<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Matthew J. Roth escribió:
<blockquote cite="mid:4A2D8FA1.6020009@imminc.com" type="cite">
<pre wrap="">Miguel Molina wrote:
</pre>
<blockquote type="cite">
<pre wrap="">I recently upgraded a production machine to asterisk 1.4.25. It seems
quite stable but after ~5 days of normal operation it core dumped with
this result:
(gdb) bt
#0 0x00516402 in __kernel_vsyscall ()
#1 0x005b3d20 in raise () from /lib/libc.so.6
#2 0x005b5631 in abort () from /lib/libc.so.6
#3 0x005ebe6b in __libc_message () from /lib/libc.so.6
#4 0x005f3b16 in _int_free () from /lib/libc.so.6
#5 0x005f7070 in free () from /lib/libc.so.6
#6 0x005e2876 in <a class="moz-txt-link-abbreviated" href="mailto:fclose@@GLIBC_2.1">fclose@@GLIBC_2.1</a> () from /lib/libc.so.6
#7 0x0809eb2a in filestream_destructor (arg=0xb1f0f200) at file.c:340
#8 0x0806e412 in ao2_ref (user_data=0xb1f0f200, delta=-1) at astobj2.c:229
#9 0x0809c1e9 in ast_closestream (f=0xb1f0f200) at file.c:902
#10 0x00b03422 in local_ast_moh_stop (chan=0xb2150fd0) at
res_musiconhold.c:1058
--- SNIP --- SNIP --- SNIP --- SNIP --- SNIP --- SNIP --- SNIP --- SNIP ---
It looks like a very random situation, as this was not a high load moment.
Also the asterisk log showed this message in the exact instant of the
failure:
[Jun 8 13:21:13] ERROR[21601] astobj2.c: refcount -1 on object 0xb1f0f200
I understand that a core dump generated by asterisk compiled with
(standard) optimized values is marked as useless information, but IMHO
it still helps to know what's failing inside it. I appreciate any input
about this, could be this a bug? A library problem? Or a server memory
problem?
</pre>
</blockquote>
<pre wrap=""><!---->Miguel,
It looks like you are running into an acknowledged bug. There are open
issues in the bug tracker for both the 1.4 and 1.6 branches:
* <a class="moz-txt-link-freetext" href="https://issues.asterisk.org/view.php?id=15109">https://issues.asterisk.org/view.php?id=15109</a>
* <a class="moz-txt-link-freetext" href="https://issues.asterisk.org/view.php?id=15195">https://issues.asterisk.org/view.php?id=15195</a>
Please create an account and add your information to the bug tracker.
Regards,
Matthew Roth
InterMedia Marketing Solutions
Software Engineer and Systems Developer
</pre>
</blockquote>
Thank you very much for identifying this known bug. I already have an
account, so I will be posting my info on the bugtracker tomorrow.<br>
<br>
Regarding the trace and the error asterisk throws,<br>
<pre wrap="">[Jun 8 13:21:13] ERROR[21601] astobj2.c: refcount -1 on object 0xb1f0f200
</pre>
<br>
And what astobj2.c comments say on here:<br>
<br>
<pre class="fragment">00223 <span class="comment">/* this case must never happen */</span>
00224 <span class="keywordflow">if</span> (current_value < 0)
00225 <a class="code"
href="http://www.asterisk.org/doxygen/1.4/logger_8c.html#0eb07c73aa8c3475ef05c5465d9b5703">ast_log</a>(<a
class="code"
href="http://www.asterisk.org/doxygen/1.4/logger_8h.html#91193576ec6eae864eeba838a8c821f5">LOG_ERROR</a>, <span
class="stringliteral">"refcount %d on object %p\n"</span>, current_value, <a
class="code"
href="http://www.asterisk.org/doxygen/1.4/structastobj2.html#cf53b89bb38cfc4a89c83c743970553c">user_data</a>);
</pre>
<br>
I think mmichelson is right, quoting him on bug 15123: "I think,
though, that this particular issue is due to poor refcounting
practices present in Asterisk 1.4.22 (which have been fixed already,
btw).". So this may be confirming his assumption.<br>
<br>
Regards,<br>
<pre class="moz-signature" cols="72">--
Ing. Miguel Molina
Grupo de Tecnología
Millenium Phone Center
</pre>
</body>
</html>