<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<style type="text/css" style="display:none"><!--P{margin-top:0;margin-bottom:0;} @font-face
{font-family:"Cambria Math"}
@font-face
{font-family:Calibri}
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
margin-bottom:.0001pt;
font-size:11.0pt;
font-family:"Calibri",sans-serif}
a:link, span.MsoHyperlink
{color:#0563C1;
text-decoration:underline}
a:visited, span.MsoHyperlinkFollowed
{color:#954F72;
text-decoration:underline}
span.EmailStyle17
{font-family:"Arial",sans-serif;
color:windowtext;
font-weight:normal;
font-style:normal}
.MsoChpDefault
{font-family:"Calibri",sans-serif}
@page WordSection1
{margin:70.85pt 70.85pt 2.0cm 70.85pt}
div.WordSection1
{}--></style>
</head>
<body dir="ltr" style="font-size:12pt;color:#000000;background-color:#FFFFFF;font-family:Calibri,Arial,Helvetica,sans-serif;">
<p>Hi Mathieu,</p>
<p><br>
</p>
<p>thank you for the explanation! </p>
<p><br>
</p>
<p>This would suggest that the problem only occurs for scheduler events, because only then we have this situation that the currently scheduled task is not the one that the tracepoint relates that. Am I interpreting that correctly?<br>
</p>
<p><br>
</p>
<p>If yes, this would certainly ease my concerns, because for most tracepoints, I get the expected behavior. I can also get the information that I want right now by looking at comm instead.<br>
</p>
<p><br>
</p>
<p>However, for future development of LTTng: Isn't the intuitive context that of the *tracepoint*? In other words, isn't it wrong to always use the currently scheduled task?<br>
</p>
<p><br>
</p>
<p>cheers,<br>
</p>
<p><br>
</p>
<div id="Signature">
<div name="divtagdefaultwrapper" style="font-family:Calibri,Arial,Helvetica,sans-serif; font-size:; margin:0">
<font color="black" face="Calibri,Arial,Helvetica,sans-serif" size="3"><span dir="ltr" style="font-size:12pt; background-color:white"><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="de"><b>Dr.
</b></span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="en-US"><b>Ingo Lütkebohle</b></span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="en-US"><b><br>
</b></span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="en-US">Software Design and Analysis -- Robotics (CR/AEA2)
</span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="en-US"><br>
</span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="en-US"><br>
Tel. </span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="de">+49(711)811-12248
</span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="de"><br>
Fax +49(711)811-0 </span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="de"><br>
Mobil +49-1525-8813417 </span></font><font face="Arial,sans-serif" size="2"><span style="font-size:10pt" lang="de"><br>
</span></font><a id="NoLP" target="_blank" href="mailto:Ingo.Luetkebohle@de.bosch.com" tabindex="0"><font face="Calibri,sans-serif" size="2"><span style="font-size:11pt" lang="de"><font face="Arial,sans-serif" size="2"><span style="font-size:10pt">Ingo.Luetkebohle@de.bosch.com</span></font></span></font></a></span></font></div>
</div>
<div style="color: rgb(33, 33, 33);">
<hr tabindex="-1" style="display:inline-block; width:98%">
<div id="divRplyFwdMsg" dir="ltr"><font style="font-size:11pt" color="#000000" face="Calibri, sans-serif"><b>Von:</b> Mathieu Desnoyers <mathieu.desnoyers@efficios.com><br>
<b>Gesendet:</b> Donnerstag, 26. November 2015 16:21<br>
<b>An:</b> Luetkebohle Ingo (CR/AEA2)<br>
<b>Cc:</b> lttng-dev<br>
<b>Betreff:</b> Re: [lttng-dev] faulty context for sched_stat_ tracepoints?</font>
<div> </div>
</div>
<div>
<div style="font-family:arial,helvetica,sans-serif; font-size:12pt; color:#000000">
<div><span id="zwchr">---- On Nov 26, 2015, at 7:34 AM, Luetkebohle Ingo (CR/AEA2) <Ingo.Luetkebohle@de.bosch.com> wrote:<br>
</span></div>
<div>
<blockquote style="border-left:2px solid #1010FF; margin-left:5px; padding-left:5px; color:#000; font-weight:normal; font-style:normal; text-decoration:none; font-family:Helvetica,Arial,sans-serif; font-size:12pt">
<style>
<!--
@font-face
{font-family:"Cambria Math"}
@font-face
{font-family:Calibri}
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
margin-bottom:.0001pt;
font-size:11.0pt;
font-family:"Calibri",sans-serif}
a:link, span.MsoHyperlink
{color:#0563C1;
text-decoration:underline}
a:visited, span.MsoHyperlinkFollowed
{color:#954F72;
text-decoration:underline}
span.EmailStyle17
{font-family:"Arial",sans-serif;
color:windowtext;
font-weight:normal;
font-style:normal}
.MsoChpDefault
{font-family:"Calibri",sans-serif}
@page WordSection1
{margin:70.85pt 70.85pt 2.0cm 70.85pt}
div.WordSection1
{}
-->
</style>
<div class="WordSection1">
