Mathieu Desnoyers [Fri, 9 Aug 2019 21:01:35 +0000 (17:01 -0400)]
Fix: test_crash: don't kill first app twice
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 9 Aug 2019 20:40:34 +0000 (16:40 -0400)]
Fix: perform a rotation to a null trace chunk on session destruction
In order to guarantee the consumption of all data by the time a
session's destruction completes, a rotation to a "null" trace chunk is
performed as part of a session's destruction.
This ensures that a session is only reclaimed when all of its
streams (on both the consumers and the relay daemon) have been
destroyed.
The notion of a "quiet" rotation is introduced in order to re-use the
session rotation infrastructure, but without the rotation being
visible to external users through either the notification system or a
rename of the trace output folder.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 9 Aug 2019 21:56:39 +0000 (17:56 -0400)]
relayd: log destination trace chunk of ROTATE_STREAMS command
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 9 Aug 2019 18:00:22 +0000 (14:00 -0400)]
Clean-up: declare dummy pipe payload as a char instead of a string
Only the first byte of the 'dummy' string is used. Hence, a single
char can be used.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Fri, 9 Aug 2019 16:25:57 +0000 (12:25 -0400)]
Fix: consumer: assert that stream chunk != NULL (not channel)
This assert sometimes triggers in
tests/regression/tools/crash/test_crash: test_shm_path_per_pid_sigint
because the destroy is performed when there is still unflushed data in
the buffers. The destroy performs a chunk close on the consumer daemon,
which effectively sets the channel chunk to NULL, while there are still
references to the chunk in the streams (which are still active).
Change the assertion in the stream read to validate that stream chunk !=
NULL instead.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 9 Aug 2019 15:18:47 +0000 (11:18 -0400)]
relayd fix: trace chunk is reclaimed before close command
The relay daemon control protocol defines a trace chunk close command
which allows the tracer to express the 'end' time bound of a trace
chunk.
However, in the event of a session rotation, the last reference to
such a trace chunk can be released before the close command is
received. This prevents the trace chunk from being renamed and moved
at the completion of a rotation.
A reference to a 'pending closure' trace chunk is kept as part of the
relay session object until its 'close' command is received.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 9 Aug 2019 15:11:28 +0000 (11:11 -0400)]
Fix: rotation of a stopped session hangs indifinitely
The relay daemon control protocol expects a rotation position as "the
sequence number of the first packet _after_ the current trace chunk.
At the moment when the positions of the buffers are sampled, the
production position does not necessarily sit at a packet boundary. The
'active' flush operation above will push the production position to
the next packet boundary _if_ it is not already sitting at such a
boundary.
Assuming a current production position that is not on the bound of a
packet, the 'target' sequence number is
(consumed_pos / subbuffer_size) + 1
Note the '+ 1' to ensure the current packet is part of the current
trace chunk.
However, if the production position is already at a packet boundary,
the '+ 1' is not necessary as the last packet of the current chunk is
already 'complete'.
In the case of a stopped session, performing the '+ 1'
indiscriminately results in a position that will not be reached until
another (not 'active') flush occurs.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 8 Aug 2019 20:39:56 +0000 (16:39 -0400)]
Fix: initialize kernel stream max subbuffer size on creation
A consumer stream's maximal subbuffer size is only initialized while
taking a snapshot. This unexpectedly leaves the max_sb_size at 0
causing the rotation target sequence number computation to crash with
a SIGFPE.
This fix initializes the value as soon as possible to provide a
behaviour in line with that of the ust-consumer.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 8 Aug 2019 20:38:48 +0000 (16:38 -0400)]
Fix: use relay_id from ust_session or kernel_session consumer
The consumer_output of an ltt_session does not contain the relay
daemon id in the case of a remote session. Hence, it must be fetched
from the domain-specific sessions.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 8 Aug 2019 20:29:38 +0000 (16:29 -0400)]
Fix: lttng-ctl: public session.h control API: restore 0 success return value
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 8 Aug 2019 19:27:51 +0000 (15:27 -0400)]
Fix: remote kernel stream paths contain too much information
The kernel stream paths, when built with a "net" consumer type,
contain the hostname and session name. This is redundant and must not
be transmitted to 2.11+ relay daemons. Otherwise, this results in an
erronerous path hierarchy on the remote end.
Stream paths transmitted to a trace chunk-aware peer are relative to a
session's output trace chunk.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 8 Aug 2019 17:53:19 +0000 (13:53 -0400)]
Fix: destroy command: send credentials with command
Not sending credentials triggers spurious destroy command failure
(EPERM) because session daemon receives "nobody.nobody" credentials.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 8 Aug 2019 04:55:51 +0000 (00:55 -0400)]
relayd: open live viewer files from the current stream's trace chunk
Since relay_streams' paths are now relative to a session's trace
chunk, file operations performed by the live protocol handler must
occur through a trace chunk.
Note that viewer streams hold a reference to the current trace chunk
of their corresponding relay stream. This is fine right now as no
session rotations may occur during the lifetime of a live session. If
this was allowed, this would be a problem since a session rotation
would not complete until all live viewer streams release their
reference to the trace chunk.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 7 Aug 2019 18:27:24 +0000 (14:27 -0400)]
Fix: don't perform an automatic session rotation in snapshot mode
An automatic session rotation is performed on destruction if more than
one trace chunk was created during the lifetime of a session. This is
not a good criterion to use in no-output/snapshot mode as multiple
trace chunks may have been created during a session's lifetime.
