<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix"><font face="Helvetica, Arial,
        sans-serif">Hi All,<br>
        <br>
        Could this backtrace possibly be related?<br>
        <br>
        #0  process_rx_data (t=0x7fae54c698a8, user_data=0x2,
        data_type=1, field_type=<optimized out>,
        buf=0x7fae11c58cda "cng", len=0) at t38_terminal.c:314<br>
        #1  0x00007fae11c22c7d in t38_core_rx_ifp_packet
        (s=0x7fae54c698a8, buf=0x7fae54c8475b "\002", len=1,
        seq_no=<optimized out>) at t38_core.c:459<br>
        #2  0x00007fae50ea96c5 in generic_fax_exec
        (chan=chan@entry=0x7fadc4548c18,
        details=details@entry=0x7fad50602c28,
        reserved=reserved@entry=0x7fad50155478, token=<optimized
        out>) at res_fax.c:1498<br>
        #3  0x00007fae50eaea9e in receivefax_exec (chan=0x7fadc4548c18,
        data=<optimized out>) at res_fax.c:1932<br>
        #4  0x0000000000530fdd in pbx_exec (c=c@entry=0x7fadc4548c18,
        app=app@entry=0x2ddca60, data=data@entry=0x7fad838b6cd0
        "/tmp/morpheus-1391681512.850.tiff") at pbx.c:1622<br>
        #5  0x000000000053656f in pbx_extension_helper
        (c=c@entry=0x7fadc4548c18, context=<optimized out>,
        exten=exten@entry=0x7fadc4549ab8 "0123489251",
        priority=priority@entry=6, label=label@entry=0x0,
        callerid=callerid@entry=0x7fadc44757b0 "0126413300",
        action=action@entry=E_SPAWN, found=found@entry=0x7fad838bad60, <br>
            combined_find_spawn=combined_find_spawn@entry=1, con=0x0) at
        pbx.c:4922<br>
        #6  0x00000000005404a4 in ast_spawn_extension
        (found=0x7fad838bad60, callerid=0x7fadc44757b0 "0126413300",
        priority=6, exten=0x7fadc4549ab8 "0123489251",
        context=<optimized out>, c=0x7fadc4548c18,
        combined_find_spawn=<optimized out>) at pbx.c:6038<br>
        #7  __ast_pbx_run (c=c@entry=0x7fadc4548c18,
        args=args@entry=0x0) at pbx.c:6513<br>
        #8  0x0000000000541c0b in pbx_thread
        (data=data@entry=0x7fadc4548c18) at pbx.c:6843<br>
        #9  0x0000000000587c5a in dummy_start (data=<optimized
        out>) at utils.c:1162<br>
        #10 0x00007fae530f2f3a in start_thread (arg=0x7fad838bb700) at
        pthread_create.c:308<br>
        #11 0x00007fae54754dad in clone () at
        ../sysdeps/unix/sysv/linux/x86_64/clone.S:113<br>
        <br>
        Had about 11 of those this morning on asterisk 11.7.0.  Codec's
        that's allowed on SIP though is g729 and gsm only, so no
        ulaw/alaw allowed.  Actually, just double checked, ulaw/alaw is
        (was now) allowed, so someone is possibly trying to run in
        bypass mode, resulting in the t38 gateway instead of t38 pass
        through.  I downgraded to 11.6.0 and hadn't had a crash since
        but I opted to disable ulaw+alaw in any case, just to be on the
        safer side.<br>
        <br>
      </font>
      <div class="moz-signature">Kind Regards,<br>
        Jaco Kroon<br>
        <img src="cid:part1.08050707.02040205@uls.co.za" usemap="#Map"
          style="color:white" border="0" width="530" height="100">
        <map name="Map" id="Map">
          <area shape="rect" coords="441,19,460,36"
            href="https://www.facebook.com/ultimatelinuxsolutions">
          <area shape="rect" coords="441,39,458,57"
            href="http://news.uls.co.za/">
          <area shape="rect" coords="354,62,461,73"
            href="http://www.uls.co.za/">
        </map>
      </div>
      On 01/02/2014 06:49, Michal Rybárik wrote:<br>
    </div>
    <blockquote cite="mid:52EC7CE6.7000208@rybarik.sk" type="cite">Hello
      Pavel,
      <br>
      <br>
      On 01/31/2014 07:59 AM, Pavel Troller wrote:
      <br>
      <blockquote type="cite">
        <blockquote type="cite">This code will translate non-slinear
          frames to slinear, just before they
          <br>
          are sent to libspandsp for v21detection. With this patch
          applied, v21
          <br>
          detection is done also for RTP (SIP) alaw/ulaw frames, so
          maybe SIP/G711
          <br>
          <->  SIP/T38 gateway will work too. I tested
          DAHDI<->  SIP/T38, gateway
          <br>
          works both ways, voice calls too. Is it better now? :o)
          <br>
        </blockquote>
        I fully understand the code, but I'm not trained enough in the
        Asterisk
        <br>
        internals to respond to questions, which immediately appeared in
        my head:
        <br>
        1) In the original code, the result from
        fax_gateway_detect_v21() is returned.
        <br>
        Now, you are returning the original frame. I quickly looked at
        the above
        <br>
        routine and it in turn calls fax_gateway_request_t38() and
        returns its
        <br>
        result (but not always), and in the fax_gateway_request_t38()
        function
        <br>
        they are also returning different things according to results of
        the
        <br>
        program flow. So, is it really safe to do this ? Are you sure,
        that the
        <br>
        real result is really unneccessary ?
        <br>
        2) Are you sure, that ast_translate() will always allocate a new
        buffer for
        <br>
        tmpframe ? Is it written somewhere ? Isn't it possible that it
        will just
        <br>
        reallocate the buffer for the original frame to increase its
        size and return
        <br>
        its pointer, so by doing ast_frfree() you would just deallocate
        the same
        <br>
        buffer, thus making big troubles ? You would find it by checking
        that
        <br>
        tmpframe != f...
        <br>
           As you can see, I'm very careful, or maybe even a bit
        conservative, with
        <br>
        patching things, unless I really DEEPLY understand, how they are
        going...
        <br>
        So, I believe, that you really studied the code enough to be
        sure, that
        <br>
        you can really clear my doubts by your deep knowledge... I
        didn't have time
        <br>
        to study the code to such extent...
        <br>
      </blockquote>
      <br>
      Answering these questions is not easy for me too, there are some
      parts of res_fax code which I don't fully understand. So I rather
      reworked the patch and moved it to another place, where
      functionality is easier to understand, and when it shouldn't harm
      anything. I uploaded diff to JIRA  -
      https://issues.asterisk.org/jira/browse/ASTERISK-20149
      <br>
      <br>
      Regards,
      <br>
      Michal Rybarik
      <br>
      <br>
    </blockquote>
    <br>
  </body>
</html>