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">Hi,</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US"> </span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">I noticed that the “procname” and “tid” *<b>context</b>* variables can differ from the “comm” and “tid” variables of the sched_stat_-tracepoints (e.g., for sched_stat_runtime).</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US"> </span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">For example, see the following two outputs:</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">[15:49:33.674072890] (+0.000029771) rng2070 sched_stat_sleep: { cpu_id = 4 }, { pid = 26974, procname = "amcl", tid = 26975 }, { comm = "lttng-sessiond", tid =
798, delay = 194122524 }</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">[15:49:33.722894563] (+0.000004851) rng2070 sched_stat_sleep: { cpu_id = 0 }, { pid = 0, procname = "swapper/0", tid = 0 }, { comm = "amcl", tid = 27078, delay
= 10043658 }</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US"> </span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">In both cases, procname differs from comm. I’ve looked at the kernel source code, and the “comm” variable is taken from task_struct.filename. Barring renames,
this would be the correct one, so I’m assuming that the context is sometimes incorrect.</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US"> </span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">Now, for this particular case, I can take the correct variables from the tracepoint, but of course, this makes me worried that the context might also be wrong in
other cases… Do you have any idea what could be going on here?</span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US"> </span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US">LTTng is Version 2.7.x+stable+bzr2948+pack17+201511192116~ubuntu14.04.1 and this is on a 3.19.0-33-generic kernel (btw, that’s not the default kernel for Ubuntu
14.04, but the LTS vivid backport kernel).</span></p>
</div>
</blockquote>
<div>Hi,<br>
</div>
<div><br>
</div>
<div>Looking at lttng-modules lttng-context-tid.c tid_record(), we see that it uses task_pid_nr(current).<br>
</div>
<div>It therefore gets the tid of the currently scheduled task. This is similar for procname.<br>
</div>
<div><br>
</div>
<div>Now looking at the Linux kernel kernel/sched/fair.c: enqueue_sleeper(), which calls<br>
</div>
<div>trace_sched_stat_sleep(), we see that it passes "tsk" as parameter to the tracepoint.<br>
</div>
<div>Already here, nothing seems to ensure that "tsk" is the currently scheduled task. It's<br>
</div>
<div>rather a task that is being put to sleep. Looking at the callers of enqueue_sleeper(),<br>
</div>
<div>we can understand that the current task is first scheduled out before being enqueued<br>
</div>
<div>as a sleeper. This is why, when we hit the sched_stat_sleep tracepoint, the current<br>
</div>
<div>task is not "tsk", which explains the difference between the context information<br>
</div>
<div>and the tracepoint payload information: they really target different tasks.<br>
</div>
<div><br>
</div>
<div>Does this help solve your concerns ?<br>
</div>
<div><br>
</div>
<div>Thanks,<br>
</div>
<div><br>
</div>
<div>Mathieu<br>
</div>
<div><br>
</div>
<blockquote style="border-left:2px solid #1010FF; margin-left:5px; padding-left:5px; color:#000; font-weight:normal; font-style:normal; text-decoration:none; font-family:Helvetica,Arial,sans-serif; font-size:12pt">
<div class="WordSection1">
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif" lang="EN-US"> </span></p>
<p class="MsoNormal"><span style="font-size:10.0pt; font-family:"Arial",sans-serif; color:black" lang="EN-US">Mit freundlichen Grüßen / Best regards
<br>
<br>
<b>Ingo Luetkebohle<br>
</b><br>
Application Software (CR/AEA2) <br>
Tel. +49(711)811-12248 | Fax +49(711)811-0 | </span><span style="font-size:10.0pt; font-family:"Arial",sans-serif; color:black"><a href="mailto:Ingo.Luetkebohle@de.bosch.com" target="_blank"><span style="color:blue" lang="EN-US">Ingo.Luetkebohle@de.bosch.com</span></a></span><span style="font-size:10.0pt; font-family:"Arial",sans-serif; color:black">
</span><span style="font-size:8.0pt; font-family:"Arial",sans-serif; color:black" lang="EN-US"><br>
</span><span style="font-size:10.0pt; font-family:"Arial",sans-serif; color:black" lang="EN-US"><br>
<br>
</span></p>
</div>
<br>
_______________________________________________<br>
lttng-dev mailing list<br>
lttng-dev@lists.lttng.org<br>
http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev<br>
</blockquote>
</div>
<div><br>
</div>
<div>-- <br>
</div>
<div>Mathieu Desnoyers<br>
EfficiOS Inc.<br>
http://www.efficios.com</div>
</div>
</div>
</div>
</body>
</html>