Below is a patch for git.lttng.org/ltt-control/liblttctl/lttctl.h. I hope I
have the patch-submission procedure and mailing list format right.
I have two other large patches to submit, for liblttctl/liblttctl.c and
lttctl/lttctl.c (more to follow), both about 1500 lines because the code
--
Date: Thu, 15 Dec 2011 14:18:23 -0500
From: Yannick Brosseau yannick.bross...@gmail.com
The format is better, thank you. Would it be possible for you to split
the patch in several git commit and thus into several different
This is very odd...
The DEBUG1: SIGCHLD catched [in sighandler() at lttng.c:179] shows you that
when this signal is caught by the lttng client, it means that the session
daemon
is ready hence the lttng modules are loaded for sure. The session daemon send
this signal only when it's ready to
You have the exact same output as mine.
51 events is correct and the return data size is 41224. We'll have to add
debug
output. Let's check the event name returned by the kernel and those we send
back. Please add this to the git head of lttng-tools:
+++ b/src/bin/lttng-sessiond/kernel.c
Of potential interest is the session log when I proceeded to kill
lttng-sessiond (which had processed only the list -k command):
DEBUG1: SIGTERM catched [in sighandler() at main.c:4179]
DEBUG1: Terminating all threads [in stop_threads() at main.c:400]
DEBUG1: Futex n to 1 wake done [in
-Message d'origine-
De : David Goulet [mailto:david.gou...@polymtl.ca]
Envoyé : 23 janvier 2012 17:43
That's very interesting.
Quick fix, let see if it works:
diff --git a/src/common/sessiond-comm/sessiond-comm.c
b/src/common/sessiond-comm/sessiond-comm.c
index 03e8931..106041f
This tiny patch fixes a small typo in lttng-tools-2.0-pre18+-4666e71
src/bin/lttng/conf.c:
diff --git a/src/bin/lttng/conf.c b/src/bin/lttng/conf.c
index d326378..4c3b1fd 100644
--- a/src/bin/lttng/conf.c
+++ b/src/bin/lttng/conf.c
@@ -171,7 +171,7 @@
fp = open_config(path, r);
lttng-tools lttng.c check_sessiond(): Improve error messages
check_args_no_sessiond(): Add --list-commands to the no daemon options
parse_args(): --help, --list-commands and --list-options yield normal exits,
not errors
main(): punctuation in comment
--
From
--
Message: 6
Date: Wed, 25 Jan 2012 14:19:34 -0500
From: Mathieu Desnoyers comp...@krystal.dyndns.org
Subject: Re: [lttng-dev] [PATCH] Improve lttng-tools lttng.c usage()
Message-ID: 20120125191934.GC10043@Krystal
- fprintf(ofp, -h, --help Show this
--
From ff37e87b82cc1e8e08d9a45f7097c07bc768b86c Wed, 25 Jan 2012 15:04:35 -0500
From: Daniel U. Thibault daniel.thiba...@drdc-rddc.gc.ca
Date: Wed, 25 Jan 2012 15:03:31 -0500
Subject: [PATCH] Adding a DBG2 to lttng/utils.c get_session_name()
diff --git
Document and ensure the return values of the cmd_add_context() chain of
functions.
Make add_context() stop looping over lttng_add_context() upon failure
(otherwise it reports success or failure depending only on the last call's
result).
Also patches lib/lttng-ctl/lttng-ctl.c
-Message d'origine-
De : David Goulet [mailto:dgou...@efficios.com]
Envoyé : 27 janvier 2012 10:51
+ /*
+* lttng_add_context() returns the size
+* of the returned payload data
+* or a negative error code.
+*/
I did not
From e85aacb4c42fc522fce58971abc85ed2fed70123 Tue, 31 Jan 2012 14:31:56 -0500
From: Daniel U. Thibault daniel.thiba...@drdc-rddc.gc.ca
Date: Tue, 31 Jan 2012 14:31:47 -0500
Subject: [PATCH] create.c : Fix memory leak
Signed-off-by: Daniel U. Thibault daniel.thiba...@drdc-rddc.gc.ca
diff --git
In lttng-tools lttng/lttng.h, we have:
struct lttng_domain {
...
char exec_name[NAME_MAX];
...
