lttng-tools.git
4 years agoconsumerd: refactor: split read_subbuf into sub-operations
Jérémie Galarneau [Sun, 10 May 2020 22:00:26 +0000 (18:00 -0400)] 
consumerd: refactor: split read_subbuf into sub-operations

The read_subbuf code paths intertwine domain-specific operations and
metadata/data-specific logic which makes modifications error prone and
introduces a fair amount of code duplication.

lttng_consumer_read_subbuffer is effectively turned into a template
method invoking overridable callbacks making most of the consumption
logic domain and data/metadata agnostic.

The goal is not to extensively clean-up that code path. A follow-up
fix introduces metadata buffering logic which would not reasonably fit
in the current scheme. This clean-up makes it easier to safely
introduce those changes.

No changes in behaviour are intended by this change.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I9366f2e2a38018ca9b617b93ad9259340180c55d

4 years agoconsumerd: move rotation logic to domain-agnostic read path
Jérémie Galarneau [Fri, 8 May 2020 20:00:11 +0000 (16:00 -0400)] 
consumerd: move rotation logic to domain-agnostic read path

The "rotation ready" logic is duplicated in both user space and kernel
specializations of the read subbuffer functions.

It is moved to the domain-agnostic caller where it is needed only
once. This makes it easier to implement a follow-up fix and reduces
code duplication.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iae952a2cd52fa458cec956ae219492557e4adf79

4 years agosessiond: enforce mmap output type for kernel metadata channel
Jérémie Galarneau [Tue, 5 May 2020 22:54:32 +0000 (18:54 -0400)] 
sessiond: enforce mmap output type for kernel metadata channel

A follow-up fix causes the consumer daemon to accumulate metadata
packets into a complete "unit" that can be parsed before sending it to
the relay daemon.

The consumer daemon will also need to extract the contents of the
metadata cache when computing a rotation position (follow-up fix too).

Hence, it is not possible to rely on the splice back-end as the
consumer daemon may need to accumulate more content than can be backed
by the ring buffer's underlying pages.

In both cases, the splice output mode could still be used when
combined with a memfd, but I see no tangible benefit. Moreover, it
would require a 3.17 kernel.

Curiously the kernel metadata channel configuration appears to be
hard-coded twice; once in the ltt_kernel_session's
ltt_kernel_metadata, and once again in
kernel_consumer_add_metadata(). kernel_consumer_add_metadata is
modified to use the kernel session's metadata configuration.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia4cad82f595d3eee50d081851c234d4c2ef7ee5f

4 years agoconsumerd: tag metadata channel as being part of a live session
Jérémie Galarneau [Tue, 5 May 2020 19:48:05 +0000 (15:48 -0400)] 
consumerd: tag metadata channel as being part of a live session

metadata channels that are part of a live session must be handled
differently than when they are part of non-live sessions since
complete "metadata units" must be accumulated before they are
forwarded to a relay daemon.

This allows a follow-up fix to use this information since the
live_timer_interval of a metadata channel is always 0.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I53db4bc717b149ed20e0309531db6f0241e873e1

4 years agoconsumerd: pass channel instance to stream creation function
Jérémie Galarneau [Tue, 5 May 2020 17:13:03 +0000 (13:13 -0400)] 
consumerd: pass channel instance to stream creation function

Both callsites of consumer_allocate_stream() set the stream's "chan"
pointer after the creation. Pass the channel directly to the stream
creation function so it can initialize the stream according to the
channel's settings.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Icea7088e7695e310585bf398e14e6443d67a30bb

4 years agoconsumerd: cleanup: use buffer view interface for mmap read subbuf
Jérémie Galarneau [Mon, 4 May 2020 23:04:02 +0000 (19:04 -0400)] 
consumerd: cleanup: use buffer view interface for mmap read subbuf

Replace explicit pointer + size parameters by an lttng_buffer_view
in lttng_consumer_on_read_subbuffer_mmap().