The "output_traces" flag is checked explicitly to prevent an automatic
session rotation from happenning on destroy.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 7 Aug 2019 18:21:11 +0000 (14:21 -0400)]
Fix: incorrect time unit used when waiting on destruction handle
DEFAULT_DATA_AVAILABILITY_WAIT_TIME is used by the lttng client to
wait on the destruction handle returned by a session destruction
command. This constant is in micro-seconds while the destruction
handle expects a timeout in ms.
This causes the client to appear unresponsive while waiting for
the destruction of a session.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 1 Aug 2019 18:42:32 +0000 (14:42 -0400)]
relayd: implement file and session rotation on top of trace chunks
Implement the file and session rotation functionality on top of the
trace chunk API. This ensures that a relay_stream and lttng_index_file
are always explicitly associated to a trace chunk and hold a reference
to it as long as their underlying files are contained within a given
trace chunk.
A number of relay_stream specific functions are moved to stream.c as
"methods" of the relay_stream interface in order to make use of
internal relay_stream helpers.
As part of this clean-up/move of the relay_stream code, raw payload
buffer handling has been replaced to use the lttng_buffer_view
interface which provides implicit bounds checking of the payload
buffers.
The stream rotation has been modified to reference a "new chunk id"
which is the ID of the trace chunk to which streams should rotate
"into". The command has also been modified to apply on a set of
streams. This is done in order to limit the number of commands on the
control socket. Conversely, all path names have been removed from the
command's payload.
The index file implementation now acquires a reference to the trace
chunk from which it is created. This affects the consumer daemon as
this code is shared with the relay daemon. This ensures that a chunk
is not released (and its close command executed, if any) before all
file descriptors related to it have been closed. Respecting this
guarantee is very important as the upcoming fd-cache will remove
the guarantee that an "fd" to a given file is always held open.
Moreover, close commands can rename a trace chunk's folders which
would cause files to be created in the wrong folder if they are
not properly created through the trace chunk.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 1 Aug 2019 20:37:15 +0000 (16:37 -0400)]
Fix: remove assertions of the existence of a trace chunk
A session can have no trace chunk on the session daemon and
consumer daemon end whenever its ouput is remote (relayd) and
doesn't support trace chunks.
The assertions removed as part of this patch are overly strict
considering that there are valid cases when trace_chunk can
be NULL.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 1 Aug 2019 20:35:27 +0000 (16:35 -0400)]
Fix: only create trace chunk if the output of a session supports it
A trace chunk should only be created on the session daemon end when
the trace's output supports this notion (2.11+).
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 31 Jul 2019 03:37:39 +0000 (23:37 -0400)]
Fix: use stream's current tracefile count rather than the max
consumer_stream_create_output_files() formats the file name of a
stream's data file by passing the tracefile count setting rather than
using `tracefile_count_current`, the current "index" within the
on-disk ring buffer. This broke the trace file rotation functionality
as the same file would be overwritten continuously.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 25 Jul 2019 19:35:21 +0000 (15:35 -0400)]
relayd: use bool for is_metadata relay_stream attribute
The `is_metadata` attribute of a relay_stream is currently
a int32_t while it is used as a boolean.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 25 Jul 2019 19:30:54 +0000 (15:30 -0400)]
relayd: create stream files relative to a session's trace chunk
Like to the consumer daemon, the relay daemon now creates its
stream files through the lttng_trace_chunk interface. This
requires changes across the session daemon and relay daemon.
The changes to the relay daemon mainly target the file creation.
In the session daemon, the snapshot command has been heavily
modified to fully initialize the ltt_session's ust and kernel
consumer_outputs _before_ the snapshot trace chunk is created.
This way, the session is setup in the same way a regular network
streaming session would be, a temporary trace chunk is created to
contain the snapshot's content, the actual snapshot is taken, and
the session's original consumer_outputs are restored.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 24 Jul 2019 21:07:00 +0000 (17:07 -0400)]
Fix: leak of consumer_output when using an explicit snapshot output
The consumer_output of a snapshot_output is leaked whenever a
snapshot is taken on a session that doesn't have pre-configured
snapshot outputs.
Steps to reproduce the leak:
$ lttng create --snapshot
$ lttng snapshot del-output 1
$ lttng enable-event -u -a
$ lttng start
$ lttng snapshot record /tmp/leak_it
This results in a leak of ~24kb/snapshot as reported by valgrind.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 23 Jul 2019 20:45:48 +0000 (16:45 -0400)]
Fix: reception buffer of control connection is leaked
The reception buffers of control connections are not "reset" on
destruction, resulting in their underlying storage buffer being
leaked. The typical (observed) size of the leak is 8kb per control
connection.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 22 Jul 2019 21:55:38 +0000 (17:55 -0400)]
Fix: ensure a newline is printed before exit on client destroy
Errors encountered during the destruction of a session can cause
the `lttng` client to exit after having printed the
'Waiting for destruction of session ...' message without printing
a newline. This causes the user's prompt to reappear after the
message and not on a new line.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 19 Jul 2019 22:54:10 +0000 (18:54 -0400)]
Send session creation time to relay daemon when supported
A recent commit changed the way the creation timestamp was added to a
session's output path; the session creation timestamp is now (2.11+)
sampled at creation time and formatted when the session's output
directory is created.
The 2.11+ version of the session and stream creation commands allow
the relay daemon to use all components of the path independently to
format an output path rather than relying on the session daemon peer
to format it ahead of time.
In order to maintain the timestamped session folder name created by
previous versions, the relay daemon now receives the session's
creation timestamp and formats it rather than relying on it being
"cooked" into the transmitted session name.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 18 Jul 2019 20:11:59 +0000 (16:11 -0400)]
relayd: remove unnecessary allocation in session path formatting
`alloc_path` merely contains the contents of `default_path`.
