Differences

This shows you the differences between two versions of the page.

--- realtime:documentation:howto:debugging:smi-latency:cyclictest-tracing [2018/08/22 10:27]
ebugden [Tracing] Put all ftrace option modification stuff together
+++ realtime:documentation:howto:debugging:smi-latency:cyclictest-tracing [2018/08/23 15:24] (current)
ebugden [Tracing] Add warning about cyclictest smi count
@@ Line 16: / Line 16: @@
 Once the CPUs are sure to never idle, a trace can be obtained using Cyclictest with the break trace and the function tracing options. The break trace option (--breaktrace) will stop tracing and end Cyclictest if it detects a latency that is longer than a specified limit. This option is intended to be used with one of the Cyclictest options that enables a specific Ftrace tracer. For this SMI detection technique, the function tracing option (--ftrace) is used so that the trace includes all the kernel function calls.
+The steps for obtaining a relevant trace using Cyclictest are explained in detail [[realtime:documentation:howto:tools:cyclictest:tracing|here]]. Since the desired tracer (function tracer) has already been selected, the section on the linked page about selecting instrumentation can be skipped.
 <WRAP center round info 60%>
-Ftrace might not behave the same way after using the Cyclictest break trace option.
+When using the breaktrace option along with the [[realtime:documentation:howto:tools:cyclictest:options:smi-count|SMI count]] option, the SMI count shown in the Cyclictest output may not increment even if the tracing was stopped after a latency caused by an SMI. This is probably because Cyclictest is stopped before the visual output of the SMI count is updated.
-When this option is used, Cyclictest changes some Ftrace options to get the desired trace format. The Ftrace options that were set before are not restored after Cyclictest runs.
-</WRAP>
-Below is an example of a Cyclictest command that includes the break trace and function tracing options. The '#' before the command is to indicate that the command must be run as root, with sudo, or as a member of the realtime group.
-<code bash>
-# cyclictest --mlockall --smp --priority=80 --interval=200 --distance=0 --breaktrace=150 --ftrace
-</code>
-The combination of options in this example is not appropriate for every test situation. More details about how to choose the right options for measuring a specific latency on a given system can be found in the Cyclictest documentation [[realtime:documentation:howto:tools:cyclictest:test-design#option-selection|here]].
-<WRAP center round important 60%>
-Remember to adjust for the tracing overhead when specifying the Cyclictest break trace limit.
 </WRAP>
-Function tracing slows down the execution of the program, so the target latency will most likely be longer after enabling tracing. It is important to adjust for this additional overhead because otherwise the tracing may not break on the correct latency and the trace may not contain the latency that was supposed to be observed. More details about how to adjust the break trace limit can be found in the documentation about debugging using Cyclictest [[realtime:documentation:howto:tools:cyclictest:tracing#measuring-the-tracing-overhead|here]].
+To make sure that the tracing stopped at the correct time, look at the break value specified below the Cyclictest output after the test stops. This value should correspond to the expected SMI related latency value even if the SMI count on the Cyclictest output is not updated. The latency should also have occurred in the same conditions as the expected SMI latency.
 ===== Analysis =====

Wiki

User Tools

Site Tools

Differences

Page Tools