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I76f35b3e295c596cdf4bbb8a6d01168a850a721a

4 years agoconsumerd: move address computation from on_read_subbuffer_mmap
Jérémie Galarneau [Mon, 4 May 2020 22:21:48 +0000 (18:21 -0400)] 
consumerd: move address computation from on_read_subbuffer_mmap

The computation of the subbuffer's address is moved outside of
lttng_consumer_on_read_subbuffer_mmap to make it usable with a regular
buffer. This facilitates an upcoming change.

Moreover this has the benefit of isolating domain-specific logic from
this function which is supposed to be domain-agnostic.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I16f8ccaa73804f98fa03e69136548e6d6b7782e5

4 years agoconsumerd: refactor: combine duplicated check_*_functions
Jérémie Galarneau [Wed, 29 Apr 2020 04:03:43 +0000 (00:03 -0400)] 
consumerd: refactor: combine duplicated check_*_functions

The check_ust_stream and check_kernel_stream functions are identical
except for the call to the domain-specific call to
consumer_flush_*_index.

A "flush_index" callback is passed to check_stream in order to share
the rest of that code.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iafdb64192322c0106a555b67f54290dadc4f0579

4 years agokerner-ctl: add RING_RING_BUFFER_GET_NEXT_SUBBUF_METADATA_CHECK
Jérémie Galarneau [Wed, 29 Apr 2020 01:40:12 +0000 (21:40 -0400)] 
kerner-ctl: add RING_RING_BUFFER_GET_NEXT_SUBBUF_METADATA_CHECK

Add a wrapper for RING_RING_BUFFER_GET_NEXT_SUBBUF_METADATA_CHECK
which gets the next metadata subbuffer and returns a boolean flag
indicating whether the metadata is guaranteed to be in a consistent
state at the end of this sub-buffer (can be parsed).

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I13fbdfe51c3c4ef04581409e0fbc9837ed6d555d

4 years agoAdd lttng_dynamic_buffer_append_view util
Jérémie Galarneau [Fri, 15 May 2020 20:04:11 +0000 (16:04 -0400)] 
Add lttng_dynamic_buffer_append_view util

Add lttng_dynamic_buffer_append_view() which appends the contents
of a buffer view to a dynamic buffer.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I4082ba2c848b79aa2116847987067453638de441

4 years agoFix: lttng: Destroying session message repeated during destruction
Jérémie Galarneau [Thu, 14 May 2020 20:08:56 +0000 (16:08 -0400)] 
Fix: lttng: Destroying session message repeated during destruction

Observed Issue
==============

The `Destroying session X` is repeated indifinitely whenever
the data pending phase lasts more than one iteration.

```
$ lttng destroy
Destroying session eloi_turcotte.Destroying session
eloi_turcotte.Destroying session eloi_turcotte.Destroying session
eloi_turcotte.D
```

Cause
=====

Missing check that the message has been printed.

Solution
========

Use the same check as is done later for
lttng_destruction_handle_wait_for_completion().

Known drawbacks
===============

None.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I6cd29d917925644a4994c515b4177bbd05ffa98e

4 years agoFix: liblttng-ctl: leak of tracker handle in lttng_[un]track_pid
Jérémie Galarneau [Wed, 20 May 2020 19:56:03 +0000 (15:56 -0400)] 
Fix: liblttng-ctl: leak of tracker handle in lttng_[un]track_pid

The lttng_track_pid and lttng_untrack_pid functions were reimplemented
on top of the new lttng_process_attr_tracker_handle API (new in 2.12).

Both functions do not destroy the tracker handle on their return
path.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ieeeace2011ba6ee1e5024306ea735f0389c3980d

4 years agoFix: common: fs_handle_seek returns negative value on success
Jonathan Rajotte [Thu, 21 May 2020 00:53:45 +0000 (20:53 -0400)] 
Fix: common: fs_handle_seek returns negative value on success

Observed issue
==============

Babeltrace 1/2 fails to fetch data from a live session.