The allocation of `alloc_path` can be removed and `default_path`
used in its place.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 18 Jul 2019 19:32:57 +0000 (15:32 -0400)]
relayd: add remote trace chunk close command
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 17 Jul 2019 19:12:49 +0000 (15:12 -0400)]
Create userspace buffers using ua_sess effective credentials
While a session's trace chunk shares its credentials, a trace
chunk's credentials are unset when the output is not local. Hence,
the session's effective credentials must be used.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 17 Jul 2019 18:46:56 +0000 (14:46 -0400)]
Use lttng_credentials instead of bespoke uid/gid members in ua_sess
Change euid/egid and uid/gid to use the lttng_credentials structure
under the "effective_credentials" and "real_credentials" names.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 16 Jul 2019 19:31:36 +0000 (15:31 -0400)]
relayd: add remote trace chunk creation command
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 16 Jul 2019 18:45:36 +0000 (14:45 -0400)]
Add .clang-format
Add a first draft of a .clang-format. I have attempted to set it up
so that it conforms to the project's usual coding convention.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 16 Jul 2019 01:24:41 +0000 (21:24 -0400)]
Don't allow slashes and dots in overriden trace chunk names
A trace chunk's name is used when building its path. Therefore,
we ensure that a chunk's name can never contain '/' or '.' in
order to prevent malicious trace chunk names that would escape
a session's output directory.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 4 Jul 2019 13:04:53 +0000 (09:04 -0400)]
relayd: Retrieve a relay_session's trace chunk on creation
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 4 Jul 2019 13:03:54 +0000 (09:03 -0400)]
relayd: Transmit current trace chunk id in create_sesssion command
The relay daemon needs the current chunk id to query (when it is set)
the sessiond trace chunk registry and retrieve the instance of the
trace chunk if it is already available during the session's
creation.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 4 Jul 2019 12:59:08 +0000 (08:59 -0400)]
Use a "comm" variant of the LTTNG_OPTIONAL helper in sessiond-comm
The LTTNG_OPTIONAL helper is currently communication safe (as long
as LTTNG_PACKED is used). However, this could change in the future.
An LTTNG_OPTIONAL_COMM alias is introduced in order to make it easier
to change the layout of LTTNG_OPTIONAL should it become necessary
in the future.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 3 Jul 2019 20:23:28 +0000 (16:23 -0400)]
Remove unused relay daemon commands
Remove the now-unused rename, mkdir, and rotation pending commands
that are replaced by the trace chunk infrastructure.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 19 Feb 2019 21:48:23 +0000 (16:48 -0500)]
relayd: create an implicit trace chunk on session creation
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 12 Mar 2019 21:20:34 +0000 (17:20 -0400)]
relayd: create sessiond trace chunk registry on session creation
Use the sessiond_trace_chunk_registry_session_created() and
sessiond_trace_chunk_registry_session_destroyed() to allow the
sessiond trace chunk registry the ability to manage trace chunk
registries associated with the various session daemons from which
the consumer daemon connections originate.
These notifiers effectively manipulate the reference count (and
create/destroy, as needed) of the lttng_trace_chunk_registry
associated with each session daemon (using their UUID).
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 2 Jul 2019 19:00:39 +0000 (15:00 -0400)]
Use directory handle to rename trace chunk directory
Replace uses of renameat() with the directory handle
compatiblity-layer for systems that don't support directory
file descriptors.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sat, 29 Jun 2019 03:01:23 +0000 (23:01 -0400)]
Docs: add directory handle header documentation
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 28 Jun 2019 21:14:11 +0000 (17:14 -0400)]
Add rmdirat and renameat to run-as commands
Add support for the rmdirat and renameat commands to the run-as
infrastructure. These commands use the directory_handle compatibility
layer to provide rename and directory deletion relative to a
(or multiple) directory file descriptors.
The rmdirat name is used even though there are no rmdirat syscall
(at least on Linux). The functionality that would be provided by
rmdirat() (vs rmdir) is provided through unlinkat(..., AT_REMOVEDIR).
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 24 Jun 2019 20:41:50 +0000 (16:41 -0400)]
Print the location of trace chunk produced at session destruction
The destruction of a session that was rotated during its lifetime
will cause an "implicit" rotation to close the last trace archive.
This change makes the CLI client use the newly added destruction
handle interface of liblttng-ctl to print the location of the
last trace archive produced during the destruction of a session
(when applicable).
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 24 Jun 2019 19:08:36 +0000 (15:08 -0400)]
Docs: misleading documentation of compat_epoll_create
The documentation of compat_epoll_create explicitly mentions that
it can't fail because of a memory allocation error. This is false
as the function allocates an array of struct epoll_event.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 24 Jun 2019 19:07:59 +0000 (15:07 -0400)]
Remove unused ust_sock member of command_ctx
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 18 Jun 2019 18:12:45 +0000 (14:12 -0400)]
Tests: clarify the tap output of the kernel session rotation test
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 18 Jun 2019 18:05:46 +0000 (14:05 -0400)]
Fix: wait for the completion of implicit session rotations
A session that has been rotated within its lifetime will be
rotated during its destruction to rename the last trace chunk.
Currently, the liblttng-ctl library's session destruction function
only waits for the data pending to indicate that all data has
been consumed. This used to be sufficient, but it is now necessary
to wait for the implicit session rotation to complete.
The "wait" variant of the session destruction function will wait
for any implicit session rotation to compete. A new
lttng_destruction_handle class is introduced to allow a client to
wait for the destruction of a session and obtain the location of
the last trace archive produced by the implicit session rotation.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 13 Jun 2019 18:41:48 +0000 (14:41 -0400)]
Fix: don't hide EBADF error subbuffer write
There are _no_ legitimate circumstances where a "EBADF" error should
be considered normal in a multi-threaded program.
