13 Aug
2014
13 Aug
'14
5:58 p.m.
On 13/08/14 16:24, Alex Balashov wrote:
On 08/13/2014 10:21 AM, AliReza Khoshgoftar Monfared
wrote:
While that is true for htable (direct) operations, one should be careful
when using hash table items in expressions.
So:
- next is an atomic increment of $sht(a=>x)
$var(x) = $shtinc(a=>x);
- but next might rise races:
$sht(a=>x) = $sht(a=>x) + 1;
because first the value is read, an add expression is evaluated and then
the item in hash table is set to new value. During the evaluation of the
expression, the current process can lose CPU, and another process can
update the item. So you can make it safer with:
sht_lock("a=>x");
$sht(a=>x) = $sht(a=>x) + 1;
sht_unlock("a=>x");
Cheers,
Daniel
--
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Next Kamailio Advanced Trainings 2014 - http://www.asipto.com
Sep 22-25, Berlin, Germany ::: Oct 15-17, San Francisco, USA
Another concern is that, if I specify a number of
child processes in my
config (e.g. children=4), then will it be safe to increment the hash
table entries in the route block upon receipt of messages? i.e will the
table entries that reside in the shared memory be precise? or do I need
to lock/unlock them while updating?
htable operations are thread-safe, as implemented under the hood. You
do not explicitly need to lock anything.