Error:

 PLUGIN/SRC.CTF.LTTNG-LIVE/VIEWER lttng_live_get_stream_bytes@viewer-connection.c:1593 [lttng-live] Received get_data_packet response: error
 PLUGIN/CTF/MSG-ITER request_medium_bytes@msg-iter.c:546 [lttng-live] User function failed: status=ERROR
 PLUGIN/CTF/MSG-ITER ctf_msg_iter_get_next_message@msg-iter.c:2881 [lttng-live] Cannot handle state: msg-it-addr=0x562d87521a40, state=DSCOPE_TRACE_PACKET_HEADER_BEGIN
 PLUGIN/SRC.CTF.LTTNG-LIVE lttng_live_iterator_next_handle_one_active_data_stream@lttng-live.c:821 [lttng-live] CTF message iterator failed to get next message: msg-iter=0x562d87521a40, msg-iter-status=ERROR
 PLUGIN/SRC.CTF.LTTNG-LIVE lttng_live_msg_iter_next@lttng-live.c:1499 [lttng-live] Error preparing the next batch of messages: live-iter-status=LTTNG_LIVE_ITERATOR_STATUS_ERROR
 LIB/MSG-ITER bt_message_iterator_next@iterator.c:865 Component input port message iterator's "next" method failed: iter-addr=0x562d8751ab70, iter-upstream-comp-name="lttng-live", iter-upstream-comp-log-level=WARNING, iter-upstream-comp-class-type=SOURCE, iter-upstream-comp-class-name="lttng-live", iter-upstream-comp-class-partial-descr="Connect to an LTTng relay daemon", iter-upstream-port-type=OUTPUT, iter-upstream-port-name="out", status=ERROR
 PLUGIN/FLT.UTILS.MUXER muxer_upstream_msg_iter_next@muxer.c:446 [muxer] Upstream iterator's next method returned an error: status=ERROR
 PLUGIN/FLT.UTILS.MUXER validate_muxer_upstream_msg_iters@muxer.c:989 [muxer] Cannot validate muxer's upstream message iterator wrapper: muxer-msg-iter-addr=0x562d87515280, muxer-upstream-msg-iter-wrap-addr=0x562d8751ca90
 PLUGIN/FLT.UTILS.MUXER muxer_msg_iter_next@muxer.c:1417 [muxer] Cannot get next message: comp-addr=0x562d8751a260, muxer-comp-addr=0x562d8751a2e0, muxer-msg-iter-addr=0x562d87515280, msg-iter-addr=0x562d8751aa90, status=ERROR
 LIB/MSG-ITER bt_message_iterator_next@iterator.c:865 Component input port message iterator's "next" method failed: iter-addr=0x562d8751aa90, iter-upstream-comp-name="muxer", iter-upstream-comp-log-level=WARNING, iter-upstream-comp-class-type=FILTER, iter-upstream-comp-class-name="muxer", iter-upstream-comp-class-partial-descr="Sort messages from multiple inpu", iter-upstream-port-type=OUTPUT, iter-upstream-port-name="out", status=ERROR
 LIB/GRAPH consume_graph_sink@graph.c:462 Component's "consume" method failed: status=ERROR, comp-addr=0x562d8751a3d0, comp-name="pretty", comp-log-level=WARNING, comp-class-type=SINK, comp-class-name="pretty", comp-class-partial-descr="Pretty-print messages (`text` fo", comp-class-is-frozen=0, comp-class-so-handle-addr=0x562d8751a110, comp-class-so-handle-path="/usr/local/lib/babeltrace2/plugins/babeltrace-plugin-text.so", comp-input-port-count=1, comp-output-port-count=0
 CLI cmd_run@babeltrace2.c:2529 Graph failed to complete successfully
 PLUGIN/SRC.CTF.LTTNG-LIVE/VIEWER lttng_live_session_detach@viewer-connection.c:1211 [lttng-live] Unknown detach return code 0