This doesn't fix a known problem, but it hid an error while testing a
development branch. It is considered a fix since not logging this
error can make debugging more difficult if this was, in fact, hiding a
problem.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 11 Jun 2019 20:03:22 +0000 (16:03 -0400)]
Tests: don't expect a UST directory after inactive per-PID rotation
The trace archive resulting from the rotation of a session containing
per-PID channels does not result in the creation of te domain
subfolders.
The trace archive is empty if none of the traced domain were active
during the trace chunk's capture.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 11 Jun 2019 09:39:46 +0000 (05:39 -0400)]
Tests: remove check for an empty chunk produced on destruction
Consider the following tracing scenario used by the rotation tests:
- Start tracing
- Rotate
- Stop tracing
- Rotate
- Destroy
The session daemon used to produce 3 trace archives when this sequence
of operations occurred. Of course, the third archive would always be
empty as no events can be produced between a stop and a destroy
operation.
As this behaviour is thankfully no longer present, the chunk
validation logic of the test causes it to fail. This patch removes
this logic.
Moreover, since chunk IDs are now 0-based, the test is adjusted to
check for names accouting for this naming scheme.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 10 Jun 2019 17:31:31 +0000 (13:31 -0400)]
Fix: metadata stream is not marked as quiescent after packet commit
When a metadata stream's wait fd is hung-up or enters an error state,
it is checked for quiescence in lttng_ustconsumer_on_stream_hangup().
If the stream is not quiescent, the current packet is closed through
the flush_buffer operation.
Currently, all commits to metadata streams are done on a packet
basis. The various code paths using the commit_one_metadata_packet
helper all perform a flush directly after the commit. Performing this
flush leaves the stream in a "quiescent" state, but does not mark it
as such.
This results in an extraneous flush being performed in the err/hup
handler, which leaves an empty packet to be consumed. This packet is
then consumed during the execution of the err/hup handler.
This bug results in an empty packet being appended to metadata
streams. This packet is typically ignored by readers, but the fact
that it is written at the time of the destruction of a session
violates the immutability guarantee of the session stop
command. Moreover, following the introduction of trace chunks, this
results in the stream attempting to serialize the empty buffer to its
output file _after_ its trace chunk has been closed, causing an
assertion to hit.
Hence, this fix performs the buffer flush and sets the stream as
quiescent directly in commit_one_metadata_packet().
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 21 Feb 2019 23:59:31 +0000 (18:59 -0500)]
Create stream files relative to a stream's current trace chunk
Create stream (and metadata) files relative to a session's current
trace chunk using the lttng_trace_chunk_open/unlink[...] functions
in the consumer daemons.
Four new commands are added to the sessiond <-> consumerd protocol:
- CREATE_TRACE_CHUNK
Command parameters:
- relayd_id: uint64_t
Unique ìd of the session's associated relay daemon connection
- override_name: optional char[]
Overriden chunk name. This field is not used by the consumer
daemon; it is forwarded to the relay daemon in order to set the
name of a trace chunk's directory when it should not follow the
`<ts begin>-<ts end>-<index>` form used by trace archives (i.e. as
produced by session rotations). This is used to preserve the
existing format of snapshot output directory names.
- sessiond_id: uint64_t
Unique id of the session (from the sessiond's perspective) to
which this trace chunk belongs.
- chunk_id: uint64_t
Unique id the the session's trace chunk.
- credentials: pair of uint32_t uid/gid
Credentials the consumer daemon should use in order to create
files within the trace chunk.
The session daemon maintains the current lttng_trace_chunk of an
ltt_session. When a session that has an output (`output_traces` == 1),
an lttng_trace_chunk is created.
In local tracing modes, the current trace chunk of a session, on the
session daemon's end, holds the ownership of the chunk's output
directory.
The CREATE_TRACE_CHUNK command is used to replicate the session
daemon's current trace chunk in the consumer daemon. This
representation of the current trace chunk has a different role. It
is created in "user" mode. Essentialy, the trace chunk's location
is opaque to the consumer daemon; it receives a directory file
descriptor from which a number of stream files will be created.
The trace chunk registry, as used by the consumer daemon, implicitly
owns the trace chunks on behalf of the session daemon. This is only
needed in the consumer since the consumer has no notion of a session
beyond session IDs being used to identify other objects.
When a channel is created, its session_id and initial chunk_id are
provided. This allows the consumer daemon to retrieve the session's
current trace chunk and associate it with the newly-created
channel. The channel holds a reference to its current trace chunk.
Streams created from a channel also hold a reference to their current
trace chunk, as retrived from their "parent" channel.
The life time of trace chunks in the consumer daemon is cooperatively
managed with the session daemon. This means
session daemon through the LTTNG_CONSUMER_CREATE_TRACE_CHUNK and
LTTNG_CONSUMER_CLOSE_TRACE_CHUNK commands.
- CLOSE_TRACE_CHUNK
[... TODO ...]
This command is used to release the global reference to a
given trace chunk in the consumer daemon. Relasing the consumer
daemon's global reference to the trace chunk leaves only the streams
to hold references until the moment when they are either closed or
they switch-over to another chunk in the event of a session rotation.
- TRACE_CHUNK_EXISTS
[... TODO ...]
- ADD_TRACE_CHUNK_CLOSE_COMMAND
[... TODO ...]
This commit changes a lot of code since it essentialy changes how
files and directories are created.
