[asterisk-bugs] [JIRA] (ASTERISK-27421) RTP source learning not working with devices that have some clock issues

[Richard Mudgett (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-27421?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=240042#comment-240042 ] 

Richard Mudgett commented on ASTERISK-27421:
--------------------------------------------

If you put the change up on Gerrit then it will be reviewed and included faster. Instructions for doing so are on the wiki[1]. By not doing this you are asking someone else to take it through the complete process and with limited resources this can take some time.

[1] https://wiki.asterisk.org/wiki/display/AST/Gerrit+Usage

> RTP source learning not working with devices that have some clock issues
> ------------------------------------------------------------------------
>
>                 Key: ASTERISK-27421
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-27421
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Resources/res_rtp_asterisk
>    Affects Versions: 13.17.0, 13.17.1, 13.17.2, 13.18.0, 13.18.1, 13.18.2
>            Reporter: nappsoft
>            Severity: Minor
>         Attachments: accept_jitter.diff
>
>
> I've observed some Softphones that do not send packets perfectly frequent but for example with 30ms between the first and the second packet and between the second and the third and <1 ms between the third and the fourth. While a jitterbuffer can handle this perfectly, the RTP source learning algorithm doesn't accept that: it will reset the learning process on every thourd packet as there are less than 5 ms between these packets. (That's not a theory: I just have observed a client that is exactly behaving as described).
> I've attached a patch that is changing the implementation in a way that not the time between two packets is taken into account to detect a flood, but the time between the first and the last packet during the learning phase. (A timeout lower than 30ms between 4 packets will not be accepted) what solves the described issue and will still detect flood attacks.



--
This message was sent by Atlassian JIRA
(v6.2#6252)