[OpenSIPS-Devel] [OpenSIPS-Users] [Releases] Scheduling OpenSIPS 1.10 beta

[Saúl Ibarra Corretgé]

Hi Vlad!

On Jul 26, 2013, at 11:36 AM, Vlad Paiu wrote:

> Hello Saúl,
> 
> The TCP reads optimizations have already been pushed to OpenSIPS trunk ( in commits [1] and [2] )
> The read improvements cover :
>    - the TCP read buffer has been moved from a per connection basis to a per process basis
>        Advantages here cover : less memory used by OpenSIPS for a large number of processes ( 64 KB of SHM memory saved per TCP connection )
>                                              : faster acceptance of new TCP connections ( less SHM memory that needs to be allocated )
>    - make the TCP read part in OpenSIPS less vulnerable to attacks
>        Improvements here are : put a limit on max number of chunks that a TCP message can arrive in ( thus blocking attacks where the attacker would open                     the connection and then just push SIP messages byte by byte )
>                                               : put a limit on max number of seconds that a TCP message can arrive in
> 

Aha. I saw those commits some days ago, I thought I had missed something :-)

> For the TCP write part, we're now finalizing asynchronous writing ( both for actual writes and also for connects ) - will be implementing by moving the currently blocking writes and connects attempts from the TCP children to a separate process.
> 

I see, this is what I couldn't find so far, I'll keep an eye on the repo!

Thanks for taking the time to summarize it :-)


Regards,

--
Saúl Ibarra Corretgé
AG Projects