Remove ltt-armtap and disarmtap from the Makefile

From: Alexandre Montplaisir <alexandre.montplaisir@gmail.com>

Signed-off-by: Alexandre Montplaisir <alexandre.montplaisir@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttngtrace update todo

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add getopt for options management

This patch also introduce the usage function.

Signed-off-by: David Goulet <david.goulet@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update lttngtrace todo

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add lttngtrace

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update version to 0.88

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add module-unloading capabilities to ltt-disarmall

Make ltt-disarmall unload all the kernel modules before exiting.

The module list is not in the same order as in ltt-armall, because
of some dependencies between the modules.

Signed-off-by: Alexandre Montplaisir <alexandre.montplaisir@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add module-loading capabilities to ltt-armall

This change makes ltt-armall load the required kernel modules
if it cannot find them.

The script will also check if it is being run as root, so
modprobe doesn't fail.

Signed-off-by: Alexandre Montplaisir <alexandre.montplaisir@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liblttctl: switch_timer for all channels fix

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update versino to 0.86

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liblttdvfs: put fadvise hint after sync writeback

On Wed, 19 May 2010, Mathieu Desnoyers wrote:
>
> A faced a small counter-intuitive fadvise behavior though.
>
>   posix_fadvise(fd, 0, 0, POSIX_FADV_DONTNEED);
>
> only seems to affect the parts of a file that already exist.

POSIX_FADV_DONTNEED does not have _any_ long-term behavior. So when you do
a

        posix_fadvise(fd, 0, 0, POSIX_FADV_DONTNEED);

it only affects the pages that are there right now, it has no effect on
any future actions.

> So after each splice() that appends to the file, I have to call fadvise
> again. I would have expected the "0" len parameter to tell the kernel to
> apply the hint to the whole file, even parts that will be added in the
> future.

It's not a hint about future at all. It's a "throw current pages away".

I would also suggest against doing that kind of thing in a streaming write
situation. The behavior for dirty page writeback is _not_ welldefined, and
if you do POSIX_FADV_DONTNEED, I would suggest you do it as part of that
writeback logic, ie you do it only on ranges that you have just waited on.

IOW, in my example, you'd couple the

        sync_file_range(fd, (index-1)*BUFSIZE, BUFSIZE,
+SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE|SYNC_FILE_RANGE_WAIT_AFTER);

with a

        posix_fadvise(fd, (index-1)*BUFSIZE, BUFSIZE, POSIX_FADV_DONTNEED);

afterwards to throw out the pages that you just waited for.

                Linus

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update ltt-control version to 0.85

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liblttd: don't fill the page cache

* Linus Torvalds (torvalds@linux-foundation.org) wrote:
>
>
> On Wed, 19 May 2010, Mathieu Desnoyers wrote:
> >
> > Good point. This discard flag might do the trick and let us keep things simple.
> > The major concern here is to keep the page cache disturbance relatively low.
> > Which of new page allocation or stealing back the page has the lowest overhead
> > would have to be determined with benchmarks.
>
> We could probably make it easier somehow to do the writeback and discard
> thing, but I have had _very_ good experiences with even a rather trivial
> file writer that basically used (iirc) 8MB windows, and the logic was very
> trivial:
>
>  - before writing a new 8M window, do "start writeback"
>    (SYNC_FILE_RANGE_WRITE) on the previous window, and do
>    a wait (SYNC_FILE_RANGE_WAIT_AFTER) on the window before that.
>
> in fact, in its simplest form, you can do it like this (this is from my
> "overwrite disk images" program that I use on old disks):
>
>       for (index = 0; index < max_index ;index++) {
>               if (write(fd, buffer, BUFSIZE) != BUFSIZE)
>                       break;
>               /* This won't block, but will start writeout asynchronously */
>               sync_file_range(fd, index*BUFSIZE, BUFSIZE, SYNC_FILE_RANGE_WRITE);
>               /* This does a blocking write-and-wait on any old ranges */
>               if (index)
>                       sync_file_range(fd, (index-1)*BUFSIZE, BUFSIZE,
+SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE|SYNC_FILE_RANGE_WAIT_AFTER);
>       }
> and even if you don't actually do a discard (maybe we should add a
> SYNC_FILE_RANGE_DISCARD bit, right now you'd need to do a separate
> fadvise(FADV_DONTNEED) to throw it out) the system behavior is pretty
> nice, because the heavy writer gets good IO performance _and_ leaves only
> easy-to-free pages around after itself.