}
While in lttng-tools src/bin/lttng-sessiond/trace-ust.h, we have:
struct ltt_ust_domain_exec {
char exec_name[LTTNG_UST_SYM_NAME_LEN];
...
}
If the
I would like to suggest adding to lttng.h an lttng_domain_type enum value of
LTTNG_DOMAIN_UNSPECIFIED = 0.
enum lttng_domain_type {
+ LTTNG_DOMAIN_UNSPECIFIED = 0,
LTTNG_DOMAIN_KERNEL = 1,
This would simplify handling of domain option switches
Date: Fri, 10 Feb 2012 10:28:06 -0500
From: Mathieu Desnoyers mathieu.desnoy...@efficios.com
Subject: Re: [lttng-dev] [PATCH] Enforce exactly oneLTTNG_DOMAIN
(add_context.c) (updated) (REPLACES PREVIOUS PATCH)
The check for more than one domain should be done when parsing the -u /
-k
-Message d'origine-
Message: 5
Date: Fri, 2 Mar 2012 18:18:07 -0500
From: Mathieu Desnoyers mathieu.desnoy...@efficios.com
Subject: Re: [lttng-dev] A problem
Fix: meaningful error message
Show a meaningful error message to users attempting to do kernel tracing
without
This is really weird. I installed lttng on three Ubuntu 12.04 LTS systems
(two of which are virtual machines), and on one of the VMs the install appears
fine but the lttng modules never load. I get no error messages whatsoever.
Any ideas why?
Daniel U. Thibault
R D pour la défense
I have this Ubuntu 12.04 machine with the lttng-tools suite installed, and I
now want to quickstart userspace tracing. Starting at
http://lttng.org/quickstart, we're told succinctly that we need to instrument
the target application (details in the lttng-ust(3) manpage) and then do the
-Message d'origine-
Envoyé : 11 janvier 2013 17:45
Turning to the Internet, I find
https://github.com/giraldeau/lttng-ust/tree/master/doc/examples/easy-ust
and grab the handful of files found there.
I then open a console within that directory and invoke make:
Please use an
Interestingly, if I do install lttng-ust 2.0.2 (a367ee6) 2013-Apr-18,
easy-ust's make compiles correctly. One key difference is that
/usr[/local]/include/lttng/tracepoint.h now includes an enum setting the
TRACE_WARNING value.
The conclusion is inescapable: the Ubuntu liblttng-ust-dev
I need help to get lttng-tools to compile.
I have grabbed the (nearly) latest git.lttng.org packages:
lttng-modules (2.1.0+-c6403de)
userspace-rcu (0.7.6-da9bed2)
lttng-ust (2.1.0+-340f776) (one new commit since)
lttng-tools (2.1.1+-281047b) (six new commits since)
The userspace-rcu
-Message d'origine-
De : David Goulet [mailto:dgou...@efficios.com]
Envoyé : 18 janvier 2013 13:30
Make sure also you have these packages:
- flex = 2.5.35
- bison = 2.4
David
-Fin du message d'origine-
Flex was the missing package (far, far from obvious). It works now!
-Message d'origine-
De : Mathieu Desnoyers [mailto:mathieu.desnoy...@efficios.com]
Envoyé : 18 janvier 2013 14:15
Flex was the missing package (far, far from obvious). It works now!
Please read the README files found in each project root directory before
building. These document,
-Message d'origine-
Quoting the lttng-tools README file:
- liburcu
Userspace RCU library, by Mathieu Desnoyers and Paul E. McKenney
- Tested with liburcu = v0.7.2
Therefore you need to install Userspace RCU 0.7.2 or later.
A dependency is not only on a package, but also on
-Message d'origine-
De : Mathieu Desnoyers [mailto:mathieu.desnoy...@efficios.com]
Envoyé : 21 janvier 2013 10:30
No. Distribution packages usually lag behind the project upstream.
Ubuntu 12.04 has been released last year, and their policy don't allow them to
update the packages within
-Message d'origine-
De : David Goulet [mailto:dgou...@efficios.com]
Envoyé : 21 janvier 2013 10:53
Also, I do still forget it still today but after installing URCU and UST:
$ sudo ldconfig
David
-Fin du message d'origine-
But 'sudo make install' invokes 'ldconfig -n
-Message d'origine-
De : Mathieu Desnoyers [mailto:mathieu.desnoy...@efficios.com]
Envoyé : 21 janvier 2013 12:07
Even though libuuid1 (includes libuuid-perl, uuid-dev, uuid-runtime) is
installed. I need to fix this new stumbling block before we can address the
liburcu-cds problem.
