[asterisk-bugs] [JIRA] Updated: (ASTERISK-17797) Coding error in find_subchannel_and_lock()

[Richard Mudgett (JIRA)]

     [ https://issues.asterisk.org/jira/browse/ASTERISK-17797?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Richard Mudgett updated ASTERISK-17797:
---------------------------------------

    Description: 
The function find_subchannel_and_lock()in mgcp_chan.c has a coding error in two locations that prevent the subchannel from being found on non-dynamic gateways. This results in responses from half the gateways to be ignored.  I built Asterisk from source on a Fedora 14 system.  Here is the C code in question from Asterisk version 1.8.3.2:

{code}
	for (g = gateways ? gateways : find_realtime_gw(name, at, sin);
                g; g = g->next ? g->next : find_realtime_gw(name, at, sin)) {

        [lines omitted]

                        /* not dynamic, check if the name matches */
                        } else if (name) {
                             if (strcasecmp(g->name, at)) {
>>>> error >>>>                 g = g->next;
                                continue;
                             }
                        /* not dynamic, no name, check if the addr matches */
                        } else if (!name && sin) {
                             if ((g->addr.sin_addr.s_addr != 
                                  sin->sin_addr.s_addr) ||
                                 (g->addr.sin_port != sin->sin_port)) {
>>>> error >>>>                  if(!g->next)
>>>> error >>>>                        g = find_realtime_gw(name, at, sin);
>>>> error >>>>                   else
>>>> error >>>>                        g = g->next;
                                  continue;
                            }
                        } else {
                            continue;
                        }
{code}

Since the "for" statement includes updating "g" from "g->next", the extra updates to "g" in the code above cause gateway entries to be skipped and responses from those gateways to continually exceed the maximum retry count.

I hope this helps!

Jeff Waltz

****** ADDITIONAL INFORMATION ******

I just downloaded the source for 1.8.4 and the erroneous code is still present.

  was:
The function find_subchannel_and_lock()in mgcp_chan.c has a coding error in two locations that prevent the subchannel from being found on non-dynamic gateways. This results in responses from half the gateways to be ignored.  I built Asterisk from source on a Fedora 14 system.  Here is the C code in question from Asterisk version 1.8.3.2:


	for (g = gateways ? gateways : find_realtime_gw(name, at, sin);
                g; g = g->next ? g->next : find_realtime_gw(name, at, sin)) {

        [lines omitted]

                        /* not dynamic, check if the name matches */
                        } else if (name) {
                             if (strcasecmp(g->name, at)) {
>>>> error >>>>                 g = g->next;
                                continue;
                             }
                        /* not dynamic, no name, check if the addr matches */
                        } else if (!name && sin) {
                             if ((g->addr.sin_addr.s_addr != 
                                  sin->sin_addr.s_addr) ||
                                 (g->addr.sin_port != sin->sin_port)) {
>>>> error >>>>                  if(!g->next)
>>>> error >>>>                        g = find_realtime_gw(name, at, sin);
>>>> error >>>>                   else
>>>> error >>>>                        g = g->next;
                                  continue;
                            }
                        } else {
                            continue;
                        }

Since the "for" statement includes updating "g" from "g->next", the extra updates to "g" in the code above cause gateway entries to be skipped and responses from those gateways to continually exceed the maximum retry count.

I hope this helps!

Jeff Waltz

****** ADDITIONAL INFORMATION ******

I just downloaded the source for 1.8.4 and the erroneous code is still present.


> Coding error in find_subchannel_and_lock()
> ------------------------------------------
>
>                 Key: ASTERISK-17797
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-17797
>             Project: Asterisk
>          Issue Type: Bug
>          Components: Channels/chan_mgcp
>    Affects Versions: 1.8.3
>            Reporter: Jeff Waltz
>            Severity: Minor
>
> The function find_subchannel_and_lock()in mgcp_chan.c has a coding error in two locations that prevent the subchannel from being found on non-dynamic gateways. This results in responses from half the gateways to be ignored.  I built Asterisk from source on a Fedora 14 system.  Here is the C code in question from Asterisk version 1.8.3.2:
> {code}
> 	for (g = gateways ? gateways : find_realtime_gw(name, at, sin);
>                 g; g = g->next ? g->next : find_realtime_gw(name, at, sin)) {
>         [lines omitted]
>                         /* not dynamic, check if the name matches */
>                         } else if (name) {
>                              if (strcasecmp(g->name, at)) {
> >>>> error >>>>                 g = g->next;
>                                 continue;
>                              }
>                         /* not dynamic, no name, check if the addr matches */
>                         } else if (!name && sin) {
>                              if ((g->addr.sin_addr.s_addr != 
>                                   sin->sin_addr.s_addr) ||
>                                  (g->addr.sin_port != sin->sin_port)) {
> >>>> error >>>>                  if(!g->next)
> >>>> error >>>>                        g = find_realtime_gw(name, at, sin);
> >>>> error >>>>                   else
> >>>> error >>>>                        g = g->next;
>                                   continue;
>                             }
>                         } else {
>                             continue;
>                         }
> {code}
> Since the "for" statement includes updating "g" from "g->next", the extra updates to "g" in the code above cause gateway entries to be skipped and responses from those gateways to continually exceed the maximum retry count.
> I hope this helps!
> Jeff Waltz
> ****** ADDITIONAL INFORMATION ******
> I just downloaded the source for 1.8.4 and the erroneous code is still present.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira