lttng-tools.git
18 months agoactions: introduce lttng_action_init
Simon Marchi [Fri, 29 Nov 2019 21:46:05 +0000 (16:46 -0500)] 
actions: introduce lttng_action_init

This function is to be used to initialize the common portion of all
action structures.  I have updated the sole currently existing action
(notify), but the function will also be used in subsequent patches.

Change-Id: I4e42554c42a4e6a5ef2da292a6dfeb708d72a602
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoactions: introduce function typedef for creating actions from buffer
Simon Marchi [Thu, 28 Nov 2019 22:51:01 +0000 (17:51 -0500)] 
actions: introduce function typedef for creating actions from buffer

The only existing action type, LTTNG_ACTION_TYPE_NOTIFY, does not
require deserializing additional data (on top of the action type field),
so lttng_action_create_from_buffer handles it in a trivial way.
Upcoming patches will introduce new action types which will need to
deserialize some additional data.  This patch prepares
lttng_action_create_from_buffer for that by making it call an
action-specific function for deserializing this additional data.  The
changes are inspired by what lttng_condition_create_from_buffer does.

No functional changes intended.

Change-Id: I469a67b744aa2cf7a45d7d970f1bbee6a994a2a9
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agobuffer-view: introduce lttng_buffer_view_contains_string
Simon Marchi [Fri, 29 Nov 2019 21:42:24 +0000 (16:42 -0500)] 
buffer-view: introduce lttng_buffer_view_contains_string

This function can be used to safely validate that a string in a buffer
view has the length we expect.  The goal is to avoid doing any reads
outside the buffer view, whatever the input is, considering that the
buffer and advertised length of the string are untrusted data.

It will be used by subsequent patches to deserialize strings from
received buffers.

Change-Id: I8afc4b6f9334e035e0ef02cb96b157df59d8107d
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoMove actions source files to src/common/actions directory
Simon Marchi [Mon, 25 Nov 2019 19:20:31 +0000 (14:20 -0500)] 
Move actions source files to src/common/actions directory

Since more actions will be added, group them under an "actions"
directory.  The files to be added are expected to have some pretty
generic/overloaded names, such as start-session.c and snapshot.c, so
having them under the actions directory make it clear that they
implement the action described by the name.

Change-Id: Ia47160dd75531eb9bcf13f875f4bc6caa2391d7b
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoust registry: Refactor representation of nested types
Mathieu Desnoyers [Wed, 1 Apr 2020 21:48:09 +0000 (17:48 -0400)] 
ust registry: Refactor representation of nested types

Those allow LTTng-UST to internally represent nested types properly,
and to serialize them when sending them over to the session
daemon. However, for now, the session daemon only accepts arrays and
sequences which contain an integer type, which is the only use-case
emitted by lttng-ust. Wait until we have the nested types wired up
within lttng-ust before accepting and supporting nested types so it
can be tested.

Given the size of this change to ust-metadata.c, use this opportunity
to make sure that every

  if (ret)
      ...

in this file is changed to conform to lttng-tools coding style:

  if (ret) {
      ...
  }

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3ac44fc3a993f73d1fb08056331ad6fed7059981
Depends-on: lttng-ust: I45bb0886c5ac970c3ca75dbefcb94adb50801294

18 months agocommon: keep libcommon_la_SOURCES list sorted
Simon Marchi [Fri, 3 Apr 2020 17:52:41 +0000 (13:52 -0400)] 
common: keep libcommon_la_SOURCES list sorted

Change-Id: I8d263f4d876047074cab3169f929fd255536d702
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-crash(1): document the command's positional argument
Philippe Proulx [Fri, 3 Apr 2020 20:14:05 +0000 (16:14 -0400)] 
lttng-crash(1): document the command's positional argument

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I919329f2e072b14ad94181e98e49cb8840991225
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-sessiond(8): append missing argument to short options
Philippe Proulx [Fri, 3 Apr 2020 02:48:57 +0000 (22:48 -0400)] 
lttng-sessiond(8): append missing argument to short options

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ifab424acfa6dedbe43167e69e2c2fb70f3ecf9a7
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-sessiond(8): sort the option list by long option name
Philippe Proulx [Fri, 3 Apr 2020 02:46:40 +0000 (22:46 -0400)] 
lttng-sessiond(8): sort the option list by long option name

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Icaeabbb95fa650c059f208b1808f0b2cf3202117
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-relayd(8): mention the `--config` option
Philippe Proulx [Fri, 3 Apr 2020 02:45:57 +0000 (22:45 -0400)] 
lttng-relayd(8): mention the `--config` option

This patch does not document the configuration file's format. This work
is reserved for a future patch.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: If36dcaf244e194b58c1e6cacbe0e771176ac2045
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: lttng-load: support legacy PID tracker specification
Jérémie Galarneau [Tue, 7 Apr 2020 04:46:39 +0000 (00:46 -0400)] 
Fix: lttng-load: support legacy PID tracker specification

The 2.12 release changes the way tracked process attributes are
expressed in the MI and save/restore formats. While the MI schema was
bumped to 4.0, the save/restore schema only undergoes a minor bump to
accomodate existing users.

The original commit introducing this change justified the breaking
change as saved PIDs being fairly unlikely. However, even the
'INCLUDE_ALL' policy will specify a 'trackers' node, which no longer
existed and made all existing configurations incompatible.

A legacy load path is introduced to support the former PID tracker
serialization format and preserve the compatibility with existing
configurations.

Configurations generated by the 2.11 releases are included to test
this new legacy path.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ied6532c42cb2d1a5c9e39785cc4e47aaf89b8288

18 months agoFix: sessiond: invalid session configuration on EXCLUDE_ALL policy
Jérémie Galarneau [Tue, 7 Apr 2020 04:14:00 +0000 (00:14 -0400)] 
Fix: sessiond: invalid session configuration on EXCLUDE_ALL policy

Saving a session with a process attribute tracker that uses the
`EXCLUDE_ALL` policy results in an invalid session configuration.

Currently, a tracker of the following form is produced:
<process_attr_values>
  <vpid/>
</process_attr_values>

This is invalid as per the XSD as 'vpid' is not a list; it is an
individual tracked attribute.

The appropriate '<process_attr_values/>' empty node is now produced.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia31e276fe7246a89e72d6808c9ed960fb04f1b3a

18 months agoFix: relayd: unchecked allocation result of unlinked file pool
Jérémie Galarneau [Mon, 6 Apr 2020 16:39:17 +0000 (12:39 -0400)] 
Fix: relayd: unchecked allocation result of unlinked file pool

`pool` is not checked for NULL after its allocation. Error out
if the allocation fails.

In lttng_unlinked_file_pool_create: Return value of function which
returns null is dereferenced without checking (CWE-476)

Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2a7717701cf3d11de557b9ecdc6609c1f6a1fd6f

18 months agolttng-crash: use `spawn_viewer()` to launch trace viewer
Francis Deslauriers [Fri, 3 Apr 2020 18:57:55 +0000 (14:57 -0400)] 
lttng-crash: use `spawn_viewer()` to launch trace viewer

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I59f8d6c1189b3b3b4cfd0704ff2c8eb22e6df44f
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-view: clean-up: move `--viewer` code to specific file
Francis Deslauriers [Fri, 3 Apr 2020 18:21:25 +0000 (14:21 -0400)] 
lttng-view: clean-up: move `--viewer` code to specific file

This code will be reuse by the lttng-crash utility.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ide72ad08577d55bbf2f7833d46e734b8a680c9d2
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-crash: clean-up: fix alignment of format string
Francis Deslauriers [Thu, 2 Apr 2020 21:27:03 +0000 (17:27 -0400)] 
lttng-crash: clean-up: fix alignment of format string

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I893a8ab7cbc0ad35a4ebabcf5017dbc83be1efc3
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-view: clean-up: rename `parse_options()` -> `parse_viewer_option()`
Francis Deslauriers [Thu, 2 Apr 2020 20:16:59 +0000 (16:16 -0400)] 
lttng-view: clean-up: rename `parse_options()` -> `parse_viewer_option()`

This seems more representative of what this function does.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I78e4282a13048f4a35df4ddeae91b89969b73941
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: python: suppress -Wmissing-prototypes warning with SWIG 3.0.10
Simon Marchi [Fri, 27 Mar 2020 01:11:35 +0000 (21:11 -0400)] 
Fix: python: suppress -Wmissing-prototypes warning with SWIG 3.0.10

SWIG 3.0.10 is used on SLES 12.  It produces this warning:

      CC       lttng_wrap.lo
    lttng_wrap.c:3411:1: error: no previous prototype for ‘SWIG_strnlen’ [-Werror=missing-prototypes]
     3411 | SWIG_strnlen(const char* s, size_t maxlen)
          | ^~~~~~

SWIG_strnlen is defined like this:

    size_t
    SWIG_strnlen(const char* s, size_t maxlen)
    {
        ...
    }

Since the function is not static and has no previous declaration, the
diagnostic is emitted.  We can see that they have fixed it in SWIG
3.0.12, where the same function is defined as:

    SWIGINTERN size_t
    SWIG_strnlen(const char* s, size_t maxlen)
    {
        ...
    }

SWIGINTERN is defined as static.

We can workaround the warning by adding our own declaration for
SWIG_strnlen in lttng.i(.in).

Tested by build with SWIG 3.0.10, 3.0.11 and 3.0.12.

Change-Id: I7d9d93cf5fc04f2044a47ad384e16d726e27f72f
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-view: clean-up: use singular form for type name
Francis Deslauriers [Thu, 2 Apr 2020 20:04:49 +0000 (16:04 -0400)] 
lttng-view: clean-up: use singular form for type name

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ibeb6363def3579761e9a5a4cb76f7c6ca22d6146
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-view: clean-up: remove references to LTTv
Francis Deslauriers [Thu, 2 Apr 2020 20:00:14 +0000 (16:00 -0400)] 
lttng-view: clean-up: remove references to LTTv

Those comments are not relevant anymore and were missed by the following
clean-up commit:
  commit 6dd26587e926671cdec2545b7d3db74cbd6a7cd8
  Author: Jérémie Galarneau <jeremie.galarneau@efficios.com>
  Date:   Thu Jan 30 12:01:57 2020 -0500

      lttng-view: clean-up: remove commented and unused references to lttv

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ie68614c8160f62a2c045153b377f036a4399ad87
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: relayd: harmonize path format in backward-compat mode
Jérémie Galarneau [Fri, 3 Apr 2020 19:05:37 +0000 (15:05 -0400)] 
Fix: relayd: harmonize path format in backward-compat mode

Observed issue
==============

Currently, the relay daemon produces the following path formats
depending on the whether a tracepath is provided, the version of the
session daemon peer, and the grouping option specified on launch of
the relay daemon.

Hostname grouping, no custom trace path
pre 2.11: $BASE/$SHOSTNAME/$SESSION-$DATETIME
    2.12: $BASE/$SHOSTNAME/$SESSION-$DATETIME

Hostname grouping, custom trace path
pre 2.11: $BASE/$HOSTNAME/$TRACEPATH
    2.12: $BASE/$HOSTNAME/$TRACEPATH

Tracing session grouping, no custom trace path
pre 2.11: $BASE/$SESSION/$HOSTNAME-$DATETIME
    2.12: $BASE/$SESSION/$HOSTNAME-$DATETIME

Tracing session grouping, custom trace path
pre 2.11: $BASE/$SESSION/$HOSTNAME/$TRACEPATH
    2.12: $BASE/$SESSION/$HOSTNAME-$DATETIME/$TRACEPATH

As you can see, there is a single case where the format
diverges based on the version of the session daemon.

Cause
=====

Pre-2.11 session daemons do not transmit a session creation time when
a TRACEPATH is specified as part of the streaming url (e.g.
`lttng create my_session --set-url net://localhost/a_path`)

Hence, the backward compatibility path formatting code does not
insert a "DATETIME" string in the resulting path.

Solution
========

The relay daemon samples the time when it creates its session and that
time is formatted into the DATETIME representation if no DATETIME is
present in the path provided by pre-2.11 peers.

Drawbacks
=========

Sampling the relay session creation time will not yield the exact same
behaviour as what a 2.11+ peer would produce, but it is a reasonable
approximation for most use-cases.

Users depending on this time being the exact same as that sampled by
the session daemon will need to adapt tools anyhow if they use the new
--group-output-by-session option, so the change doesn't introduce more
problems.

This behaviour can be surprising when snapshots are streamed by
pre-2.11 peers as the session creation DATETIME will be different for
all snapshots. This is not ideal, but still less jarring than getting
a completely different path format depending on a peer's version.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3316aa35a34985e83ae759851af3a899b0011789
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoBump session.xsd version to 2.12
Jérémie Galarneau [Fri, 3 Apr 2020 14:43:18 +0000 (10:43 -0400)] 
Bump session.xsd version to 2.12

The session schema has changed to include new tracked process
attributes. This change is technically non-backward compatible if
tracked PIDs were saved.

As it is extremely unlikely that anyone saves PIDs to a session
configuration to load it in another lttng-tools version, the schema
does not allow the pre-2.12 PID tracking specification to be
expressed. In my opinion, this is unlikely enough not to warrant a
major version bump.

This can be changed if someone really encounters this in the wild or
has a legitimate use case for going through this trouble.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I8fd8dac02065b09c3e3c75b38699caea3223ce50
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-relayd(8): normalize style and add details
Philippe Proulx [Thu, 2 Apr 2020 19:32:59 +0000 (15:32 -0400)] 
lttng-relayd(8): normalize style and add details

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ifd2e90686bf9955f0c68fe158c60344d346814d0
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agodoc/man: refer to Babeltrace 2 instead of Babeltrace 1
Philippe Proulx [Thu, 2 Apr 2020 19:03:15 +0000 (15:03 -0400)] 
doc/man: refer to Babeltrace 2 instead of Babeltrace 1

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Icafd69f28304b14e49e8cf8e21d2c6ebe3e319b6
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agolttng-clear(1): normalize style and add details
Philippe Proulx [Thu, 2 Apr 2020 18:44:36 +0000 (14:44 -0400)] 
lttng-clear(1): normalize style and add details

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I50709df27a04bbb5f300c734b9a85ae3a7c311c2
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: filter-grammar-test: add dependencies between steps
Simon Marchi [Tue, 31 Mar 2020 03:24:24 +0000 (23:24 -0400)] 
Fix: filter-grammar-test: add dependencies between steps

If the user specifies only the -B switch of the filter-grammar-test
program, the program will try to print the bytecode without having
generated it first, leaving to a segfault.  Similarly, if the user
specifies only the -b switch, the program will try to generate the
bytecode from the IR, without having generated the IR first, also
leading to a segfault.

This patch adds some kind of dependency between the steps, such that if
the user is interested in a particular step (let's say, print the
bytecode), all the required steps will also be done.

Change-Id: Idc365a12b992a950566f227759fd3223cf5c5fd8
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: relayd: assertion fails on creation of session by peer < 2.11
Jérémie Galarneau [Thu, 2 Apr 2020 19:47:30 +0000 (15:47 -0400)] 
Fix: relayd: assertion fails on creation of session by peer < 2.11

Observed issue
==============

An assertion that a chunk has no active directory handle fails when
creating an anonymous chunk. More specifically, this occurs when
associating an fd tracker to the newly created anonymous trace chunk.

This occurs when a session is created by a peer that is older than
2.11.

Cause
=====

Trace chunks that should monitor their file descriptors with a file
descriptor tracker must be associated with the tracker before any
other operation occurs on the chunk. This is to ensure that "raw" file
descriptors are not created when they were meant to be tracked.

Here, the credentials and session output directory are set before the
file descriptor tracker was provided to the anonymous chunk which is a
breach of the API contract (enforced by the assert()).

Solution
========

Associate the fd tracker immediately to the anonymous chunk before
providing it with a reference to the file descriptor
tracker. Moreover, a leak of the output_directory is prevented by not
setting it to NULL. The trace chunk will acquire a reference to the
trace chunk; it is not transferred to the trace chunk.

Note
====

The problem was introduced during the 2.12 release cycle (clear
feature); this doesn't need to be backported.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I92ca2f156f500dfe02e09f8b4783447e46710246

18 months agoFix: relayd: crash on creation of session by peer < 2.11
Jérémie Galarneau [Thu, 2 Apr 2020 18:08:12 +0000 (14:08 -0400)] 
Fix: relayd: crash on creation of session by peer < 2.11

Observed issue
==============

A NULL pointer dereference occurs during the creation of
a session that is associated with a peer older than 2.11.

The resulting backtrace follows:

 Program terminated with signal SIGSEGV, Segmentation fault.

 #0  0x0000564af45b755b in lttng_trace_chunk_set_as_owner (chunk=0x7f8ca8004730, session_output_directory=0x7f8ca8004680) at trace-chunk.c:1033
 1033 if (chunk->path[0] != '\0') {
 [Current thread is 1 (Thread 0x7f8cb808d700 (LWP 7300))]

 #0  0x0000564af45b755b in lttng_trace_chunk_set_as_owner (chunk=0x7f8ca8004730, session_output_directory=0x7f8ca8004680) at trace-chunk.c:1033
 #1  0x0000564af45a6a78 in session_set_anonymous_chunk (session=0x7f8ca8001380) at session.c:229
 #2  session_create (session_name=<optimized out>, hostname=<optimized out>, base_path=<optimized out>, live_timer=<optimized out>, snapshot=<optimized out>,
     sessiond_uuid=<optimized out>, id_sessiond=<optimized out>, current_chunk_id=<optimized out>, creation_time=<optimized out>, major=<optimized out>,
     minor=<optimized out>, session_name_contains_creation_time=<optimized out>) at session.c:416
 #3  0x0000564af459207e in relay_create_session (conn=0x7f8ca0000f60, payload=<optimized out>, recv_hdr=<optimized out>) at main.c:1428
 #4  0x0000564af4594f12 in relay_process_control_command (payload=0x7f8cb808c940, header=0x7f8ca0001000, conn=0x7f8ca0000f60) at main.c:3218
 #5  relay_process_control_receive_payload (conn=0x7f8ca0000f60) at main.c:3361
 #6  0x0000564af45980b0 in relay_process_control (conn=0x7f8ca0000f60) at main.c:3478
 #7  relay_thread_worker (data=<optimized out>) at main.c:3927
 #8  0x00007f8cbba9a46f in start_thread () from /usr/lib/libpthread.so.0
 #9  0x00007f8cbb9ca3d3 in clone () from /usr/lib/libc.so.6

Cause
=====

lttng_trace_chunk_set_as_owner() correctly handles the case
where a trace chunk has no output path, but expects the path
to be an empty string rather than being NULL.