Great! I just implemented it in LTTng and it works very well !

A faced a small counter-intuitive fadvise behavior though.

  posix_fadvise(fd, 0, 0, POSIX_FADV_DONTNEED);

only seems to affect the parts of a file that already exist. So after each
splice() that appends to the file, I have to call fadvise again. I would have
expected the "0" len parameter to tell the kernel to apply the hint to the whole
file, even parts that will be added in the future. I expect we have this
behavior because fadvise() was initially made with read behavior in mind rather
than write.

For the records, I do a fadvice+async range write after each splice(). Also,
after each subbuffer write, I do a blocking write-and-wait on all pages that are
in the subbuffer prior to the one that has just been written, instead of using
the fixed 8MB window.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix ltt-armall/disarmall /debugfs auto-find

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix lttctl flight recorder wait for daemon (temp fix)

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttctl: wait for lttd to complete in overwrite mode

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add missing liblttdvfs.[ch]

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update version to 0.81

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liblttd cleanup and error handling fixes

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liblttd fix

Add one line in liblttd to ensure that every field of the instance struct are
initialized correctly.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Create utility methods in the library to support local write to disk

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Adds the multiple instances feature

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Move the code from lttd to liblttd and adapt everything so it works

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

update version to 0.80

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

ltt-armall script: disable input subsystem by default

Add -i option to enable input subsystem.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Document LTT_DAEMON in lttctl --help

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update licensing: lgplv2.1 for libs, gplv2 for programs

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttctl scripts: ensure sh compatibility

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

ltt-armall: head -1 is deprecated, replace by head -n 1

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Remove dependency of ltt-arm/disarmall on bash idioms

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

update ltt-armall/disarmall

- Protect variables
- use /bin/sh instead of /bin/bash

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Unify ltt-armall* scripts

Adds switches to ltt-armall to control locking-related and network-related
marker activation (off by default).
Modifies ltt-disarmall to simply disconnect all active markers (and no more).

Signed-off-by: Benjamin Poirier <benjamin.poirier@polymtl.ca>
Cc: Pierre-Marc Fournier <pierre-marc.fournier@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Remove commented code

No need to keep old cruft there! It's in the source control.

Signed-off-by: Benjamin Poirier <benjamin.poirier@polymtl.ca>
Cc: Pierre-Marc Fournier <pierre-marc.fournier@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Remove old userspace support

Removes the support for arming and disarming old style userspace tracing. This
tracing was based on syscalls and was removed after lttng-0.88

Signed-off-by: Benjamin Poirier <benjamin.poirier@polymtl.ca>
Cc: Pierre-Marc Fournier <pierre-marc.fournier@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Add git ignore files

Based on the kernel's .gitignore

Signed-off-by: Benjamin Poirier <benjamin.poirier@polymtl.ca>
Cc: Pierre-Marc Fournier <pierre-marc.fournier@polymtl.ca>

Support variable sized subbuffers

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

lttd: 0.75, improve error handling of file open

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

liblttctl: fix missing return value

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Add lttctl support for switch_timer

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Add network sync arm/disarm

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

Fix liblttctl closedir

Here's my patch to liblttctl.c so it compiles and runs using uClibc. I'm using
Linux 2.6.30.