My two Ubuntu installations have the same set of LTTng modules and tools
(they were all built from the same git.lttng.org packages) but different
kernels and installation histories. The clean machine has kernel
linux-virtual (3.2.0-36-virtual), the other one has linux-generic
To answer my own earlier question, the Ubuntu lttng packages (but not the
git.lttng.org packages) automate the launch of the lttng-sessiond daemon
through the Upstart service, by adding /etc/init/lttng-sessiond.conf:
description LTTng 2.0 central tracing registry session daemon
author
-Message d'origine-
Date: Mon, 21 Jan 2013 14:34:29 -0500
Subject: Re: [lttng-dev] Odd dependency for lttng-ust
-Message d'origine (intérieur)-
De?: Mathieu Desnoyers [mailto:mathieu.desnoy...@efficios.com]
Envoy??: 21 janvier 2013 13:39
As the error messages suggests, please
-Message d'origine-
De : Alexandre Montplaisir [mailto:alexmon...@voxpopuli.im]
Envoyé : 25 janvier 2013 12:57
On 13-01-25 12:32 PM, Thibault, Daniel wrote:
[...]
Until this old gold bug is fixed (which looks unlikely considering its
age), maybe each package's configure.ac could
The default value of the option is 'yes' (re: configure.ac), which means
the fixup is applied. Setting the option to another value, such as 'no',
means the fixup is not applied.
Why then is the option named *disable*-libtool-linkdep-fixup? Shouldn't it
be named
The lttng man pages are a little vague as to the relationship between lttng,
the lttng-sessiond, and the tracing group. My experiments have shown that a
user belonging to the tracing group should more or less be able to issue lttng
commands as if they were prefixed with sudo, but that's not
-Message d'origine-
De : christian.bab...@0x80.ca [mailto:christian.bab...@0x80.ca] De la part de
Christian Babeux
Envoyé : 6 février 2013 12:52
First observation
I'm not able to reproduce your first observation on my machine.
[...]
$ sudo lttng -vvv create rootlevel
I'm operating under the assumption that an event's fully qualified
identifier is something like session:domain:channel:name. Tracepoints and
syscalls have system-defined names, and of course the domain can only be
'kernel' or 'userspace', but everything else is pretty much arbitrary. My
According to
http://git.dorsal.polymtl.ca/?p=lttngtop.git;a=shortlog;h=refs/heads/live
README (2012-Aug-30), one needs special versions of the lttng-tools and
babeltrace packages in order to experiment with the live lttngtop alpha/beta.
For technical reasons, I can't use git: directly and
-Message d'origine-
De : Julien Desfossez [mailto:jdesfos...@efficios.com]
Envoyé : 21 février 2013 11:16
The gitwebs for these branches are :
http://git.dorsal.polymtl.ca/~jdesfossez?p=lttng-tools;a=shortlog;h=refs/heads/lttngtop-live
check_field_requirements() reads in part:
if (*tid_check == 0) {
if (strncmp(name, tid, 3) == 0)
(*tid_check)++;
}
if (*pid_check == 0) {
if (strncmp(name, tid, 3) == 0)
(*pid_check)++;
}
Shouldn't that second if check for pid instead of tid?
The verbose output of babeltrace (-v, --verbose) is missing its closing
comment. Looking at babeltrace.c, line 670:
fmt_write-close_trace(td_write);
bt_context_put(ctx);
printf_verbose(finished converting. Output written to:\n%s\n,
opt_output_path ? : stdout);
goto end;
I
--
Message: 1
Date: Fri, 22 Feb 2013 20:13:15 +
From: Thibault, Daniel daniel.thiba...@drdc-rddc.gc.ca
Subject: [lttng-dev] babeltrace omits its closing verbose comment
The verbose output of babeltrace (-v, --verbose
Is anyone else bothered by the slight irregularity of babeltrace's output?
Here is the same trace record printed using various babeltrace settings. In
all cases I used '-f trace:hostname --no-delta', varying only the '-n' part:
-n scope
[11:02:08.064705355] edge-vb-u12 exit_syscall:
The last few times I rebuilt the lttng-ust package, make's log contains some
ominous complaints, although it seems to nevertheless complete normally.