The relevant relayd log:

 DEBUG2 - Relay get data packet (in viewer_get_packet() at live.c:1770)
 PERROR - Failed to seek file system handle of viewer stream 4 to offset 2244861952: Success (in viewer_get_packet() at live.c:1810)
 DEBUG1 - Sent 262156 bytes for stream 4 (in viewer_get_packet() at live.c:1852)

Cause
=====

The fs_handle_seek function calls lseek on a stream file of ~2.5GB. The
return value of the lseek call is downcasted from off_t (signed 64 bit
on my system) to int. The resulting value is negative and force an error
at the call sites.

Solution
========

Use off_t as the return type.

Note that current call sites of fs_handle_seek already expect an off_t
return value.

ag fs_handle_seek:

 src/bin/lttng-relayd/stream.c
 249: lseek_ret = fs_handle_seek(previous_stream_file, previous_stream_copy_origin, SEEK_SET);

 src/bin/lttng-relayd/live.c
 1804: lseek_ret = fs_handle_seek(vstream->stream_file.handle,

 src/bin/lttng-relayd/viewer-stream.c
 176: lseek_ret = fs_handle_seek(

ag lseek_ret:

 src/bin/lttng-relayd/stream.c
 193: off_t lseek_ret, previous_stream_copy_origin;

 src/bin/lttng-relayd/live.c
 1760: off_t lseek_ret;

 src/bin/lttng-relayd/viewer-stream.c
 174: off_t lseek_ret;

Known drawbacks
=========

This limitation existed before this patch.

On 32bit system without _FILE_OFFSET_BITS=64 defined at compile time
(-D_FILE_OFFSET_BITS=64) the lseek operation could return EOVERFLOW for
stream file bigger then 2,147,483,647 bytes. Anybody working on a 32bit
system should be aware of the limitation of working in 32bit.

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ib6c7bb3c9c402bdfbe9b447b1f8f298de6058caa

4 years agoFix: common: abort on rotation after time manipulation
Jonathan Rajotte [Tue, 19 May 2020 16:23:18 +0000 (12:23 -0400)] 
Fix: common: abort on rotation after time manipulation

Observed issue
==============
Core dump:

 #0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
 #1  0x0000003eb4025548 in __GI_abort () at abort.c:79
 #2  0x0000003eb402542f in __assert_fail_base (fmt=0x3eb4184ae0 "%s%s%s:%u: %s%sAssertion `%s' failed.\n%n", assertion=0x4cdee0 "(trace_chunk->timestamp_close).is_set",
     file=0x4cde78 "../../../lttng-tools-2.11.3/src/common/trace-chunk.c", line=903, function=0x4cf4a0 <__PRETTY_FUNCTION__.6756> "lttng_trace_chunk_move_to_completed")
     at assert.c:92
 #3  0x0000003eb4033af2 in __GI___assert_fail (assertion=assertion@entry=0x4cdee0 "(trace_chunk->timestamp_close).is_set",
     file=file@entry=0x4cde78 "../../../lttng-tools-2.11.3/src/common/trace-chunk.c", line=line@entry=903,
     function=function@entry=0x4cf4a0 <__PRETTY_FUNCTION__.6756> "lttng_trace_chunk_move_to_completed") at assert.c:101
 #4  0x000000000047f37e in lttng_trace_chunk_move_to_completed (trace_chunk=0x7fcb5c00e570) at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:903
 #5  0x0000000000480755 in lttng_trace_chunk_release (ref=0x7fcb5c00e598) at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1117
 #6  urcu_ref_put (release=<optimized out>, ref=0x7fcb5c00e598) at /usr/include/urcu/ref.h:68
 #7  lttng_trace_chunk_put (chunk=0x7fcb5c00e570) at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1150
 #8  0x0000000000429c22 in cmd_rotate_session (session=0x7fcb5c003ff0, rotate_return=rotate_return@entry=0x7fcb6b7ed470, quiet_rotation=quiet_rotation@entry=false)
     at ../../../../lttng-tools-2.11.3/src/bin/lttng-sessiond/cmd.c:5037
 #9  0x00000000004451d7 in process_client_msg (cmd_ctx=0x7fcb5c00e760, sock=sock@entry=0x7fcb6b7fd4c0, sock_error=sock_error@entry=0x7fcb6b7fd4c4)
     at ../../../../lttng-tools-2.11.3/src/bin/lttng-sessiond/client.c:1852
 #10 0x00000000004474c6 in thread_manage_clients (data=<optimized out>) at ../../../../lttng-tools-2.11.3/src/bin/lttng-sessiond/client.c:2199
 #11 0x00000000004422f2 in launch_thread (data=0x4f97a0) at ../../../../lttng-tools-2.11.3/src/bin/lttng-sessiond/thread.c:75
 #12 0x0000003eb4408ed4 in start_thread (arg=<optimized out>) at pthread_create.c:479
 #13 0x0000003eb40f8e6f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