 Sincerely,
 Scott Esters, AF6RT
 scott@af6rt.com

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

update version

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

add SPLICE_F_MORE to lttd splice

    SPLICE_F_MORE      More  data  will  be  coming in a subsequent splice.
                       This is a helpful hint when the fd_out refers  to  a
                       socket  (see  also  the  description  of MSG_MORE in
                       send(2), and the description of TCP_CORK in tcp(7))

Will be a useful guideline for network sending daemons.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>

move ltt-control out of trunk/ltt-control

remove unneeded dir attic/

remove unneeded directory tags/

fix lttd append

git-svn-id: http://ltt.polymtl.ca/svn@3419 04897980-b3bd-0310-b5e0-8ef037075253

move unused dirs to attic

git-svn-id: http://ltt.polymtl.ca/svn@3418 04897980-b3bd-0310-b5e0-8ef037075253

fix leak, enhance error values

git-svn-id: http://ltt.polymtl.ca/svn@3404 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3341 04897980-b3bd-0310-b5e0-8ef037075253

update lttctl help

git-svn-id: http://ltt.polymtl.ca/svn@3340 04897980-b3bd-0310-b5e0-8ef037075253

update lttctl version

git-svn-id: http://ltt.polymtl.ca/svn@3332 04897980-b3bd-0310-b5e0-8ef037075253

add lttctl help detail

git-svn-id: http://ltt.polymtl.ca/svn@3331 04897980-b3bd-0310-b5e0-8ef037075253

lttctl ltt-armall updated to new debugfs

git-svn-id: http://ltt.polymtl.ca/svn@3316 04897980-b3bd-0310-b5e0-8ef037075253

move old packages to obsolete

git-svn-id: http://ltt.polymtl.ca/svn@3246 04897980-b3bd-0310-b5e0-8ef037075253

apply zhaolei updates

git-svn-id: http://ltt.polymtl.ca/svn@3212 04897980-b3bd-0310-b5e0-8ef037075253

lttctl: Rewrite -o option processing

git-svn-id: http://ltt.polymtl.ca/svn@3193 04897980-b3bd-0310-b5e0-8ef037075253

really fix newline

git-svn-id: http://ltt.polymtl.ca/svn@3191 04897980-b3bd-0310-b5e0-8ef037075253

update newline fix

git-svn-id: http://ltt.polymtl.ca/svn@3190 04897980-b3bd-0310-b5e0-8ef037075253

update compat, fix newline

git-svn-id: http://ltt.polymtl.ca/svn@3189 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3179 04897980-b3bd-0310-b5e0-8ef037075253

update lttctl

git-svn-id: http://ltt.polymtl.ca/svn@3174 04897980-b3bd-0310-b5e0-8ef037075253

update lttctl

git-svn-id: http://ltt.polymtl.ca/svn@3164 04897980-b3bd-0310-b5e0-8ef037075253

update lttd inotify

git-svn-id: http://ltt.polymtl.ca/svn@3163 04897980-b3bd-0310-b5e0-8ef037075253

When compile lttctl 0.58 in fc9, I got following error message:

Making all in lttd
make[2]: Entering directory `/home/zl/ltt-control-0.58-23112008/lttd'
gcc -DHAVE_CONFIG_H -I.. -I..     -g -O2 -MT lttd.o -MD -MP -MF
.deps/lttd.Tpo
+-c -o lttd.o lttd.c
In file included from /usr/include/asm/fcntl.h:2,
                 from /usr/include/linux/fcntl.h:5,
                 from /usr/include/linux/inotify.h:12,
                 from lttd.c:55:
/usr/include/asm-generic/fcntl.h:117: error: redefinition of 'struct
flock'
/usr/include/asm-generic/fcntl.h:140: error: redefinition of 'struct
flock64'
lttd.c: In function 'read_subbuffer':
lttd.c:467: warning: passing argument 2 of 'splice' from incompatible
pointer
+type
make[2]: *** [lttd.o] Error 1
make[2]: Leaving directory `/home/zl/ltt-control-0.58-23112008/lttd'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/zl/ltt-control-0.58-23112008'
make: *** [all] Error 2
$

git-svn-id: http://ltt.polymtl.ca/svn@3162 04897980-b3bd-0310-b5e0-8ef037075253

Fixed following compile warning:

gcc -DHAVE_CONFIG_H -I.. -I..     -g -O2 -MT lttd.o -MD -MP -MF
.deps/lttd.Tpo
+-c -o lttd.o lttd.c
lttd.c: In function 'read_subbuffer':
lttd.c:467: warning: passing argument 2 of 'splice' from incompatible pointer type

git-svn-id: http://ltt.polymtl.ca/svn@3161 04897980-b3bd-0310-b5e0-8ef037075253

update lttctl 0.58

git-svn-id: http://ltt.polymtl.ca/svn@3158 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3141 04897980-b3bd-0310-b5e0-8ef037075253

update lttd

git-svn-id: http://ltt.polymtl.ca/svn@3140 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3123 04897980-b3bd-0310-b5e0-8ef037075253