Here's an excerpt:
## Begin ##
CDPATH=${ZSH_VERSION+.}: cd . /bin/bash
I'm running an up-to-date LTTng suite (except for lttng-ust, where I'm using
the 5 Mar 2013 20:17:22 commit). I've compiled the
lttng-ust/doc/examples/easy-ust/sample application. While I run it (as a user
*not* belonging to the tracing group), I notice something unexpected:
* The local
@@ -78,6 +80,8 @@ static struct poptOption long_options[] = {
{read-timer, 0, POPT_ARG_INT, 0, OPT_READ_TIMER, 0, 0},
{list-options, 0, POPT_ARG_NONE, NULL, OPT_LIST_OPTIONS, NULL,
NULL},
{output, 0, POPT_ARG_STRING, opt_output, 0, 0, 0},
+
In the lttng-ust man pages, there is a paragraph that reads:
Note about dlopen() usage: Due to locking side-effects caused by the way libc
lazily resolves Thread-Local Storage (TLS) symbols when a library is dlopened,
linking the tracepoint probe or liblttng-ust with dlopen() is discouraged.
Message: 1
Date: Tue, 19 Mar 2013 11:51:29 -0400
I don't see this warning and error here. Which release are you using?
lttng-modules 2.1.1+ (74b9312, 2013-02-27)
lttng-tools 2.1.1+ (e58727c, 2013-03-01)
lttng-ust 2.1.1+ (009769c, 2013-03-05)
userspace-rcu 0.7.6+ (108a92e,
-Message d'origine-
Message: 6
Date: Wed, 20 Mar 2013 12:38:59 -0400
Meanwhile I've been playing with some very simple examples of .so use, and
found that the behaviour of LD_PRELOAD is incompletely documented. For
instance, if I run:
Take a look at man 8 ld.so. It should answer
In my previous note on this mailing list, forget the bit about LD_PRELOAD
being beholden to the remaining library search. I got fooled into thinking
that for unrelated reasons. Turns out LD_PRELOAD loads the so regardless of
the search's outcome: it's just added to the list of dependencies
Here are proposed semantics for the tracefile-size/tracefile-count
enable-channel options (are they still planned for later inclusion in
lttng-tools?):
-C, --tracefile-size SIZE
This is a recent proposal (2013-Mar-08) that has yet to be added to the
command. Maximum size of each
@@ -179,6 +193,15 @@ static int enable_channel(char *session_name)
set_default_attr(dom);
+ if ((chan.attr.tracefile_size 0)
+ (chan.attr.tracefile_size chan.attr.subbuf_size)) {
+ ERR(Tracefile_size must be superior or equal to subbuf_size
--subbuf-size accepts sizes such as:
- 123 - 123
- 123k or 123K - 123 * 1024
- 123m or 123M - 123 * 1024 * 1024
- 123g or 123G - 123 * 1024 * 1024 * 1024
I don't really like using non-SI prefixes such as K, m or g
(well, in fact m is an SI prefix but for milli, not mega).
I think that
The Google Summer of Code LTTng entry
(http://bugs.lttng.org/projects/lttng/wiki/Google_Summer_of_Code_2013) is a
promising initiative. May I propose another little project idea?
Babeltrace support for the debugfs (/sys/kernel/debug/tracing) trace format
Brief explanation:
Provide a
, Thibault, Daniel
daniel.thiba...@drdc-rddc.gc.camailto:daniel.thiba...@drdc-rddc.gc.ca
wrote:
Even if the debugfs tracing facility is destined for eventual withdrawal
from the Linux kernel, it may be of interest to some to recover older traces
generated in this way -if only in order
I have now looked at download page http://lttng.org/download
Below section and links appears to be relevant for me.
LTTng 2.1 Basse Messehttp://lttng.org/version-naming - Tracer toolchain
LTTng-tools
2.1.1http://lttng.org/files/lttng-tools/lttng-tools-2.1.1.tar.bz2
(Changelog)
Absolutely: no patching required.
I decided to find out if the lttng-modules were required at all if one needs
to do only user-space tracing.
