henning,

I believe we met the same problem. For us, only one flag in an AVP (5 AVPs
for a call) is set and the logic is simple. So, we did not meet a crash
condition. It happened 17 times in a 1M call test at 50 cps.

Klaus Darilion suggested we read the tm code for the AVP lock. I will study
it.

Thanks,

Di-Shi Sun.

----- Original Message ----- 
From: "Henning Westerholt" <henning.westerholt at 1und1.de>
To: <devel at openser.org>
Cc: "Di-Shi Sun" <di-shi at transnexus.com>; "Support of TransNexus"
<support at transnexus.com>
Sent: Thursday, October 18, 2007 4:09 PM
Subject: Re: [OpenSER-Devel] need advice for avp lock


> On Thursday 18 October 2007, Di-Shi Sun wrote:
> > All,
> >
> > We met a race condition for accessing avp values. We use avps to store
some
> > info. Both 180 and 200 messages may trigger a logic to access the avp
> > values. Sometimes they arrive at almost the same time. Would anyone
please
> > give us some advice on what lock mechanism we should use for it?
>
> Hello Di-Shi,
>
> i guess we observed the the same race condition here.. Sometimes the
internal
> avp list get corrupted, and then contains invalid data. After some time
the
> child dies with a SIG 11 in search_first_avp or other avp related
functions.
>
> Do you have the same problem?
>
> The problem appears recently after we're starting to use more AVPs. As
this
> happens only sometimes, and only with high load, the problem is dificult
to
> debug. So any further hints are appreciated, too.
>
> Cheers,
>
> Henning
>
>