This is not correct as an anonymous chunk, created in backward
compatibility mode when interacting with older peers, has no
path; the path is transmitted as part of the streams' attributes
upon their creation.

Solution
========

Simply check for a NULL pointer in the same place where the empty
chunk path string is created. The rest of the code in trace-chunk.c
doesn't assume that the chunk's path is non-NULL.

Note
====

The problem was introduced during the 2.12 release cycle (clear
feature); this doesn't need to be backported.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iaeb41e1648d61fbbe78d70b21191fd6d720900df

18 months agoFix: consumer: fallback to flush when flush empty is unsupported
Jérémie Galarneau [Thu, 2 Apr 2020 04:57:38 +0000 (00:57 -0400)] 
Fix: consumer: fallback to flush when flush empty is unsupported

Session destruction fails on older (<= 2.8) lttng-modules as the
flush_empty fails on the kernel streams during the quiet rotation.

Fallback to the regular flush as the semantics of regular rotations
are not expected here; we merely want to flush any pending data and
destroy the session.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ifdf8a4e60b55dbf582747d71f5c2485d24c11964

18 months agoFix: consumerd: incorrect clear logging statement
Jérémie Galarneau [Thu, 2 Apr 2020 04:45:21 +0000 (00:45 -0400)] 
Fix: consumerd: incorrect clear logging statement

A logging statement was apparently copy-pasted from the rotation code
and not adapted for the consumer_clear_buffer command and would
indicate that a flush operation failed when a clear operation was
performed.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I909a391fbd8b8bf48c7481d394571ad19d7f6332

18 months agoFix: sessiond: error reported on session destruction for old modules
Jérémie Galarneau [Thu, 2 Apr 2020 04:25:11 +0000 (00:25 -0400)] 
Fix: sessiond: error reported on session destruction for old modules

The session destruction command will return
-LTTNG_ERR_ROTATION_NOT_AVAILABLE_KERNEL when the kernel tracer
version does not support packet sequence numbers which prevents
rotations from being performed.

It is okay to not perform an implicit rotation in this case since we
know that no rotations have occurred during the session's lifetime (as
it is not supported). Thus, the client/library only needs to stop the
session, wait for pending data, and destroy the session.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ibccf73f08eecb6431ea3cc358bf8dd6af3ba4427

18 months agoFix: sessiond: erroneous error code returned on rotation failure
Jérémie Galarneau [Thu, 2 Apr 2020 02:16:32 +0000 (22:16 -0400)] 
Fix: sessiond: erroneous error code returned on rotation failure

`LTTNG_ERR_KERN_CONSUMER_FAIL` is returned by the kernel domain
rotation handling code. This code is associated with a failure to
launch the kernel consumer daemon which is not the case here.

The `LTTNG_ERR_ROTATION_FAIL_CONSUMER` is used instead and returned to
the client.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ifdc9e46addcc4dc7ca2002d6dff5b2d7ac6c31f3

18 months agoFix: lttng-destroy: missing newline on session destruction message
Jérémie Galarneau [Wed, 1 Apr 2020 22:34:59 +0000 (18:34 -0400)] 
Fix: lttng-destroy: missing newline on session destruction message

The lost packets/discarded events statistics are printed on the same
line as the session destruction progress message when the session is
stopped as part of the `destroy` command.

This is a consequence of printing the statistics as they are
retrieved; the statistics must be fetched before the destruction,
but the progress indicator is still being printed.

The statistics output is now formatted to a buffer and printed
after the session's destruction has completed.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2105056b109774a57c83be3e30984038880c0fb7

18 months agorelayd: clean-up: reference is repeated in comment
Jérémie Galarneau [Wed, 1 Apr 2020 20:16:05 +0000 (16:16 -0400)] 
relayd: clean-up: reference is repeated in comment

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2d963d5a3c0a2980eb745780b342b1fa9b14184d

18 months agoTypo: 'Descritptor' -> 'Descriptor'
Michael Jeanson [Wed, 1 Apr 2020 19:13:54 +0000 (15:13 -0400)] 
Typo: 'Descritptor' -> 'Descriptor'

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I3b616777c9d39e23b84224cb1a6a92fa43fceb45
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoTypo: 'Accomodate' -> 'Accommodate'
Michael Jeanson [Wed, 1 Apr 2020 19:09:22 +0000 (15:09 -0400)] 
Typo: 'Accomodate' -> 'Accommodate'

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I1f5408d82546db7df14d4899d4d2c52ec1421b52
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoClean-up: trace-ust comment still refers to only PID trackers
Jérémie Galarneau [Wed, 1 Apr 2020 14:51:56 +0000 (10:51 -0400)] 
Clean-up: trace-ust comment still refers to only PID trackers

An application must meet all process attribute trackers restrictions
to be traced.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I9d8cd00b95c7db70ebf4fe91d6aea60646e3c525

18 months agoFix: tracker: NULL pointer dereference after NULL check
Jérémie Galarneau [Tue, 31 Mar 2020 02:10:36 +0000 (22:10 -0400)] 
Fix: tracker: NULL pointer dereference after NULL check

value_view can be NULL and must thus be checked before use.

Moreover, the fix introduced in 1ad5cb59 is erreneous: the
function must validate that either:
  - value is a 'name' type, value_view is not null, and not len == 0,
  - value is an integer and value_view does not contain more data.

In process_attr_value_from_comm: Pointer is checked against null but
then dereferenced anyway (CWE-476)

Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia130ef57e10118960f1023338b90f7a10d588ee2

18 months agoFix: sessiond: NULL pointer dereference after NULL check
Jérémie Galarneau [Fri, 27 Mar 2020 15:27:13 +0000 (11:27 -0400)] 
Fix: sessiond: NULL pointer dereference after NULL check

The process attribute value deserialization allows the buffer view to
be NULL when the value's type is not USER_NAME nor GROUP_NAME. This is
not checked when ensuring that no string is passed (len == 0) in the
case of integral values.

A NULL check is added to the condition.

Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I343f747c325f739196284dadd3c407cfb4084268

18 months agoFix: sessiond: missing goto in error handler
Jérémie Galarneau [Fri, 27 Mar 2020 15:07:10 +0000 (11:07 -0400)] 
Fix: sessiond: missing goto in error handler

The trace_ust inclusion set add/remove methods do not jump to the
end label after checking the `tracker` variable. This can result
in a NULL pointer dereference when an invalid process attribute
is specified.

The same problem appears in save_process_attr_trackers() and
process_attr_value_from_comm().

The missing jump (goto) is added in all cases.

Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I473e008e5330a4c3820c8ab7c57ce4f2961e79b2

18 months agoFix: sessiond: user/group name can be leaked on malformed command
Jérémie Galarneau [Fri, 27 Mar 2020 15:01:05 +0000 (11:01 -0400)] 
Fix: sessiond: user/group name can be leaked on malformed command

process_attr_value_from_comm() can leak a copy of the user/group
name when the value type is erroneous. This is not reachable in
"normal" execution, but could be triggered by invalid "crafter"
lttng-ctl commands.

In process_attr_value_from_comm: Leak of memory or pointers to
system resources (CWE-404).

Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7ef55c0743c954a93e3d27ce17e6478708b49437

18 months agoconfigure: add -Wmissing-declarations, -Wmissing-prototypes, and more
Simon Marchi [Wed, 25 Mar 2020 22:40:02 +0000 (18:40 -0400)] 
configure: add -Wmissing-declarations, -Wmissing-prototypes, and more

Here's the rationale for each:

- -Wmissing-declarations: Make sure the definition of a function can
  "see" a corresponding (usually in a header file), if it isn't static.
  This makes sure that the declaration and definition don't go out of
  sync, which can lead to hard to debug problems (because it still
  builds, but the function doesn't receives what it thinks it receives).
  On top of pointing out out-of-sync declarations, it can help point out
  that a foo.c file misses including its header foo.c, or that a
  function should actually be made static.

- -Wmissing-prototypes: makes sure that functions without parameters are
  declared as `foo(void)` instead of `foo()`.  In C, the former declares
  a function that takes no parameters, whereas the latter declares a
  function without specifying its parameters.  The latter could be
  called with any number of parameters, which is a recipe for confusion.

- -Wmissing-parameter-type, -Wold-style-definition,
  -Wold-style-declarations, -Wstrict-prototypes: makes sure there's no
  function declared with parameters without types specified, or using
  the old style:

  int foo(bar)
  int bar;
  {
    ...
  }

  Unlikely, but there's no harm in enabling them.

Change-Id: I7ddf5ff61b4466c0bd7b03485ef29156c399e2a8
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: sessiond: make the --without-lttng-ust version of launch_application_notificatio...
Simon Marchi [Wed, 25 Mar 2020 22:39:56 +0000 (18:39 -0400)] 
Fix: sessiond: make the --without-lttng-ust version of launch_application_notification_thread static

When building with --without-lttng-ust, a simple version of
launch_application_notification_thread, implemented in the header file,
is used.  We get this warning:

      CC       main.o
    In file included from /home/simark/src/lttng-tools/src/bin/lttng-sessiond/main.c:61:
    /home/simark/src/lttng-tools/src/bin/lttng-sessiond/notify-apps.h:17:6: error: no previous prototype for ‘launch_application_notification_thread’ [-Werror=missing-prototypes]
       17 | bool launch_application_notification_thread(int apps_cmd_notify_pipe_read_fd)
          |      ^~~~~~~

Make the function `static inline` to avoid that.  The `inline` is not
strictly required here, but if that header ended up included by some
other source file that didn't use
launch_application_notification_thread, we would get a -Wunused-function
warning.  The `inline` avoids that.

Change-Id: I19605e0594af0d7997951def2da3a6313bf65e11
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: tests: include callsites.h from callsites.c
Simon Marchi [Wed, 25 Mar 2020 22:39:52 +0000 (18:39 -0400)] 
Fix: tests: include callsites.h from callsites.c

In commit:

    commit f12eb9c1ceb619db54be0842323a32cda12651cd
    Author: Simon Marchi <simon.marchi@efficios.com>
    Date:   Mon Nov 25 16:41:29 2019 -0500

        Fix all -Wmissing-declarations warning instances

I've fixed the -Wmissing-declarations warning in callsites.c by adding a local
declaration.  That was wrong, since there is actually a callsites.h header file
that needs to be included, which contains the declaration.  This is nicely
pointed out when building with clang and -Wstrict-prototypes:

      CC       exec_with_callsites-multi-lib-test.o
    In file included from /home/simark/src/lttng-tools/tests/regression/ust/multi-lib/multi-lib-test.c:15:
    /home/simark/src/lttng-tools/tests/regression/ust/multi-lib/callsites.h:10:21: error: this function declaration is not a prototype [-Werror,-Wstrict-prototypes]
    void call_tracepoint();
                        ^
                         void

Remove the local declaration and include callsites.h in callsites.c.

Change-Id: Ib656d96c2ed3b389697a2022e343e98ac0b66447
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: relayd: cast idigit argument to unsigned char
Simon Marchi [Wed, 25 Mar 2020 22:39:48 +0000 (18:39 -0400)] 
Fix: relayd: cast idigit argument to unsigned char

This diagnostic is emitted when building on cygwin:

    main.c:233:34: warning: array subscript has type ‘char’ [-Wchar-subscripts]
      233 |    if (errno != 0 || !isdigit(arg[0])) {
          |                               ~~~^~~

This is due to passing a `char` argument to isdigit.  According to the
man page of isdigit, the arguments of type `char` must be cast to
`unsigned char`, so they are able to represent the EOF value.  This
patch does that, and should get rid of the warning.

Change-Id: Iaed4c0b494a79b917761e65f18388f43478b97e1
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: tests: make some functions static
Simon Marchi [Wed, 25 Mar 2020 22:39:43 +0000 (18:39 -0400)] 
Fix: tests: make some functions static

Make two functions static, they are only used in their respective file.

Change-Id: I022dd064249683c9414ab36602e9645865373c51
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: tests: add `void` parameter to functions that take no parameters
Simon Marchi [Wed, 25 Mar 2020 22:39:39 +0000 (18:39 -0400)] 
Fix: tests: add `void` parameter to functions that take no parameters

In C, empty parenthesis declare a function without a prorotype (without
specifying its parameters).  This is not the same as having a `void`
parameter, which declares a function which has no parameters.

It's safer to use the later, otherwise it makes it possible to
erroneously call the function with some arguments.

Change this `test_function` to add `void`.  It fixes diagnostics like:

  CC       userspace-probe-elf-binary.o
/home/simark/src/lttng-tools/tests/utils/testapp/userspace-probe-elf-binary/userspace-probe-elf-binary.c:12:33: error: no previous prototype for ‘test_function’ [-Werror=missing-prototypes]
   12 | void __attribute__ ((noinline)) test_function()
      |

Change-Id: Iceb7636e44d45f51889667ec76f2c04c032b5df8
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: common: make lttng_trace_chunk_remove_subdirectory_recursive static
Simon Marchi [Wed, 25 Mar 2020 22:39:35 +0000 (18:39 -0400)] 
Fix: common: make lttng_trace_chunk_remove_subdirectory_recursive static

This function is only used in this file, make it static.  Fixes this
diagnostic:

  CC       trace-chunk.lo
/home/simark/src/lttng-tools/src/common/trace-chunk.c:1498:5: error: no previous prototype for ‘lttng_trace_chunk_remove_subdirectory_recursive’ [-Werror=missing-prototypes]
 1498 | int lttng_trace_chunk_remove_subdirectory_recursive(struct lttng_trace_chunk *chunk,
      |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Change-Id: I19e03b64557ce845b94f02fcf0d6fbafba654d95
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoFix: common: add `void` parameter to log_add_time declaration
Simon Marchi [Wed, 25 Mar 2020 22:39:30 +0000 (18:39 -0400)] 
Fix: common: add `void` parameter to log_add_time declaration

It makes it match the definition and fixes this warning:

      CC       error.lo
    /home/smarchi/src/lttng-tools/src/common/error.c:33:13: error: no previous prototype for ‘log_add_time’ [-Werror=missing-prototypes]
     const char *log_add_time(void)
                 ^~~~~~~~~~~~

Change-Id: Ibbf5eebd8171504bc7050c754e2a5d113b6b5387
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
18 months agoCONTRIBUTING.md: harmonize list style with the rest of the docs
Jérémie Galarneau [Tue, 10 Mar 2020 20:01:56 +0000 (16:01 -0400)] 
CONTRIBUTING.md: harmonize list style with the rest of the docs

Append full stops (periods, for our non-British friends) and
capitalize list elements as they are complete sentences on their own.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7646cafa8a23a3dc6cfabea2dd5a54ce244eb2b3

18 months agoCONTRIBUTING.md: clarify the guidelines for commit messages
Jérémie Galarneau [Tue, 10 Mar 2020 17:22:02 +0000 (13:22 -0400)] 
CONTRIBUTING.md: clarify the guidelines for commit messages

In order to streamline the code review process, I am adding a more
detailed explanation of the desired commit message format.

Of note, the commit title format `Fix: sub-system`, used informally
for a couple of months, is adopted for bug fixes. A template of the
sections expected in the commit message body of those patches is also
included.

More general guidelines are also added for feature contributions.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id0097949ca984e88367e8e78cb6111adb30c5b99

18 months agoFix: lttng-list: don't warn when the kernel domain has no channels
Jérémie Galarneau [Mon, 23 Mar 2020 23:12:26 +0000 (19:12 -0400)] 
Fix: lttng-list: don't warn when the kernel domain has no channels

Some commands beside lttng-enable-channel have the side-effect of
creating a domain. For instance, the lttng-track and lttng-untrack
commands will implicitly create their target domains if they don't
exist. Thus, it is not unexpected for a domain to exist without
channels.

Currently, tracking process attributes in the user space and kernel
domains will result in a warning being printed when lttng-status
(or lttng-list `the_session`) is invoked.

Example output:
Tracing session arielle_bolduc: [inactive]
    Trace output: /home/jgalar/lttng-traces/arielle_bolduc-20200323-191128

=== Domain: Linux kernel ===

Tracked process attributes
  Process IDs: all
  Virtual Process IDs: 12365, 526, 41
  User IDs: all
  Virtual User IDs: all
  Group IDs: all
  Virtual Group IDs: all

Warning: No kernel channel
=== Domain: User space ===

Buffering scheme: per-user

Tracked process attributes
  Virtual Process IDs: 12365, 526, 41
  Virtual User IDs: all
  Virtual Group IDs: all

The warning is removed since it can only confuse users.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I733ee6c3d8c1d297194842c7bdbdf3f237661e51

18 months agoRefactor: lttng-ctl: follow terminology of the tracker documentation
Jérémie Galarneau [Tue, 17 Mar 2020 20:14:30 +0000 (16:14 -0400)] 
Refactor: lttng-ctl: follow terminology of the tracker documentation

This commit harmonizes the process attribute tracker API and
serialization formats (save/restore and MI) with the documentation
with regards to the terminology used.

The message of the parent commit adjusting the manual pages of the
lttng-track and lttng-untrack commands details those terminology
changes and their rationale.

Some problems with the API introduced during the 2.12 development
cycle are also adressed:

Type safety:
  - The process attribute tracker is made type safe with regards
    to the platform's native types to express process
    attributes. Where the original API casted all integral values to
    integers, this change introduces accessors for all process
    attribute types (pid_t, uid_t, gid_t). This makes it easier to
    use the API safely and without producing warnings in user's
    code.

    Another benefit of adopting this explicit type-safe approach is
    that is will make it easier to add new attributes which are not
    expressible (or non-ambiguously expressible) using `int` and
    `string` types (e.g. tracking a virtual PID within a given
    namespace).

Ambiguity of INCLUDE_ALL, EXCLUDE_ALL, and INCLUDE_SET states:
  - The original tracker API has a notion of 'enabled' pid_tracker
    which is confusing to users:
      - enable = 0: everything is tracked,
      - enable = 1: a list of tracked pids is provided, which may be
        empty.
      - pid '-1' is *special* and tracks or untracks everything.

    This was replaced with a 'special' opaque value meaning 'ALL'
    which, while being clearer, was still confusing and hard to
    document.

    The revised API explicitly expresses the notion of a tracking
    policy (`enum lttng_tracking_policy`). When that policy is set
    to `LTTNG_TRACKING_POLICY_INCLUDE_SET`, the inclusion set can
    be queried and/or mutated.

    On top of being clearer, this aligns more closely with the
    internal lttng-sessiond daemon API which gets rid of a lot
    of code to handle those special cases. The resulting code is
    more verbose, but a lot easier to understand.

    Moreover, the types introduced (e.g. lttng_process_attr_values)
    are meant to be re-used if a new
    `LTTNG_TRACKING_POLICY_EXCLUDE_SET` tracking policy is added in
    the future.

Documentation:
  - The revised API includes a complete documentation. It documents
    the API usage, but also adds implementation notes such explicitly
    mentionning when/where user names and group names are resolved.

Client:
  - While making the changes to use this new API, some error messages
    are clarified (or added). The resulting output when listing the
    trackers was also changed to be more compact.

    The CLI output now also makes use of the terminology used in
    the documentation for all commands interacting with process
    attribute trackers.

    It is now also possible to specify multiple process attribute
    trackers along with the --all option for the lttng-track and
    lttng-untrack command.  For instance: `lttng tracker --userspace
    --vpid --vuid --all` is now allowed.

    The same process attribute tracker can also be specified more than
    once in a command, as follows:
    `lttng track --userspace --vpid 43,11 --vpid 55,77`

Since the serialization had been changed during the 2.12 cycle, I
changed them further to use the API's terminology in the element
names.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I5f0ea2849eb024ea6944b2ee7dbadfd893cc0be4

18 months agolttng: list: replace domain headers with the official names
Jérémie Galarneau [Fri, 20 Mar 2020 21:27:00 +0000 (17:27 -0400)] 
lttng: list: replace domain headers with the official names

The lttng-list command uses the following domain headers
when printing domain listings:
  - `Kernel`
  - `UST global`
  - `JUL (Java Util Logging)`
  - `LOG4j (Logging for Java)`
  - `Python (logging)`

Those don't match the official project names and/or names
used in the LTTng online documentation.

They are replaced by:
  - `Linux kernel`
  - `User space`
  - `java.util.logging (JUL)`
  - `log4j`
  - `Python logging`

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I549b309eebf0d00dfaacbed00e263f4e5e3d718b

18 months agolttng: list: print `per-user` and `per-process` buffer types
Jérémie Galarneau [Fri, 20 Mar 2020 21:17:34 +0000 (17:17 -0400)] 
lttng: list: print `per-user` and `per-process` buffer types

The lttng-list command prints `per UID` and `per PID` for the
buffer type which is inconsistent with the terms used in the
documentation.

Moreover, `Buffer type` is replaced by `Buffering scheme` for
the same reason.

Change the print-out to match the online documentation.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id20098386ec0369fae017d01786ecc13c78820c0

18 months agoDocs: overhaul of lttng-track(1) and lttng-untrack(1)
Jérémie Galarneau [Wed, 4 Mar 2020 21:37:32 +0000 (16:37 -0500)] 
Docs: overhaul of lttng-track(1) and lttng-untrack(1)

- Replace the term whitelist by "inclusion set"

  The use of whitelist (and its counterpart, blacklist) is somewhat
  controversial and doesn't align with the terms used in the tracker
  API.

  The "inclusion set" nomenclature is inspired by that used by ACL,
  namely "include list" and "exclude list". Set is preferred to list
  as there is no implied notion of order.

- Adapt process attribute descriptions

  The process attribute descriptions were apparently copy-pasted from
  the original PID description (all are referencing "process IDs").

  A suitable description is added for each process attribute.

- Add new process attributes to lttng-untrack(1)

  The lttng-untrack man page was not updated when the resource types
  introduced in LTTng 2.12 were added to lttng-track(1).

- Group all process attributes with their 'virtual' counter-parts

  The order in which process attributes were described was: pid, uid,
  gid, vpid, vuid, vgid.

  Grouping process attributes with their virtual counterpart (e.g. pid
  followed by vpid) makes it easier to understand the difference at a
  glance.

  The order becomes: pid, vpid, uid, vuid, gid, vgid

- Remove all PID-specific wordings in lttng-untrack(1)

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I46486c4e90003da3e860682f6c22a0e4a154492b

19 months agoFix: MI: bump MI schema version to 4.0 in mi-lttng.c
Jérémie Galarneau [Thu, 19 Mar 2020 01:02:03 +0000 (21:02 -0400)] 
Fix: MI: bump MI schema version to 4.0 in mi-lttng.c

The MI schema has been bumped to 4.0, but the internal constant
in mi-lttng.c still indicates a 3.0 schema. Synchronize both
versions to 4.0 for the 2.12 release.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia27e15e2c154b2fd39eeaaa6c71df922a233bc80

19 months agoFix: sessiond: occasional badfd error on repeated SIGTERM
Jérémie Galarneau [Thu, 5 Mar 2020 21:52:32 +0000 (16:52 -0500)] 
Fix: sessiond: occasional badfd error on repeated SIGTERM

The session daemon occasionally prints the following messages
when it received multiple SIGTERM signals:

PERROR - 16:50:18.505585257 [49845/49845]: write poll pipe: Bad file
descriptor (in notify_thread_pipe() at utils.c:35)

This is caused by a (somewhat inevitable) race between the teardown of
the daemon and the closing of its quit pipe. This happens more often
when kernel modules take a long time to be unloaded and the user
spams ctrl+c in the hope of convincing the daemon process to close
faster since modules are unloaded after closing the quit pipe.

Setting closed pipe fds to '-1' is safe anyway and is already
handled by the notify_thread_pipe() util.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: If08a455c65055dcb56ffbad3703173e6a45cfba5

19 months agoFix: lttng: incorrect domain list printed when no domain is provided
Jérémie Galarneau [Wed, 4 Mar 2020 23:27:28 +0000 (18:27 -0500)] 
Fix: lttng: incorrect domain list printed when no domain is provided

The following commands make use of a common utility function to
validate the count of domains specified and print an error when it
is invalid:
  - lttng-enable-event,
  - lttng-disable-event,
  - lttng-track,
  - lttng-untrack,
  - lttng-add-context,
  - lttng-enable-channel,
  - lttng-disable-channel.

Those commands do not allow the same domains to be used. In fact, they
all expect --kernel or --userspace only, except for the
lttng-enable-event, lttng-disable-event, and lttng-add-context
commands which allow the --log4j, --jul, and --python domain options
to be used.

Currently, the error message when no domain is specified is incorrect
for all of those commands. The error reads as follows:

`Error: Please specify a domain (-k/-u/-j).`

For most commands, the -j option cannot be used. For those that allow
agent domains, the message is missing the -l and -p domains.

This ensures that the expected domains are printed for each of those
commands.

Moreover, the message is clarified by using the long form option
names.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I45aee075dbf6c62c4120bdeb06697b88b2d8716c

19 months agoFix all -Wdiscarded-qualifiers warning instances
Simon Marchi [Thu, 13 Feb 2020 15:36:31 +0000 (10:36 -0500)] 
Fix all -Wdiscarded-qualifiers warning instances

Change-Id: I85246856d284d5a2dad9609f5fb06d32a89dbb7c
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoRemove part of last name to fit in a 80 character line
Jonathan Rajotte [Tue, 17 Mar 2020 23:04:32 +0000 (19:04 -0400)] 
Remove part of last name to fit in a 80 character line

It plays much more nicely with the .clang-format since it ensure that
the line stay inside the 80 column limit.

It also match my signed-off and other files already present in the
project.

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Change-Id: Ib1ce1f101e9f24c03a05888b5fb1901341d8c005
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix: Tests: trace path wildcards not expanded
Francis Deslauriers [Tue, 17 Mar 2020 21:48:40 +0000 (17:48 -0400)] 
Fix: Tests: trace path wildcards not expanded

Issue
=====
Some testcases related to `--live` sessions in
`tools/clear/{test_ust,test_kernel}` fail with the following error:
  ls: cannot access '/tmp/tmp.gi1eVM3fZ9/tmp.dNL0WHRTQ3/ci-node-bionic-amd64-02-04/LUP49x7iV7xexehS*': No such file or directory
  not ok 61 - Failed to list content of directory "/tmp/tmp.gi1eVM3fZ9/tmp.dNL0WHRTQ3/ci-node-bionic-amd64-02-04/LUP49x7iV7xexehS*"
  FAIL: tools/clear/test_ust 61 - Failed to list content of directory "/tmp/tmp.gi1eVM3fZ9/tmp.dNL0WHRTQ3/ci-node-bionic-amd64-02-04/LUP49x7iV7xexehS*"
  #   Failed test 'Failed to list content of directory "/tmp/tmp.gi1eVM3fZ9/tmp.dNL0WHRTQ3/ci-node-bionic-amd64-02-04/LUP49x7iV7xexehS*"'
  #   in ./tools/clear//../../../utils/tap/tap.sh:fail() at line 159.

They fail because the wildcard character `*` is used in a path and is
not expanded when comes the time to list the content of the expanded
path.

In `--live` related tests, we use a wildcard in those paths because we
can't know the full name of the output directory as it's named by the
sessiond.

This bug was introduced (or rather not fixed) by the following commit:
  commit 94360c17201a28466af49058735166c73f9ae130
  Author: Francis Deslauriers <francis.deslauriers@efficios.com>
  Date:   Fri Mar 6 18:18:14 2020 -0500

      Fix: Tests: utils.sh: merge `validate_{directory,folder_is}_empty` functions

Solution
========
Remove the double quotes around the path variable so that wildcards can
be expanded as expected.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: If4a72fab0d487cf6e6b021dfad9eca6d0d60d5f4
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix: Tests: `gen-ust-events` doesn't error out on invalid option
Francis Deslauriers [Wed, 4 Mar 2020 21:18:37 +0000 (16:18 -0500)] 
Fix: Tests: `gen-ust-events` doesn't error out on invalid option

Issue
=====
When running `gen-ust-events` with an invalid option:
  ./gen-ust-events -h

The `getopt_long()` function prints the following error:
  ./gen-ust-events: invalid option -- 'h'

which is very kind of it.

The problem is that the process keep running and go on to generate
events. It should exit right away.

Solution
========
Remove the `break` statement and so that we execute the `goto end`
right away.

Apply the same changes to `gen-ust-nevents`.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ib1c96f4c9ed8f98395bf842215f858a69db2bbf0
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix: Tests: utils.sh: merge `validate_{directory,folder_is}_empty` functions
Francis Deslauriers [Fri, 6 Mar 2020 23:18:14 +0000 (18:18 -0500)] 
Fix: Tests: utils.sh: merge `validate_{directory,folder_is}_empty` functions

Issues
======
1. Both functions aim to do the same thing.

2. The current `validate_folder_is_empty` function uses an erroneous
   way of checking if a folder is empty. `ls -A` returns zero on
   success; it doesn't return the number of files and sub-folders. A
   correct way to get the number of elements in a directory is to pipe
   the output of `ls -A` to `wc -l`.

3. The `local_path` variable is used undefined in the current
   `validate_directory_empty` function.

   It just happens that the `local_path` variable is defined in all
   but one of the callsites of this function so those calls work as
   expected.

   The other call to this function was
   bogus (tests/regression/tools/clear/test_ust:323) because it uses
   the `$TRACE_PATH` variable name.

Fixes
=====
- Merge both functions into one, keeping the name
  `validate_directory_empty`,

- Fix the emptiness check,

- Fix usages of `validate_directory_empty` in tests.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Id75baf16f42866e3b978389f9aada4a2f6b6f2ae
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix: Tests: undefined `NR_USEC_WAIT` bash variable
Francis Deslauriers [Thu, 5 Mar 2020 19:09:51 +0000 (14:09 -0500)] 
Fix: Tests: undefined `NR_USEC_WAIT` bash variable

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I18eff11e47340245c595d4883f280d0fed23ca48
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agodirectory-handle: print `errno` reason on `unlinkat()` error
Francis Deslauriers [Mon, 9 Mar 2020 14:49:51 +0000 (10:49 -0400)] 
directory-handle: print `errno` reason on `unlinkat()` error

This is helpful while troubleshooting clear and rotation issues.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I92f9551a49fced7603c8dca832566de71b188528

19 months agoFix: lttng-ctl: _handle can be NULL
Jonathan Rajotte [Tue, 3 Mar 2020 20:30:10 +0000 (15:30 -0500)] 
Fix: lttng-ctl: _handle can be NULL

lttng_destroy_session_no_wait does not care for the returned handle
and passes a NULL _handle.

This leads to an immediate LTTNG_ERR_INVALID failure on
lttng_destroy_session_no_wait calls.

lttng_destroy_session_ext already performs a null check on _handle
before assigning it to transfer ownership of the handle.

Fixes #1241

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Change-Id: I9e8bee28382fa2250dee720037efdf77b4e776b8
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix: sessiond: domain subdirectory not deleted on empty clear
Jonathan Rajotte [Mon, 9 Mar 2020 18:45:20 +0000 (14:45 -0400)] 
Fix: sessiond: domain subdirectory not deleted on empty clear

Scenario
=====

$ lttng create
$ lttng enable-event -u -a
$ lttng start

// no events are produced.

$ lttng clear

Sessiond fails on clear with this message:

Error: Error removing subdirectory '.tmp_new_chunk' file when deleting
chunk

Cause
=====

The "ust" domain directory is still present preventing the removal of
the ".tmp_new_chunk" directory.

Solution
=====

Use lttng_trace_chunk_create_subdirectory, ensuring that the domain
folder is moved/deleted as necessary.

Fixes #1244

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Change-Id: I81d23ffc9f330f50f718957d460ee7a2718c95a1
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoIgnore -Wincomplete-setjmp-declaration warnings
Simon Marchi [Thu, 13 Feb 2020 15:35:34 +0000 (10:35 -0500)] 
Ignore -Wincomplete-setjmp-declaration warnings

We currently get this when building a test that requires UST:

    make[4]: Entering directory '/home/smarchi/build/lttng-tools-clang/tests/regression/ust/linking'
      CC       tp.lo
    In file included from /home/smarchi/src/lttng-tools/tests/regression/ust/linking/tp.c:11:
    In file included from /home/smarchi/src/lttng-tools/tests/regression/ust/linking/./ust_tests_demo.h:38:
    In file included from /home/smarchi/install/include/lttng/tracepoint-event.h:58:
    In file included from /home/smarchi/install/include/lttng/ust-tracepoint-event.h:28:
    In file included from /home/smarchi/install/include/lttng/ust-events.h:41:
    /usr/include/pthread.h:744:12: error: declaration of built-in
    function '__sigsetjmp' requires the declaration of the 'jmp_buf'
    type, commonly provided in the header
    <setjmp.h>. [-Werror,-Wincomplete-setjmp-declaration]
    extern int __sigsetjmp (struct __jmp_buf_tag *__env, int __savemask) __THROWNL;
               ^

I'm not sure what we can do about it, and I believe the warning is
bogus.  I do have a definition for the "jmp_buf" type in
/usr/include/setjmp.h:

    typedef struct __jmp_buf_tag jmp_buf[1];

And even with I include <setjmp.h>, the warning does not go away.  I'm
not sure if that's right, but this patch disables the warning.

Change-Id: Ibe7451dc0afc9aaca59431296d42011d9e4306f9
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agotests: put -no-pie in LDFLAGS instead of CFLAGS
Simon Marchi [Thu, 13 Feb 2020 15:36:04 +0000 (10:36 -0500)] 
tests: put -no-pie in LDFLAGS instead of CFLAGS

When building with clang, we get:

    make[4]: Entering directory
    '/home/smarchi/build/lttng-tools-clang/tests/utils/testapp/gen-syscall-events-callstack'
      CC       gen-syscall-events-callstack.o
    clang: error: argument unused during compilation: '-no-pie'
    [-Werror,-Wunused-command-line-argument]

Indeed, -no-pie should be in LDFLAGS, not CFLAGS.

To make sure that Makefiles can use `AM_LDFLAGS += ...`, I have added
an AC_SUBST for AM_LDFLAGS in configure.ac.  This makes it so that if
we ever set AM_LDFLAGS for real in configure.ac, the Makefiles won't
inadvertently overwrite that value.

Change-Id: I7bad985bb135e50750917db6a928f2705a85b445
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoSilence `POSIX Yacc` warnings
Jérémie Galarneau [Tue, 10 Mar 2020 02:06:01 +0000 (22:06 -0400)] 
Silence `POSIX Yacc` warnings

Flex/Bison warn that we make use of two non-POSIX directives,
%code and %define. Coincidentally, we don't use POSIX Yacc!

From the bison documentation, "[...] -Wno-yacc will hide the warnings
about POSIX Yacc incompatibilities." [1].

Thus, this flag is added to the YFLAGS.

  YACC     filter-parser.c
/home/jgalar/EfficiOS/src/lttng-tools/src/lib/lttng-ctl/filter/filter-parser.y:293.1-5:
warning: POSIX Yacc does not support %code [-Wyacc]
  293 | %code provides
      | ^~~~~
/home/jgalar/EfficiOS/src/lttng-tools/src/lib/lttng-ctl/filter/filter-parser.y:301.1-7:
warning: POSIX Yacc does not support %define [-Wyacc]
  301 | %define api.pure
      | ^

[1] https://www.gnu.org/software/bison/manual/html_node/Diagnostics.html

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iaf65534bc47514d3a99b97ce414880a6afc723b9

19 months agoFix -Wmissing-declarations warnings in filter-parser.y
Simon Marchi [Mon, 25 Nov 2019 20:32:19 +0000 (15:32 -0500)] 
Fix -Wmissing-declarations warnings in filter-parser.y

Make gc_string_append and yyerror static, as they are only used in
this file.

Remove yywrap, apparently, it's not used.  This probably has to do
with the fact that our lexer uses the noyywrap option.

setstring is used by filter-lexer.c (generated from filter-lexer.l),
which has its own local setstring declaration.  This is not very good,
because the prototype of setstring could change in the implementation,
and the users of the function would still compile, but then it would
probably crash inexplicably at runtime.