Moved daemon() after channels_init() to make lttd return !0 on error.
Applies on top of ltt-control-0.55-16102008.

Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
git-svn-id: http://ltt.polymtl.ca/svn@3122 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3116 04897980-b3bd-0310-b5e0-8ef037075253

Hi Mathieu,

this patch make lttctl select channel_root in debugfs automatically
 when no specify.
Applies on top of ltt-control-0.54-10102008.

Signed-off-by: Zhaolei <zhaolei@cn.fujitsu.com>
git-svn-id: http://ltt.polymtl.ca/svn@3115 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3097 04897980-b3bd-0310-b5e0-8ef037075253

update armall

git-svn-id: http://ltt.polymtl.ca/svn@3089 04897980-b3bd-0310-b5e0-8ef037075253

lttd support splice

git-svn-id: http://ltt.polymtl.ca/svn@3074 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@3071 04897980-b3bd-0310-b5e0-8ef037075253

Looks like we overlooked some marker redirects in our ltt-armall patch
from February (see Jan's patch in [1]). As I can see, you already caught
some of them, but some are still missing. See the attached patch.

Gernot Hillier
Siemens AG, CT SE 2, Corporate Competence Center Embedded Linux
Gernot Hillier <gernot.hillier@siemens.com>

git-svn-id: http://ltt.polymtl.ca/svn@3070 04897980-b3bd-0310-b5e0-8ef037075253

add tags

git-svn-id: http://ltt.polymtl.ca/svn@3000 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@2999 04897980-b3bd-0310-b5e0-8ef037075253

update tap scripts

git-svn-id: http://ltt.polymtl.ca/svn@2997 04897980-b3bd-0310-b5e0-8ef037075253

update tap scripts

git-svn-id: http://ltt.polymtl.ca/svn@2996 04897980-b3bd-0310-b5e0-8ef037075253

move obsolete ltt-modules to attic

git-svn-id: http://ltt.polymtl.ca/svn@2986 04897980-b3bd-0310-b5e0-8ef037075253

move all projects into the trunk directory

git-svn-id: http://ltt.polymtl.ca/svn@2975 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@2968 04897980-b3bd-0310-b5e0-8ef037075253

fix lttd mutexes

git-svn-id: http://ltt.polymtl.ca/svn@2967 04897980-b3bd-0310-b5e0-8ef037075253

update

git-svn-id: http://ltt.polymtl.ca/svn@2961 04897980-b3bd-0310-b5e0-8ef037075253

64 bits

git-svn-id: http://ltt.polymtl.ca/svn@2960 04897980-b3bd-0310-b5e0-8ef037075253

64 bits

git-svn-id: http://ltt.polymtl.ca/svn@2959 04897980-b3bd-0310-b5e0-8ef037075253

64 bits

git-svn-id: http://ltt.polymtl.ca/svn@2958 04897980-b3bd-0310-b5e0-8ef037075253

add marker user activation

git-svn-id: http://ltt.polymtl.ca/svn@2957 04897980-b3bd-0310-b5e0-8ef037075253

add marker user activation

git-svn-id: http://ltt.polymtl.ca/svn@2956 04897980-b3bd-0310-b5e0-8ef037075253

add armalluser

git-svn-id: http://ltt.polymtl.ca/svn@2955 04897980-b3bd-0310-b5e0-8ef037075253

update version

git-svn-id: http://ltt.polymtl.ca/svn@2823 04897980-b3bd-0310-b5e0-8ef037075253