A number of commands no longer need to specify a `trace_archive_id`
since the CREATE_TRACE_CHUNK and CLOSE_TRACE_CHUNK allow the consumer
daemon to keep track of the current trace chunk of a channel at any
given time.
Creation and ownership of channel sub-directories
---
The path expressed in consumer channel objects is now relative to
the current trace chunk rather than being absolute. For example,
the `pathname` of a consumer channel is now of the form
`ust/1000/64-bit` rather than containing the full output path
`/home/me/lttng-traces/session-[...]/ust/1000/64-bit/`.
The subdirectory of a channel (relative to a trace chunk, e.g.
`ust/1000/64-bit`) is lazily created when a stream's output files are
created.
To do so, the `lttng_consumer_channel` now has a
`most_recent_chunk_id` attribute. When a stream creates its output files
(i.e. at the beginning of a session, or during a session rotation),
the stream's current trace chunk `id` is compared to the channel's
`most_recent_chunk_id`. If it is determined that the channel is
entering a new trace chunk, its channel subdirectory is created
relative to the stream's chunk.
Since this new state is within the `lttng_consumer_channel`, the
channel lock must be held on code paths that may result in the
creation of a new set of output files for a given stream.
Note that as of this commit, there is now a clear ownership boundary
between directories, owned by the session daemon through its trace
chunk, and files, owned by the consumer daemon.
Down-scoping of channel credentials
---
Since files are now created relative to their stream's current trace
chunk (which has credentials set), the fewer sites need access to the
channel's credentials. The only reason credentials are kept as part
of the consumer channel structure is the need to open and unlink UST
shared memory mappings. Since the credentials must only be used for
this purpose, they are now stored as an `LTTNG_OPTIONAL` field,
buffer_credentials, that is only set for UST channels.
Stream files should never need those credentials to be created.
The following commands sessiond <-> consumerd commands have been
removed:
- LTTNG_CONSUMER_ROTATE_RENAME
- LTTNG_CONSUMER_CHECK_ROTATION_PENDING_LOCAL
- LTTNG_CONSUMER_CHECK_ROTATION_PENDING_RELAY
- LTTNG_CONSUMER_MKDIR
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 22 May 2019 20:28:39 +0000 (16:28 -0400)]
Cleanup: remove duplicated code in snapshot record command
The session daemon's snapshot record command contains duplicated
code depending on the configuration of the snapshot output. This
makes it harder to modify the code in a follow-up commit.
Those code paths are made const-correct by the same occasion.
No behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 23 May 2019 17:34:37 +0000 (13:34 -0400)]
Fix: run_as_mkdirat returns value of errno
The run_as_mkdirat() function returns the _errno value returned
by the run-as worker instead of returning `ret`. This causes
run_as_mkdirat() to return an error when errno was already set
to a non-zero value by a previous libc/system call.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 15 May 2019 15:37:55 +0000 (11:37 -0400)]
Fix: double close of directory fd in runas worker
The run-as worker implements its mkdirat and mkdirat_recursive commands
on top of the lttng_directory_handle interface. When a directory
handle is created from a directory file descriptor, it assumes the
ownership of this file descriptor and it will release it when
lttng_directory_handle_fini() is invoked.
The runas worker's post-command clean-up closes any file descriptor
received from its client. The command structure's `fd` is set to
`-1` in order to make the received file descriptor invalid and
prevent the double-close.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 11 Mar 2019 21:25:45 +0000 (17:25 -0400)]
relayd: add the sessiond_trace_chunk_registry interface
The sessiond_trace_chunk_registry interface allows the relay
daemon to maintain an association between multiple consumer daemon
connections originating from the same session daemon and a common
trace chunk registry.
This trace chunk registry allows separate, but related, connections
to manipulate the same trace chunk.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sat, 30 Jun 2018 18:51:55 +0000 (14:51 -0400)]
Move index initialization to ctf-index.h
This initialization code is moved to a common header to re-use
it in a follow-up patch.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 10 May 2019 16:41:26 +0000 (12:41 -0400)]
Export utils_stream_file_path outside of common/utils.c
utils_stream_file_name is used by the stream file creation and
unlinking utilities. Those utilities perform actions that
will be performed on a trace chunk in a future commit.
Making this util public allows the reuse of most of the code
of the utilities.
The function is renamed from utils_stream_file_name to
utils_stream_file_path since it returns a path and not a
name.
The function's output parameter is moved to the end of the argument
list to honor the convention used in most of the code base. A length
parameter is added rather than relying on an implicit assumption
that the buffer's length is >= PATH_MAX.
The existing callers' buffer size is changed from PATH_MAX to
the preferred LTTNG_PATH_MAX.
The implementation of the function has been reworked to eliminate
a number of unncessary copies and allocations. It is now simpler,
handles path truncations, and does not duplicate the path separator
if it is already contained in `path_name`.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 9 May 2019 18:10:58 +0000 (14:10 -0400)]
Add file creation/unlinking utils to directory handle
This commit adds file creation and unlinking functions to the
directory handle interface and corresponding run_as wrappers.
This allows the creation and unlinking of files relative to
an existing directory handle using either openat/unlinkat when
directory file descriptors are supported or open/unlink when
they are not.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 5 Mar 2019 03:16:10 +0000 (22:16 -0500)]
relayd: send sessiond uuid and session id as part of create session
Send the sessiond's instance UUID and session id as part of the
relay daemon control protocol's CREATE_SESSION command.
