[ltt-dev] CONFIG_HAVE_TRACE_CLOCK_GENERIC timestamp problem on PXA255

Steve Langstaff steve.langstaff at pebblebay.com
Fri May 29 07:27:56 EDT 2009


I have built LTTng for kernel 2.6.30-rc5 + patch set 0.137 using
CONFIG_HAVE_TRACE_CLOCK_GENERIC on my PXA255.

All of the LTTng stuff is configured to be built-in to the kernel, rather
than loadable modules.

I am having trouble with the timestamps that are logged in the traces - they
don't seem to 'tick' at the right rate.

Here is my test program... it starts a trace, logs a START user event,
sleeps for 1 second, logs a STOP user event and then stops the trace:

/tmp # cat sleeper.sh
#!/bin/sh
lttctl -C -w /tmp/sleeper sleeper
echo `date` START > /tmp/debugfs/ltt/write_event
sleep 1
echo `date` STOP >/tmp/debugfs/ltt/write_event
lttctl -D sleeper

----

Here is the test program running:

/tmp # ./sleeper.sh
Linux Trace Toolkit Trace Control 0.67-05032009

Controlling trace : sleeper

lttctl: Creating trace
lttctl: Forking lttd
Linux Trace Toolkit Trace Daemon 0.67-05032009

Reading from debugfs directory : /tmp/debugfs/ltt/sleeper
Writing to trace directory : /tmp/sleeper

lttctl: Starting trace
Linux Trace Toolkit Trace Control 0.67-05032009

Controlling trace : sleeper

lttctl: Pausing trace
lttctl: Destroying trace

----

I then tar up the trace directory and copy it to my host machine - that
shouldn't be a problem because the traces are self-describing, right?

Here are the timestamps that I see when I examine the trace with lttv -m
textDump sleeper | fgrep UTC

userspace.event: 4661.842551269
(/home/steve/Projects/TEMP/sleeper/userspace_0), 351, 351, sleeper.sh, ,
259, 0x0, SYSCALL { string = "Thu Jan 1 01:33:27 UTC 1970 START" }

userspace.event: 4661.844299316
(/home/steve/Projects/TEMP/sleeper/userspace_0), 351, 351, sleeper.sh, ,
259, 0x0, SYSCALL { string = "Thu Jan 1 01:33:28 UTC 1970 STOP" }

See how the 'seconds' field is stuck at 4661 - this appears to be the case
on all traces that I take.

Any ideas where I could start looking?

Am I just flogging a dead horse trying to get meaningful data from LTTng on
a PXA255?






More information about the lttng-dev mailing list