Instead, put the declaration in filter-parser.h, using a "%code
provides" directive.  That seems to be the intent of "%code provides"
[1]:

    Purpose: This is the best place to write additional definitions and
    declarations that should be provided to other modules.

[1] https://www.gnu.org/software/bison/manual/html_node/_0025code-Summary.html

Change-Id: I04ce69ae9808b72e96bf9e602772f0e5758dfc10
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoInclude cmd-2-2.h in cmd-2-1.h
Simon Marchi [Mon, 25 Nov 2019 20:39:22 +0000 (15:39 -0500)] 
Include cmd-2-2.h in cmd-2-1.h

Fixes:

      CC cmd-2-2.o
    /home/smarchi/src/lttng-tools/src/bin/lttng-relayd/cmd-2-2.c:36:5:
    error: no previous declaration for ‘cmd_recv_stream_2_2’
    [-Werror=missing-declarations] int cmd_recv_stream_2_2(const
    struct lttng_buffer_view *payload, ^~~~~~~~~~~~~~~~~~~

... and helps ensure that the declarations are in sync with the
definitions.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I889d7306c157b0d68d24bf610fce18c8949422d8
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoMake create_file function static in gen-ust-tracef.c
Simon Marchi [Mon, 25 Nov 2019 21:00:56 +0000 (16:00 -0500)] 
Make create_file function static in gen-ust-tracef.c

Fixes:

      CC gen-ust-tracef.o
    /home/smarchi/src/lttng-tools/tests/utils/testapp/gen-ust-tracef/gen-ust-tracef.c:36:6:
    error: no previous declaration for ‘create_file’
    [-Werror=missing-declarations] void create_file(const char *path)
    ^~~~~~~~~~~

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I12359b30054da609be2aca998de960719120083e
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoMake remove_file_from_hierarchy function static in test_directory_handle.c
Simon Marchi [Mon, 25 Nov 2019 21:16:32 +0000 (16:16 -0500)] 
Make remove_file_from_hierarchy function static in test_directory_handle.c

Fixes:

      CC test_directory_handle.o
    /home/smarchi/src/lttng-tools/tests/unit/test_directory_handle.c:139:5:
    error: no previous declaration for ‘remove_file_from_hierarchy’
    [-Werror=missing-declarations] int
    remove_file_from_hierarchy(struct
    lttng_directory_handle *test_dir_handle,
    ^~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I060a89d88f2de8e12368496968708d273cb318f2
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoMake fd_count function static in test_fd_tracker.c
Simon Marchi [Mon, 25 Nov 2019 21:17:37 +0000 (16:17 -0500)] 
Make fd_count function static in test_fd_tracker.c

Fixes:

      CC test_fd_tracker.o
    /home/smarchi/src/lttng-tools/tests/unit/test_fd_tracker.c:65:5:
    error: no previous declaration for ‘fd_count’
    [-Werror=missing-declarations] int fd_count(void) ^~~~~~~~

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I02775d5d601211052b4cad323bc33623beabc8f9
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoAdd declarations for exported functions in health_exit.c
Simon Marchi [Mon, 25 Nov 2019 21:21:52 +0000 (16:21 -0500)] 
Add declarations for exported functions in health_exit.c

Silences a bunch of -Wmissing-declarations warnings.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: Ib8ca00c2f9eaeec1821a15c3978d988e58804367
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoAdd declarations for exported functions in health_fail.c
Simon Marchi [Mon, 25 Nov 2019 21:25:52 +0000 (16:25 -0500)] 
Add declarations for exported functions in health_fail.c

Silences a bunch of -Wmissing-declarations warnings.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: If909cb5552aa9a8c9602410cbc608dd57a152e9f
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoMake functions in live_test.c static
Simon Marchi [Mon, 25 Nov 2019 21:27:13 +0000 (16:27 -0500)] 
Make functions in live_test.c static

Fixes:

      CC live_test.o
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:154:5:
    error: no previous declaration for ‘establish_connection’
    [-Werror=missing-declarations] int establish_connection(void)
    ^~~~~~~~~~~~~~~~~~~~
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:198:5:
    error: no previous declaration for ‘list_sessions’
    [-Werror=missing-declarations] int
    list_sessions(uint64_t *session_id) ^~~~~~~~~~~~~
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:245:5:
    error: no previous declaration for ‘create_viewer_session’
    [-Werror=missing-declarations] int create_viewer_session(void)
    ^~~~~~~~~~~~~~~~~~~~~
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:283:5:
    error: no previous declaration for ‘attach_session’
    [-Werror=missing-declarations] int attach_session(uint64_t id)
    ^~~~~~~~~~~~~~
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:372:5:
    error: no previous declaration for ‘get_metadata’
    [-Werror=missing-declarations] int get_metadata(void) ^~~~~~~~~~~~
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:466:5:
    error: no previous declaration for ‘get_next_index’
    [-Werror=missing-declarations] int get_next_index(void)
    ^~~~~~~~~~~~~~
    /home/smarchi/src/lttng-tools/tests/regression/tools/live/live_test.c:631:5:
    error: no previous declaration for ‘detach_viewer_session’
    [-Werror=missing-declarations] int detach_viewer_session(uint64_t
    id) ^~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I7dc82076439e1bccd7bc253d084291f6281f6887
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoAdd declarations for exported functions in consumer_testpoints.c
Simon Marchi [Mon, 25 Nov 2019 21:28:16 +0000 (16:28 -0500)] 
Add declarations for exported functions in consumer_testpoints.c

Silences some -Wmissing-declarations warnings.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: Icb6b79ea5ad4a076a6ae03c61b9f0691d45f3c1c
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoMake parse_arguments static in base_client.c
Simon Marchi [Mon, 25 Nov 2019 21:29:04 +0000 (16:29 -0500)] 
Make parse_arguments static in base_client.c

Fixes:

      CC base_client.o
    /home/smarchi/src/lttng-tools/tests/regression/tools/notification/base_client.c:60:5:
    error: no previous declaration for ‘parse_arguments’
    [-Werror=missing-declarations] int parse_arguments(char **argv) {
    ^~~~~~~~~~~~~~~

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I4e431009d6e174eaadd82d206f9bedde9efdd30b
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix all -Wmissing-declarations warning instances
Simon Marchi [Mon, 25 Nov 2019 21:41:29 +0000 (16:41 -0500)] 
Fix all -Wmissing-declarations warning instances

This fixes all the remaining -Wmissing-declarations warning occurences.
The changes either:

- Make functions static
- Remove unused functions
- Add declarations for functions that are meant to be exported in a
  shared object to override some other function (e.g.
  lttng_ust_clock_plugin_init).  There isn't really a better place for
  these declarations.

Change-Id: If08855b75bf44dfdcfbdd654c272474ad8ebef39
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
19 months agoFix: fd-tracker: mark symbols as hidden
Jérémie Galarneau [Mon, 2 Mar 2020 22:45:17 +0000 (17:45 -0500)] 
Fix: fd-tracker: mark symbols as hidden

The fd-tracker is part of libcommon and, as such, its symbols must be
marked as hidden since it is linked into liblttng-ctl.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I007d950c0e3f7e059ba372c4d20b096c3069febb

19 months agoFix: liblttng-ctl: hide new tracker config symbols
Jérémie Galarneau [Mon, 2 Mar 2020 22:12:17 +0000 (17:12 -0500)] 
Fix: liblttng-ctl: hide new tracker config symbols

Session configuration constants are supposed to remain internal.
Some symbols were errneously exported in previous releases, but
we mark new ones as "hidden".

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iae51c7e479ba9f9e377bb426f35f9e3a73c2d449

19 months agoTests: remove unused libhealthexit code
Jérémie Galarneau [Mon, 2 Mar 2020 19:48:41 +0000 (14:48 -0500)] 
Tests: remove unused libhealthexit code

libhealthexit is no longer used since 89c453960. Remove the now-unused
code of that test library. A comment referencing that library is also
adjusted.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I1d088e9032a3b2e1a9f7956e81c7cb662473a7fd

19 months agoFix: remove broken health monitoring test `test_thread_exit`
Jérémie Galarneau [Fri, 28 Feb 2020 23:02:17 +0000 (18:02 -0500)] 
Fix: remove broken health monitoring test `test_thread_exit`

The `test_thread_exit` test uses the testpoint() infrastructure to
simulate daemon threads dying at "random" times by calling
pthread_exit() and checks that dead threads are properly reported by
the health check API.

The health check system is implemented as a list of structures that
live in the TLS of the various monitored threads. When the health
thread receives a health-check request, it iterates on this list and
reports the current health status of the threads to the client.

When a thread is stalled, this works fine; the health TLS has not been
updated in some time and that can be observed by the client. However,
for a 'spurious' thread exit, this is a bit more fragile.

Essentially, the test assumes that the thread will not have
unregistered its health TLS, but that it will remain valid until the
thread is joined.

Unfortunately, this behaviour relies on the fact that threads were not
joined until late in the tear down of the session daemon in the
past. This is no longer the case for all threads.

To provide an example of a test sequence that results in a
crash:
  - the test kills the client thread,
  - the session daemon received SIGTERM,
  - the client thread is joined immediately,
  - the next thread to shutdown (gracefully) unregisters itself from
    the health monitoring sybsystem and, in doing so, accesses an
    invalid element while removing itself from the health_app list,
    causing a crash.