Those attributes will be used to uniquely identify trace chunks
in follow-up commits.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 14 Dec 2018 21:04:44 +0000 (16:04 -0500)]
Transmit sessiond uuid to consumerd on launch
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 12 Mar 2019 20:53:27 +0000 (16:53 -0400)]
Add lttng_uuid_copy() util
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 12 Mar 2019 20:53:00 +0000 (16:53 -0400)]
Add lttng_uuid_is_nil() util
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 11 Mar 2019 21:23:21 +0000 (17:23 -0400)]
Add lttng_uuid_is_equal util to lttng_uuid
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 12 Dec 2018 21:16:26 +0000 (16:16 -0500)]
Add a consumer daemon INIT command
This commit adds an INIT commands which allows the session daemon
to provide its UUID as soon as the consumer daemon is started.
The name of this command is generic since other attributes may
need to be passed at the initialization of a consumer daemon.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 12 Dec 2018 20:39:53 +0000 (15:39 -0500)]
Generate a UUID on lttng-sessiond launch
This commit adds a global sessiond_uuid which can be used to
uniquely identify a session daemon from a relay daemon's
perspective even when it communicates with multiple consumers.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 27 Feb 2019 17:16:29 +0000 (12:16 -0500)]
Remove unneeded creation of the UST domain sub-directory
ust_app_start_trace() always attempts to create its domain
sub-directory after the channel directory has been created. This is
redundant and can be safely removed.
To provide an example, a "domain sub-directory" is:
/home/jgalar/lttng-traces/allo-
20190227-115205/ust
A "channel sub-directory" is, in per-uid buffering mode:
/home/jgalar/lttng-traces/allo-
20190227-115205/ust/uid/1000/64-bit
or, in per-pid buffering mode:
/home/jgalar/lttng-traces/cloe_fafard-
20190227-115609/ust/pid/hello-ust-20186-
20190227-115700
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 6 May 2019 21:14:02 +0000 (17:14 -0400)]
.gitignore: ignore ccls and clangd files
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 26 Feb 2019 03:11:08 +0000 (22:11 -0500)]
Create an initial trace chunk on first start of an ltt_session
Add a current trace chunk to an ltt_session on session start.
The session base output directory is created at that moment and an
handle to it is provided to the session daemon's trace chunk when
it is set in 'owner' mode.
Setting the trace chunk in "owner" mode results in the creation of
the chunk's output directory. In the case, of the first chunk,
the chunk output directory matches the session output directory.
Hence, this has no visible effect.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sun, 5 May 2019 22:47:26 +0000 (18:47 -0400)]
Allow direct access to the dirfd of a directory handle
Serialization functions need access to the dirfd of a directory handle
in order to pass the fd and re-build the directory handle in
another process.
This function should not be used for other purposes. It is
also only available in the 'enable-dirfd' configuration, which
is assumed for the session daemon.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 21 Feb 2019 16:18:24 +0000 (11:18 -0500)]
Create a consumer daemon trace chunk registry on launch
A trace chunk registry is needed by the consumer daemon since there is
no concrete representation of a tracing session object in those
processes.
The trace chunk registry allows the consumer daemon to match
newly-created streams to their current trace chunk. This integration
is performed in a follow-up commit.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 19 Feb 2019 21:44:36 +0000 (16:44 -0500)]
Add the trace chunk and trace chunk registry interfaces
A trace chunk is a set of stream files. It maps to the user-visible
concept of "trace archive chunks" produced following a tracing session
rotation.
The concept of a "chunk" is introduced to make it possible to
associate a group of stream files together, store common
properties (e.g. the epoch, base path, list of files, credentials,
etc.), and perform an action once all files have been closed/released.
The "trace chunk" interface is to be used by the session, consumer,
and relay daemons in different ways, through the OWNER or USER roles.
The lttng_dynamic_pointer_array, lttng_dynamic_array, and optional
utils are added since they are used by the trace chunk implementation.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 17 May 2019 20:03:25 +0000 (16:03 -0400)]
Add a time_to_iso8601_str() utility
time_to_iso8601_str() formats a Unix timestamp to an ISO 8601
compatible string. For now, the only ISO 8601 time format used
in LTTng-tools is the "YYYYmmddTHHMMSS+HHMM" form used by the
session rotation feature.
Since ISO 8601 allows many formats, this functions may need to be
renamed at some point.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 1 May 2019 21:34:34 +0000 (17:34 -0400)]
Add a method to create a directory handle relative to another one
Adds both implementations (dirfd present or not) of the
lttng_directory_handle_init_from_handle method of the
lttng_directory_handle.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 21 May 2019 23:44:30 +0000 (19:44 -0400)]
Maintain a channel-per-session_id hash table in the consumers
Add and maintain an hash table indexed by `session_id` that allows the
consumer daemons to perform actions on a set of channels belonging to
a given session.
This hash table holds no ownership of the channels; it is merely an
"index" to the existing set of channels maintained in the channel_ht.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 1 May 2019 18:35:34 +0000 (14:35 -0400)]
Document lttng_directory_handle_init_from_dirfd
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 1 May 2019 14:50:55 +0000 (10:50 -0400)]
Allow lttng_directory_handle to be moved
Implement a "move" method (akin to C++'s move semantics) which allows
the transfer of the contents/ownership of a stack-allocated directory
handle to another object.