I installed (on an otherwise lttng-virgin system) just the userspace-rcu,
lttng-ust, and lttng-tools packages. The lttng commands fail for the
did you run ldconfig on the machine without lttng-modules ?
Not the first time, I must admit. I had forgotten, silly me. That was why
I had to use LD_LIBRARY_PATH or LD_PRELOAD. With ldconfig done, there is no
need to do so.
does this machine recognise libraries in /usr/local
I was fooling around with checkinstall in order to create installation
packages for lttng-modules, userspace-rcu, etc., when I ran into what seems a
very obscure bug (googling for the error message unable to create
`/proc/self/coredump_filter.dpkg-new' brings up nothing).
It occurs when
-Message d'origine-
De : Alexandre Montplaisir [mailto:alexmon...@voxpopuli.im]
Envoyé : 24 avril 2013 17:48
On 13-04-24 05:25 PM, Thibault, Daniel wrote:
dpkg: error processing
/home/daniel/Documents/git.lttng.org/lttng-ust-2.2.0-rc1+-a16877a-Java/lttng-ust-java_2.2.0-rc1-12_amd64
Message: 1
Date: Fri, 3 May 2013 10:42:22 -0400
There are now two tracing modes implemented in lttng-ust (user-space tracer),
each with their own advantages (+)/disadvantages(-):
1) per-pid buffer tracing,
[...]
2) per-uid (shared) buffer tracing, (buffers shared across applications
Let's assume I want to do a kernel-only LTTng installation. After putting
lttng-modules in place, I get lttng-tools and, at its configure step, I do:
./configure --disable-lttng-ust
However, this fails with:
checking whether cds_list_add is declared... no
configure: error: liburcu =
I ran a little experiment whose results surprised me. Using a non-tracing
group user account, I first made sure I had a root lttng-sessiond running and a
user-space lttng-sessiond. In the first I set up a trace for some kernel
events plus all user-space events. In the second I set up a
(Question: If there were several users active at once, would there be
several root user-space lttng-consumerd daemons or still just one?)
no
Please clarify: no, there will be several root user-space lttng-consumerd
daemons or no, there will be just one root user-space lttng-consumerd
Date: Wed, 22 May 2013 10:04:41 -0400
From: Matthew Khouzam matthew.khou...@ericsson.com
I was given an interesting corner case and want to explain how I believe it
should be handled.
You are given a trace with no timestamps, and n channels. How do you display
the events?
I propose (and have
I tried deploying the tarball to another directory and building. The tests
refuse to run (Could not execute (unit/test_kernel_data): open3: exec of
unit/test_kernmel_data failed at
/usr/share/perl/5.14/TAP/Parser/Iterator/Process.pm line 168) until after the
make step (naturally). This
In lttng-tools/src/bin/lttng/lttng.c, the check_args_no_sessiond function
checks the command-line arguments for specific options and commands that must
not trigger a session daemon. The function detects the options -h, --h
(for --help), --list-options, and --list-commands, and the commands
Harking back to Mathieu's post of 3 May 2013 10:42:22, I'd like to see if I
understand the options correctly. I'm pretty sure I don't.
When a channel has been set to use per-UID buffering, if an event-producing
user-space process dies suddenly it may create a hole in one of the channel's
Inspired by Jérémie's https://gist.github.com/jgalar/5706595, I tinkered
with lttng-ust/doc/examples/easy-ust to make the sample application use a small
C++ object (and turning itself into sample.cpp). It links and works fine
except if LTTNG_UST_HAVE_SDT_INTEGRATION is #defined, in which
-Message d'origine-
1. Re: [PATCH lttng-ust] Fix: Check g++ version before compiling
'hello.cxx' test (Christian Babeux)
2. Re: [PATCH lttng-ust] Fix: Check g++ version before compiling
'hello.cxx' test (Mathieu Desnoyers)
-Message d'origine-
Message: 7
Date: Mon, 24 Jun 2013 17:28:26 +0300
From: Avi Goren a...@reduxio.com
I'm using lttng-ust and in the process of the attempt to gain some flush
frequency control I encountered a problem.
I'm having difficulties with the use of enable-channel.
it doesn't seem
-Message d'origine-
Date: Tue, 25 Jun 2013 09:49:36 +0300
From: Avi Goren a...@reduxio.com
I'm using lttng 2.2.0 and encountered an infinite loop:
-
# lttng create
Session auto-20130624-165047 created.