Reproduction:

 Disable ntp/any time management mechanism.

 lttng create
 lttng enable-event -u 'lttng_ust_tracef:*'
 lttng start
 lttng rotate
 date --set="$(date --date='-1 hour')"
 lttng rotate auto-20200515-142503
    Waiting for rotation to complete
    Error: Failed to query the state of the rotation.

Logs:
 DEBUG1 - 12:25:28.570037987 [2660/2717]: Setting trace chunk close command to "move to completed chunk folder" (in lttng_trace_chunk_set_close_command() at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1073)
 Error: Failed to set trace chunk close timestamp: close timestamp is before creation timestamp
 Error: Failed to set the close timestamp of the current trace chunk of session "auto-20200515-142503"
 lttng-sessiond: ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:903: lttng_trace_chunk_move_to_completed: Assertion `(trace_chunk->timestamp_close).is_set' failed.

 ...

 Aborted (core dumped)
 root@X10SDV-8C-TLN4F:~# DEBUG1 - 12:25:29.534263017 [2739/2739]: Releasing trace chunk registry to all trace chunks (in lttng_trace_chunk_registry_put_each_chunk() at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1414)
 DEBUG1 - 12:25:29.534317468 [2739/2739]: Releasing reference to trace chunk: session_id = 0chunk_id = 2, name = "20200515T122528+0000-2", status = closed (in lttng_trace_chunk_registry_put_each_chunk() at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1435)
 DEBUG1 - 12:25:29.534365653 [2739/2739]: Releasing reference to trace chunk: session_id = 0chunk_id = 1, name = "20200515T142520+0000-1", status = closed (in lttng_trace_chunk_registry_put_each_chunk() at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1435)
 DEBUG1 - 12:25:29.534400638 [2739/2739]: Released reference to 2 trace chunks in lttng_trace_chunk_registry_put_each_chunk() (in lttng_trace_chunk_registry_put_each_chunk() at ../../../lttng-tools-2.11.3/src/common/trace-chunk.c:1447)
 Error: 2 trace chunks are leaked by lttng-consumerd. This can be caused by an internal error of the session daemon.

Cause
=====
The trace_chunk->timestamp_close is not set since the result from time()
is smaller than the creation timestamp.

The close timestamp is smaller because the calendar system time is
modified by an administrator.

time() offers no monotonicity guarantee and hence is exposed to time
modification of the system.

The begin and close timestamps are strictly used in the name generation
of the chunk/archives. Given the current usage of these timestamps
validating monotonicity should not be a fatal error. Name uniqueness is
provided by the chunk name suffix (auto increment).

Solution
========
Do not enforce monotonicity for the begin and close timestamps but warn
on unexpected return (begin > close).

Known drawbacks
=========
None.