A moved directory handle must no longer be used after the "move"
operation. The original moved handle is invalidated in the hope
of catching erroneous uses of this API quickly.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 24 Apr 2019 19:33:45 +0000 (15:33 -0400)]
Add a copy method to lttng_directory_handle
This method allows the duplication of a directory handle to
existing storage. It duplicates the underlying directory file
descriptor or path, depending on the configured implementation.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 16 May 2019 16:18:10 +0000 (12:18 -0400)]
Remove unused bitfield.h header
There are no users of the bitfield.h header. It was previously
used to list syscalls from a kernel channel in
834978fd, but
this function was removed in
9897fbc9.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 18 Jul 2019 19:51:49 +0000 (15:51 -0400)]
Tests build fix: undefined MAGIC_VALUE macro
The MAGIC_VALUE macro is only defined when the build is configured
in epoll mode. This value should be defined unconditionally as it
is used in both poll and epoll test modes.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 18 Jul 2019 19:49:28 +0000 (15:49 -0400)]
Build fix: undeclared variable in poll compat
No `ipfd` variable exists in the compat_poll_wait function. The
author meant to use `idle_pfd`.
Reported-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Geneviève Bastien [Mon, 17 Jun 2019 16:56:21 +0000 (12:56 -0400)]
Fix: error when listing sessions with no session
lttng_list_sessions() returns a "fatal error" code when
lttng_ctl_ask_sessiond() returns 0. This was interpreted as the
control socket being shutdown unexpectedly. However, it is
(more often) caused by no sessions being available. Given that, it
makes more sense to report that no sessions are available.
More clean-up/refactoring would be needed to report unexpected socket
shutdowns.
Fixes #1188
Signed-off-by: Geneviève Bastien <gbastien@versatic.net>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Thu, 23 May 2019 18:11:35 +0000 (14:11 -0400)]
Update base test for binding
This test is not run for now as it is not part of the test suite.
Use a temporary directory to store trace.
Split in 2 test suite, one for ust and the other for kernel.
Partially fix formatting.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Thu, 23 May 2019 18:02:26 +0000 (14:02 -0400)]
Fix: python binding: expose domain buffer type
On enable_channel the domain buffer type is used to create a temporary
channel. This currently fail for kernel channel since the buffer type is
not exposed at the binding level and default to LTTNG_BUFFER_PER_PID.
Channel for the kernel domain can only be created in LTTNG_BUFFER_GLOBAL
mode.
Exposing the buffer type also allow userpsace channel to use the per uid
buffering scheme.
The current bindings are in a rough state. This is to at least get them
to work with kernel domain.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:34 +0000 (18:23 -0400)]
Clean-up: correct typo from epoll to poll
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:33 +0000 (18:23 -0400)]
Clean code base from redundant verification
Remove redundant verification for file descriptors with 0 revents in
code base.
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:32 +0000 (18:23 -0400)]
Change lttng_poll_wait behaviour of compat-poll to match compat-epoll
This removes the need to verify for idle file descriptors and mitigates
risks of bug due to behaviour mismatch.
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:31 +0000 (18:23 -0400)]
Fix: hang in thread_rotation when using compat-poll
Add missing verification to prevent a blocking read on an empty fd.
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:30 +0000 (18:23 -0400)]
Adapt poll layer behaviour to match the epoll layer
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:29 +0000 (18:23 -0400)]
Change LTTNG_POLL_GETNB behaviour for poll flavor
Modify LTTNG_POLL_GETNB to provide compatibility with the epoll flavor.
Since it is only used after a lttng_poll_wait call with no modification
(add, del, mod) between, this change does not modify the behaviour in
its current usage while ensuring similar API behavior between
compatibility layer implementations.
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:28 +0000 (18:23 -0400)]
Add Unit test to poll compatibility layer
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 25 Apr 2019 22:23:27 +0000 (18:23 -0400)]
Fix: lttng_poll_mod calls compat_(e)poll_add
lttng_poll_mod should call compat_(e)poll_mod.
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Tue, 30 Apr 2019 23:09:24 +0000 (19:09 -0400)]
Fix: getenv can return null
On system with LANG not defined getenv will return null.
An example of such system is the lava runner used by ci.lttng.org.
https://ci.lttng.org/view/System%20Tests/
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 30 Apr 2019 18:59:51 +0000 (14:59 -0400)]
Bump LTTNG_UST_ABI to 8.0
See commit
6c737d05 in lttng-ust for the rationale behind this
change.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sun, 28 Apr 2019 22:06:11 +0000 (18:06 -0400)]
Fix: directory handle credentials parameter is not const
There is no reason for the "as user" operations on a directory
handle not to take the credentials as a const parameter. Not
passing credentials as const makes their ownership ambiguous
and makes it harder to write const-correct code.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Geneviève Bastien [Thu, 4 Apr 2019 19:18:07 +0000 (15:18 -0400)]
doc: Add reference to USDT probes
SDT probes are known by most as USDT probes. People may be looking for them
by that name.
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Signed-off-by: Geneviève Bastien <gbastien+lttng@versatic.net>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Yannick Lamarre [Thu, 28 Mar 2019 19:07:43 +0000 (15:07 -0400)]
Clean-up: Remove double buffer initialisation
Signed-off-by: Yannick Lamarre <ylamarre@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Mon, 1 Apr 2019 20:33:41 +0000 (16:33 -0400)]
Fix: getgrnam is not MT-Safe, use getgrnam_r
Running the test suite under a Yocto musl build resulted in musl
coredump due to double freeing.
