<!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 text="#000000" bgcolor="#ffffff">
    Hi Jaco,<br>
    <br>
    if I understand correctly, your segfault did not happen during in
    T38gateway, but while receiving fax to tiff file (ReceiveFax), am I
    right?<br>
    I haven't checked neither patched this (because my Asterisks are
    only relaying faxes, not terminating/originating to/from tiff file),
    but if your segfault happen when data are passed to libspandsp, it
    should be the same situation as mine was. Code in res_fax allows
    slinear/alaw/ulaw frames to be passed to res_fax_spandsp and then to
    libspandsp, but libspandsp accepts only slinear. When ulaw/alaw
    frames are passed here, bad things can happen.<br>
    <pre class="moz-signature" cols="72">-- 
Regards,
Michal Rybarik</pre>
    <br>
    <br>
    Dňa 6. 2. 2014 12:07, Jaco Kroon  wrote / napísal(a):
    <blockquote cite="mid:52F36CDD.5060608@uls.co.za" type="cite">
      <meta content="text/html; charset=ISO-8859-1"
        http-equiv="Content-Type">
      <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:part1.08030301.01020604@rybarik.sk"
            usemap="#Map" style="color: white;" width="530" border="0"
            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>
    </blockquote>
    <br>
  </body>
</html>