Traces will be
-Message d'origine-
Message: 5
Date: Tue, 25 Jun 2013 17:37:57 -0400
From: David Goulet dgou...@efficios.com
[...]
Per UID buffers makes every application sharing the same UID to use the same
buffers.
[...]
Also, the buffers for UST application are now created in the consumer
and passed to
-Message d'origine-
Per UID buffers makes every application sharing the same UID to use the same
buffers.
[...]
Also, the buffers for UST application are now created in the consumer and
passed to the application. Thus, the buffer ownership has been transfered from
the UST tracer to the
I was reading Michael Kerrisk's Namespaces in operation
(http://lwn.net/Articles/531114/), which explains most of the namespace
functionalities, and this brings to mind a number of questions.
Are trace paths namespace-ready? Per-PID traces are apparently not much at
risk since a process
In the process of hunting down what is going on with a bug that occurs on
one machine but not on the other, I ran into this:
$ sudo -H lttng -vvv create hsession
DEBUG3 [30339/30339]: URI string:
file:///root/lttng-traces/hsession-20130704-112000 (in uri_parse() at uri.c:291)
DEBUG3
-Message d'origine-
Date: Thu, 4 Jul 2013 16:27:25 +
From: Thibault, Daniel daniel.thiba...@drdc-rddc.gc.ca
To: lttng-dev@lists.lttng.org lttng-dev@lists.lttng.org
[...] what strikes me as odd is the lsm cmd type value: it is
LTTNG_ENABLE_EVENT (6) in both cases. Shouldn't
Date: Mon, 8 Jul 2013 12:33:41 -0400
From: Mathieu Desnoyers mathieu.desnoy...@efficios.com
* Thibault, Daniel (daniel.thiba...@drdc-rddc.gc.ca) wrote:
But what of per-UID traces? What happens to a trace in progress if
one or more processes switch user namespaces?
nothing
--
Date: Tue, 09 Jul 2013 14:32:54 -0400
From: Mohamad Gebai mohamad.ge...@polymtl.ca
To: linuxtools-...@eclipse.org, lttng-dev@lists.lttng.org
For the sake of our experience, we pinned vCPU0 of VM1 and vCPU0 of VM2 on
the
The lttng command has a --group option which the --help describes succinctly
as the Unix tracing group name. I have yet to manage to get this to work.
Is there a simple use case that showcases this?
I'm looking for a use case where 'lttng list' spawns a local session daemon
while
lttng-tools-2.2.0\src\bin\lttng-sessiond\trace-kernel.c, lines 190-198:
case LTTNG_EVENT_FUNCTION:
attr-instrumentation = LTTNG_KERNEL_KRETPROBE;
attr-u.kretprobe.addr = ev-attr.probe.addr;
attr-u.kretprobe.offset = ev-attr.probe.offset;
In LTTng 2.3's src/bin/lttng/commands/snapshot.c, we have:
static int add_output(const char *url)
{
int ret;
struct lttng_snapshot_output *output = NULL;
if (!url (!opt_data_url || !opt_ctrl_url)) {
ret = CMD_ERROR;
goto error;
--
Date: Fri, 26 Jul 2013 23:40:33 +0800
From: Zifei Tong soar...@gmail.com
Command Line Interface
--
To enable a dynamic probe in a running process at given address, you can use:
lttng enable-event
-Message d'origine-
Envoyé : 29 juillet 2013 10:20
On Mon, Jul 29, 2013 at 9:49 AM, Thibault, Daniel
daniel.thiba...@drdc-rddc.gc.ca wrote:
--
Unless I'm mistaken, currently lttng's 'enable-event -u ...' works
-Message d'origine-
Date: Wed, 31 Jul 2013 14:29:12 -0400
From: Simon Marchi simon.mar...@polymtl.ca
I encountered a problem while trying to put UST tracepoints in GDB
today. Since GDB defines a struct tracepoint, it conflicts with
LTTng's struct tracepoint. Renaming all struct tracepoint
--
Date: Wed, 7 Aug 2013 12:48:53 -0700
From: Vomlehn vomle...@amazon.com
-- ./configure
-- make
-- make install
+ $ ./configure
+ $ make
+ $ make install
+
+To compile the source tree from a git checkout, use:
+
+ $ autoreconf
+ $ automake --add-missing --gnu
+
-Message d'origine-
De : Mathieu Desnoyers [mailto:mathieu.desnoy...@efficios.com]
Envoyé : 8 août 2013 11:22
* Thibault, Daniel (daniel.thiba...@drdc-rddc.gc.ca) wrote:
Unless there's been a change in the code, that is just not true. You can
trace the kernel just fine without
--
Message: 5
Date: Fri, 9 Aug 2013 10:54:49 -0400
From: J?r?mie Galarneau jeremie.galarn...@efficios.com
On Fri, Aug 9, 2013 at 4:01 AM, Stanislav Vovk stanislav.v...@ericsson.com
wrote:
In my system there are two users,
-Message d'origine-
Envoyé : 12 août 2013 16:11
On Mon, Aug 12, 2013 at 9:23 AM, Thibault, Daniel
daniel.thiba...@drdc-rddc.gc.ca wrote:
--
Actually, a non-root user can also access the root daemon if he has sudo
Since the 8692d4e lttng-tools commit (12 July), the default configuration
for user-space channels is per-UID (it would be nice if 'lttng enable-channel
--help' said so, by the way). However, in
lttng-tools\src\lib\lttng-ctl\lttng-ctl.c, one can still read:
void
--
Date: Fri, 16 Aug 2013 12:08:56 +
A question on the suggested feature below:
If we do the following set of commands:
lttng enable-event a* -u
lttng enable-event !ab -u
The intention is to enable all events
Fixes #588, which was closed prematurely
'lttng-sessiond -h' documents its short options; '-Z' no longer requires an
argument; '-s' removed; '-GHTU' handled; usage formatted to 80 columns; usage,
parse_args and long_options[] now in synch
(The attached diff was prepared semi-manually and is
Suppose we do something like this (edited for brevity):
$ lttng enable-event -u sample:message --loglevel notice
$ lttng list session
Events:
sample:message (loglevel: TRACE_NOTICE (5)) [enabled]
$ lttng disable-event -u sample:message
$ lttng list session
Events:
sample:message (loglevel:
De : Ikaheimonen, JP [mailto:jp_ikaheimo...@mentor.com]
Envoyé : 10 septembre 2013 07:41
As you can see, things easily get too complicated.
Currently in LTTng you can only disable events that are enabled in the first
place. In other words, an event is born disabled, as if 'lttng
-Message d'origine-
From: Julien Desfossez jdesfos...@efficios.com
The live trace reading mechanism offers guarantees that the viewer can never be
in a position where it has to assume that it can read the trace safely.
--
Could you clarify what you mean
-Message d'origine-
De : Mathieu Desnoyers [mailto:mathieu.desnoy...@efficios.com]
Envoyé : 10 septembre 2013 16:18
And why is tracer_patchlevel dropped alongside the process-related
attributes?
Because different processes could use different tracer versions.
How is such a
I notice that the metadata of per-PID user-space traces has three attributes
which are dropped from the equivalent per-UID trace:
Per-PID trace:
env {
hostname = sds-dut-vb;
domain = ust;
tracer_name = lttng-ust;
tracer_major = 2;
tracer_minor = 3;
Date: Tue, 10 Sep 2013 11:28:53 -0400
From: Julien Desfossez jdesfos...@efficios.com
-Message d'origine-
The live trace reading mechanism offers guarantees that the viewer can
never be in a position where it has to assume that it can read the trace
safely.
We keep the enable-event command as it is. However, we enhance the event name
specification syntax somewhat. We could of course do a full regex matching,
but I
fear that would get too slow. Instead we'd have something like
$ lttng enable-event -u *!mine*!alsomine
This would enable any
In https://www.kernel.org/doc/Documentation/trace/tracepoints.txt, there is
prominent mention of an include/trace/events/subsys.h. Where is that header
supposed to be? I can't find it in either of the kernel or lttng sources.
Daniel U. Thibault
Protection des systèmes et contremesures
In https://www.kernel.org/doc/Documentation/trace/tracepoints.txt, there
is prominent mention of an include/trace/events/subsys.h. Where is that
header supposed to be? I can't find it in either of the kernel or lttng
sources.
subsys is the name of your subsystem. There is no
1 - 100 of 262 matches
Mail list logo