We get the following backtraces:
0 a_crash () at ./arch/x86_64/atomic_arch.h:108
1 unmap_chunk (self=<optimized out>) at src/malloc/malloc.c:515
2 free (p=<optimized out>) at src/malloc/malloc.c:526
3 0x00007f46d9dc3849 in __getgrent_a (f=f@entry=0x7f46d9d1f7e0, gr=gr@entry=0x7f46d9e24460 <gr>, line=line@entry=0x7f46d9e26058 <line>, size=size@entry=0x7f46d92db550, mem=mem@entry=0x7f46d9e26050 <mem>, nmem=nmem@entry=0x7f46d92db558, res=0x7f46d92db548) at src/passwd/getgrent_a.c:45
4 0x00007f46d9dc2e6b in __getgr_a (name=0x487242 "tracing", gid=gid@entry=0, gr=gr@entry=0x7f46d9e24460 <gr>, buf=buf@entry=0x7f46d9e26058 <line>, size=size@entry=0x7f46d92db550, mem=mem@entry=0x7f46d9e26050 <mem>, nmem=0x7f46d92db558, res=0x7f46d92db548) at src/passwd/getgr_a.c:30
5 0x00007f46d9dc3733 in getgrnam (name=<optimized out>) at src/passwd/getgrent.c:37
6 0x0000000000460b29 in utils_get_group_id (name=<optimized out>) at ../../../lttng-tools-2.10.6/src/common/utils.c:1241
7 0x000000000044ee69 in thread_manage_health (data=<optimized out>) at ../../../../lttng-tools-2.10.6/src/bin/lttng-sessiond/main.c:4115
8 0x00007f46d9de1541 in start (p=<optimized out>) at src/thread/pthread_create.c:195
9 0x00007f46d9dee661 in __clone () at src/thread/x86_64/clone.s:22
From another run:
0 a_crash () at ./arch/x86_64/atomic_arch.h:108
1 unmap_chunk (self=<optimized out>) at src/malloc/malloc.c:515
2 free (p=<optimized out>) at src/malloc/malloc.c:526
3 0x00007f5abc210849 in __getgrent_a (f=f@entry=0x7f5abc2733e0, gr=gr@entry=0x7f5abc271460 <gr>, line=line@entry=0x7f5abc273058 <line>, size=size@entry=0x7f5abaef5510, mem=mem@entry=0x7f5abc273050 <mem>, nmem=nmem@entry=0x7f5abaef5518, res=0x7f5abaef5508) at src/passwd/getgrent_a.c:45
4 0x00007f5abc20fe6b in __getgr_a (name=0x487242 "tracing", gid=gid@entry=0, gr=gr@entry=0x7f5abc271460 <gr>, buf=buf@entry=0x7f5abc273058 <line>, size=size@entry=0x7f5abaef5510, mem=mem@entry=0x7f5abc273050 <mem>, nmem=0x7f5abaef5518, res=0x7f5abaef5508) at src/passwd/getgr_a.c:30
5 0x00007f5abc210733 in getgrnam (name=<optimized out>) at src/passwd/getgrent.c:37
6 0x0000000000460b29 in utils_get_group_id (name=<optimized out>) at ../../../lttng-tools-2.10.6/src/common/utils.c:1241
7 0x000000000042dee4 in notification_channel_socket_create () at ../../../../lttng-tools-2.10.6/src/bin/lttng-sessiond/notification-thread.c:238
8 init_thread_state (state=0x7f5abaef5560, handle=0x7f5abbf9be40) at ../../../../lttng-tools-2.10.6/src/bin/lttng-sessiond/notification-thread.c:375
9 thread_notification (data=0x7f5abbf9be40) at ../../../../lttng-tools-2.10.6/src/bin/lttng-sessiond/notification-thread.c:495
10 0x00007f5abc22e541 in start (p=<optimized out>) at src/thread/pthread_create.c:195
11 0x00007f5abc23b661 in __clone () at src/thread/x86_64/clone.s:22
The problem was easily reproducible (~6 crash on ~300 runs). A prototype fix
using mutex around the getgrnam yielded no crash in over 1000 runs. This
patch yielded the same results as the prototype fix.
Unfortunately we cannot rely on a mutex in liblttng-ctl since we cannot
enforce the locking for the application using the lib.
Use getgrnam_r instead.
The previous implementation of utils_get_group_id returned the gid of
the root group (0) on error/not found. lttng_check_tracing_group needs
to know if an error/not found occured, returning the root group is not
enough. We now return the gid via the passed parameter. The caller is
responsible for either defaulting to the root group or propagating the
error.
We also do not want to warn when used in liblttng-ctl context. We might
want to move the warning elsewhere in the future. For now, pass a bool
if we need to warn or not.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 2 Apr 2019 17:41:17 +0000 (13:41 -0400)]
Fix: logging: log_add_time() save/restore errno
The debugging logging macros (e.g. DBG()) are used as printf in the
lttng-tools source files. The printf() implementation does not alter the
errno value, so the fact that log_add_time() (through clock_gettime())
can alter errno is unexpected. For instance, adding a logging statement
for debugging purposes within a function for which errno is expected to
stay unchanged on return will change the behavior between execution with
-vvv and non-verbose.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Wed, 24 Apr 2019 22:56:05 +0000 (18:56 -0400)]
Fix relayd: initialize beacon to -1ULL
The relayd stream beacon_ts_end field is expected to have the value
-1ULL when unset (no beacon has been received since last index).
However, the initial state is wrong. It is left at the value 0, which
indicates that a live beacon has indeed been received (which is untrue),
which in turn causes a live beacon with ctf_stream_id of -1ULL to be
sent to babeltrace, which does not expect it, and fails.
This issue can be triggered with the following scenario:
1) create live session
2) setup UST per-uid buffers tracing
3) start tracing, without any active traced application
4) hook with babeltrace live client to view the trace
5) run a traced application
Step 5) will cause the babeltrace live client to receive a stream_id of
-1ULL, and error out.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Wed, 3 Apr 2019 20:26:45 +0000 (16:26 -0400)]
Fix: relayd: handling of lttng_read errors >= 0
errno is only set when lttng_read returns a negative value. Else, we
need to print a ERR() statement rather than use PERROR().
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
This page took 0.049003 seconds and 5 git commands to generate.