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ic4b17285d150358d1569d6821c451c243e64e9a1

4 years agoTests: test_exclusion: exclusion after tracing active
Francis Deslauriers [Tue, 5 May 2020 16:27:52 +0000 (12:27 -0400)] 
Tests: test_exclusion: exclusion after tracing active

This testcase tests for a UST bug where exclusions made when the tracing
is active are not excluding the undesired events.

This UST bug was fixed by the following commit:
  commit de713d8a77cbd77e60f58537b0fc222f98fde395
  Author: Francis Deslauriers <francis.deslauriers@efficios.com>
  Date:   Tue May 5 11:51:58 2020 -0400

      Fix: event probes attached before event enabled

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I6d896682d5f5e16582ab195c6f4d6946de394843
Depends-on: lttng-ust: Id984f266d976f346b001db81cd8a2b74965b5ef2

4 years agoTests: `gen-ust-nevents`: add syncpoints
Francis Deslauriers [Tue, 5 May 2020 16:19:13 +0000 (12:19 -0400)] 
Tests: `gen-ust-nevents`: add syncpoints

Adds two sync points:
  `--sync-in-main`: create a file when `gen-ust-nevents` when app is in
  main,
  `--sync-before-first-event`: wait on a file before starting to
  generate any events.

Those two sync points allow for tests to do work when the testapp has
reached main BUT before any events are generated.

This is useful to perform actions on a UST channel once tracing has
started and is active on a particular app.

For example, we want to test a scenario where events are enabled once an
app is already generating other events.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id501c8b373e1a9b43aad6caef11672fe6c30a55a

4 years agoTests: accept built-in kernel modules
Francis Deslauriers [Thu, 14 May 2020 18:56:05 +0000 (14:56 -0400)] 
Tests: accept built-in kernel modules

When validating that the kernel tracer is available on the system also
consider that it might be built directly in the kernel.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2b0e1767b8a6d561ad94ba38546cb183d9e98a95

4 years agoUpdate version to v2.12.1 v2.12.1
Jérémie Galarneau [Tue, 12 May 2020 15:12:08 +0000 (11:12 -0400)] 
Update version to v2.12.1

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
4 years agoAPI: missing includes in lttng.h
Jonathan Rajotte [Mon, 11 May 2020 18:21:51 +0000 (14:21 -0400)] 
API: missing includes in lttng.h

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I8ffc15f1ea855cdc6e08da6387936406ce9f7d75

4 years agoAPI: missing clear and clear-handle includes in lttng.h
Jonathan Rajotte [Mon, 11 May 2020 18:03:34 +0000 (14:03 -0400)] 
API: missing clear and clear-handle includes in lttng.h

Refs: #1266

Reported-by: Shuo Yang <shuoyang@didiglobal.com>
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia933831e806e51a071f05b2c0425799f2f6410cf

4 years agoAPI: sort lttng.h includes
Jonathan Rajotte [Mon, 11 May 2020 18:02:00 +0000 (14:02 -0400)] 
API: sort lttng.h includes

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I59699d6252c8b840cc3dc80316e0462c9420b465

4 years agoFix: API: missing end brace for C++ linkage specification.
Jonathan Rajotte [Mon, 11 May 2020 13:57:53 +0000 (09:57 -0400)] 
Fix: API: missing end brace for C++ linkage specification.

Fixes: #1266
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia8f2b4b55acc5770297d20df1b95575cf7fa480a

4 years agoFix: tests: `-Wstringop-overflow` warning
Francis Deslauriers [Mon, 4 May 2020 19:59:22 +0000 (15:59 -0400)] 
Fix: tests: `-Wstringop-overflow` warning