As a side note, I could not find a definitive answer on the lifetime
of TLS variables. Are they guaranteed to be accessible until a
pthread_join() or is it undefined to access them after a thread has
returned? I'm guessing this is a very internal implementation detail
of the pthread implementation being used and that we should not rely
on that behaviour.

There are multiple ways we could fix this problem, such as using
heap-allocated structures and ref-counting them to share ownership
with the health thread, using pthread_atexit() to clean-up, etc.

However, the LTTng daemons never pthread_exit() their threads to
handle errors (or even call it directly); handling this behaviour is
of dubious interest at the moment.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I16ec70eb683832fe5f2810885c995988b6194690

20 months agoFix: directory-handle: use of free'd handle on fstat() error
Jérémie Galarneau [Tue, 18 Feb 2020 01:13:49 +0000 (20:13 -0500)] 
Fix: directory-handle: use of free'd handle on fstat() error

On an error to fstat a directory handle, a directory handle is
released and is still initialized as if the error had not occurred.

Return NULL early from lttng_directory_handle_create_from_dirfd()
to prevent those erroneous accesses.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia2a3bb1cfe8c90d3f7f87a68286a9b8524694d3c

20 months agoFix: relayd: use of relay_session ref count before initialization
Jérémie Galarneau [Tue, 18 Feb 2020 01:05:22 +0000 (20:05 -0500)] 
Fix: relayd: use of relay_session ref count before initialization

The relay_session's reference count is used before it is initialized
on multiple code paths of session_create(). The initialization of the
reference count, mutexes, and intrusive data structure nodes are
initialized earlier to make their use safe in the event of an error.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I1be53ad88a3e783b85b4c568527df1a75ce58d3a

20 months agoFix: relayd: unchecked return value when opening relay socket
Jérémie Galarneau [Tue, 18 Feb 2020 00:32:54 +0000 (19:32 -0500)] 
Fix: relayd: unchecked return value when opening relay socket

fd_tracker_open_unsuspendable_fd may fail because the underlying
socket() call fails or there may be too many open file descriptors at
the time of the call. In both cases, these errors must be logged and
handled.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I8b8c4fcc9de08746a91778b58c74b2118e98667b

20 months agoconfigure: add --enable-Werror
Simon Marchi [Tue, 11 Feb 2020 22:29:09 +0000 (17:29 -0500)] 
configure: add --enable-Werror

If one wants to build with -Werror, it's not possible to simply
configure with -Werror in the CFLAGS.  This makes a bunch of configure
checks fail, which would have otherwise passed.

This patch adds an --enable-Werror option to configure, which has the
effect of adding -Werror to AM_CFLAGS.  It therefore ends up in the
CFLAGS used to build the project, but it doesn't interfere with the
configure checks.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I18c33125c717305aac8f1d8a19fee7e065d70c31
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agoconfigure: use AX_APPEND_COMPILE_FLAGS to detect supported warning flags
Simon Marchi [Fri, 31 Jan 2020 18:05:20 +0000 (13:05 -0500)] 
configure: use AX_APPEND_COMPILE_FLAGS to detect supported warning flags

I would eventually like to enable some additional warnings by default
when building lttng-tools.  However, some warnings are
compiler-specific or are not present in older versions of some compilers
we need to support.  We can therefore not add them unconditionally to
CFLAGS.

This patch uses the AX_APPEND_COMPILE_FLAGS macro to address that.  This
macro tests each individual flag we pass it with the current compiler.
If it finds that the flag is supported (the compiler exits with status 0
when compiling a file with that flag), it appends it to the given
variable, WARN_CFLAGS in our case).  WARN_CFLAGS is then added to
AM_CFLAGS.

With time, we'll be able to throw any warning flag in there that we
think is useful, even if just available in a recent version of one
compiler.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: Id2ae4b4e8882af788c835ce89a979544531370e9
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agotests: append to AM_CFLAGS instead of overriding it
Simon Marchi [Mon, 3 Feb 2020 22:38:23 +0000 (17:38 -0500)] 
tests: append to AM_CFLAGS instead of overriding it

The Makefiles modified by this patch currently override the AM_CFLAGS
value, which means that anything put in AM_CFLAGS by configure (for
example, the warning flags) is lost.

I believe the intention is to add some flags to CFLAGS, so modify them
such that they append instead of override.

notification/Makefile.am overrides AM_LDFLAGS with nothing.  It feels
like it was not the intention to actually clear that variable, but
rather that it is just the by-product of a copy paste.  If it was really
the intention to clear the value of AM_LDFLAGS, there would have been a
comment to explain it, right?

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I4ef926d9135b16200e5f17d09461506a5e955068
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agoTests: gen-ust-nevents: use options instead of arguments
Francis Deslauriers [Wed, 12 Feb 2020 00:27:05 +0000 (19:27 -0500)] 
Tests: gen-ust-nevents: use options instead of arguments

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I59c648c650304e12b30bf8a3eaedaf9727c48700
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agoTests: Cleanup: test_exclusion: more detailed output
Francis Deslauriers [Wed, 12 Feb 2020 00:17:36 +0000 (19:17 -0500)] 
Tests: Cleanup: test_exclusion: more detailed output

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I97baaf30385b644766f2e825d8884ea75770b330
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agoFix: Tests: `test_exclusion` passing for the wrong reason
Francis Deslauriers [Tue, 11 Feb 2020 23:27:35 +0000 (18:27 -0500)] 
Fix: Tests: `test_exclusion` passing for the wrong reason

Issue
=====
The following commit added `-i` and `-w` flags to the test app arguments
of the `test_exclusion` tests:
  commit 6c4a91d639747f260ab46decebc50998ef063712
  Author: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
  Date:   Mon Aug 26 14:22:06 2019 -0400

      tests: gen-ust-events: use options instead of arguments

      Remove argument dependency and ease usage of features individually.

The `gen-ust-nevents` was not modified to support those flags. I suspect this
mistake was caused by the name similarity of the `gen-ust-nevents` and the
`gen-ust-events` test applications.

We ended up calling the following command:
  ./gen-ust-nevents -i 100 -w -1

When called with such arguments the `gen-ust-nevents` parsed the first
argument (`-i`) using `atoi()` which retuned 0. This was interpreted as
the number of iterations requested by the user so the app immediately
exited without generating any events.

So, the test was not seeing any of the excluded events in the trace
which was then considered as a successful result but no events were ever
excluded because none were generated in the first place.

Solution
========
Remove the use of `-i` and `-w` flags.

I also added a `dry_run` test to confirm that we do indeed get events
when exclusions are not used to prevent this error from happening in the
future.

Notes
=====
- I changed the wildcard used in the enable-event command so to only
  enable events from the testapp and not the `lttng_ust_statedump:` events
  as those are generated even if we didnt' ask for them.

- I add a stderr redirection to `/dev/null` in the trace reading
  pipeline because we now end up with traces with no events. This has
  changed because we now only enable events from the application (see
  previous note).

- In a future commit, I will change the `gen-ust-nevents` application to
  take those `-i` and `-w` flags.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Id37dcd59a18b3401d97439bce1191a8c5cac87d5
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agoTests: notification.c: remove extra space
Francis Deslauriers [Tue, 11 Feb 2020 20:51:52 +0000 (15:51 -0500)] 
Tests: notification.c: remove extra space

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Iee893ab9d4aeb59e335af83b068adf716ac869f8
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
20 months agoFix: trace-chunk: useless assignment to 'ret'
Jérémie Galarneau [Tue, 18 Feb 2020 00:51:54 +0000 (19:51 -0500)] 
Fix: trace-chunk: useless assignment to 'ret'

'ret' is not used to report error in
lttng_trace_chunk_rename_path_no_lock(); status is used. Therefore,
this assignment is useless.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I1804d616a24c41d3956b021f90bb77d0e6efef1a

20 months agoFix: lttng: track-untrack: error assigned to wrong variable
Jérémie Galarneau [Tue, 18 Feb 2020 00:49:04 +0000 (19:49 -0500)] 
Fix: lttng: track-untrack: error assigned to wrong variable

CMD_ERROR is assigned to 'ret' rather than 'command_ret' when
an unknown left-over argument is passed to the track/untrack
command.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I5889a934d9859d0499eb26555658bf15af73f927

20 months agoFix: relayd: live: unchecked poll set creation return value
Jérémie Galarneau [Tue, 18 Feb 2020 00:46:18 +0000 (19:46 -0500)] 
Fix: relayd: live: unchecked poll set creation return value

The fd_tracker_util_poll_create function can fail because of fd
exhaustion or because the underlying epoll call fails.

In both cases, report and handle the error.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ie79fdc011afda43395ac883c6648f983118cfddb

20 months agoFix: relayd: live: unchecked return value when opening relay socket
Jérémie Galarneau [Tue, 18 Feb 2020 00:40:47 +0000 (19:40 -0500)] 
Fix: relayd: live: unchecked return value when opening relay socket

fd_tracker_open_unsuspendable_fd may fail because the underlying
socket() call fails or there may be too many open file descriptors at
the time of the call. In both cases, these errors must be logged and
handled.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3205896a5e8c83ceba02005a2b73f9466d26427c

This page took 0.087958 seconds and 4 git commands to generate.