I get the following warning when compiling with `-Wall -Werror` with gcc
9.3.0:
  In file included from ../../src/common/macros.h:15,
                   from ../../src/common/lttng-kernel.h:14,
                   from ../../src/bin/lttng-sessiond/trace-kernel.h:14,
                   from test_kernel_data.c:16:
  In function ‘lttng_strnlen’,
      inlined from ‘lttng_strncpy’ at ../../src/common/macros.h:120:6,
      inlined from ‘test_create_kernel_event’ at test_kernel_data.c:136:2:
  ../../src/common/compat/string.h:28:8: error: ‘memchr’ reading 256 bytes from a region of size 11 [-Werror=stringop-overflow=]
     28 |  end = memchr(str, 0, max);
       |        ^~~~~~~~~~~~~~~~~~~

Fix this warning by using the RANDOM_STRING_LEN value as the max number
of bytes to copy.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I61752ee17163c4d642aad21b296c0fc4fad5b7a6
(cherry picked from commit 1dd622b10db0821d77490c937caee80c65332f14)
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
4 years agoFix: load: incomplete error handling for load_session_from_file
Jérémie Galarneau [Fri, 17 Apr 2020 19:49:52 +0000 (15:49 -0400)] 
Fix: load: incomplete error handling for load_session_from_file

This fix is adapted from a fix against the stable-2.11 branch.  The
commit message of the stable-2.11 branch follows.

An equivalent fix was already in place in `load_session_from_path()`,
but the same problem as the stable-2.11 branch is present in
`load_session_from_file()`.

Original message:

Observed issue
==============

lttng-ivc test fails to fail.

test_save_load_blocking_timeout[lttng-tools-2.12-lttng-tools-2.11-False]

Here we load a xml created by lttng-tools-2.12 and try to load it using
lttng-tools 2.11. We expect this to fail on the load.

The command report an error on the stderr but the command return code
value is zero.

From lttng-ivc test runtime.log:
Command #0
  Return value: 0
  Command: lttng load --input-path=/home/joraj/lttng/lttng-ivc/.tox/py3/tmp/test_save_load_blocking_timeou0/save_load saved_trace
  STDOUT:
      Session saved_trace has been loaded successfully
  STDERR:
      XML Error: Element 'process_attr_trackers': This element is not expected.
      Error: Session configuration file validation failed

Cause
=====

The error coming from load_session_from_file is not handled correctly.

Solution
========

Rework error handling in load_session_from_path and
load_session_from_file.

LTTNG_ERR_LOAD_SESSION_NOENT is NOT an error when session_name is
specified in load_session_from_path. In this scenario, we are actively
looking for the configuration of the session.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Change-Id: Ic68c253aa194bf8ab72c3c271f10d443118bdeee

4 years agoUpdate version to v2.12.0 v2.12.0
Jérémie Galarneau [Tue, 7 Apr 2020 18:01:56 +0000 (14:01 -0400)] 
Update version to v2.12.0

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7de0cfaebbd642d51fdc5229d2dcdf34765ca846

4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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

4 years 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

4 years 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

4 years 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

4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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>
4 years 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>
4 years 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

4 years agoUpdate version to v2.12.0-rc3 v2.12.0-rc3
Jérémie Galarneau [Fri, 27 Mar 2020 18:39:18 +0000 (14:39 -0400)] 
Update version to v2.12.0-rc3

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7e72fd3a7b710ae373dac88956772b1fe74619af

4 years agoFix: remove duplicated AC_INIT directive
Jérémie Galarneau [Fri, 27 Mar 2020 18:33:32 +0000 (14:33 -0400)] 
Fix: remove duplicated AC_INIT directive

The release script got confused by the new SPDX header which caused
the AC_INIT directive to get duplicated. This causes the configure
script to hang.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: If91e3db17ca46b697870ab37dc998b050ee7284c

4 years agoUpdate version to v2.12.0-rc2 v2.12.0-rc2
Jérémie Galarneau [Fri, 27 Mar 2020 16:53:42 +0000 (12:53 -0400)] 
Update version to v2.12.0-rc2

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I285efedca101d130eef5de700b8a8a7a0bd7558a

4 years 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

4 years 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

4 years 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

4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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

4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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

4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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

4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
4 years 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>
This page took 0.053067 seconds and 4 git commands to generate.