Jérémie Galarneau [Wed, 11 Sep 2019 00:02:32 +0000 (20:02 -0400)]
Clean-up: remove dead assignment from ht cleanup thread launcher
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 11 Sep 2019 00:01:32 +0000 (20:01 -0400)]
Clean-up: remove dead assignment from thread_rotation
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:59:20 +0000 (19:59 -0400)]
Clean-up: remove dead assignment in setup_channel_trace_path
setup_channel_trace_path() returns NULL on error. There is no
need to update a return code.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:56:18 +0000 (19:56 -0400)]
Clean-up: remove dead assignments while setting session trace chunk
The error path always returns -1 making more specific error reporting
impossible. Don't bother reporting more precise error codes.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:53:20 +0000 (19:53 -0400)]
Fix: report path truncation on addition of local uri to consumer
Return an error whenever a session's destinatio path exceeds the
maximal allowed length (LTTNG_PATH_MAX).
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:47:28 +0000 (19:47 -0400)]
Clean-up: lttng: silence warning in regenerate command handler
argv[0] is used before checking argc, resulting in a warning
that argv could be of zero-length.
This is not a reachable bug; it really is a false positive.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:43:16 +0000 (19:43 -0400)]
Clean-up: lttng: silence warning in metadata command handler
argv[0] is used before checking argc, resulting in a warning
that argv could be of zero-length.
This is not a reachable bug; it really is a false positive.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:35:32 +0000 (19:35 -0400)]
Clean-up: remove NULL check on sesison_create mandatory arguments
session_name, base_path, and hostname are assumed to be non-null in
the session_create() function. Checking the pointers for null is
therefore useless.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 23:30:33 +0000 (19:30 -0400)]
Fix: session may be NULL in relay_create_session error path
A reply is sent in response to a session's creation even if it failed
to be created. In those cases, session is null and its output_path
should not be accessed.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 17:44:26 +0000 (13:44 -0400)]
Clean-up: silence erroneous leak warning
Add an assert to validate that a newly created uevent is not enabled.
If it was enabled at creation time, it would be leaked before it is
published.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 17:18:32 +0000 (13:18 -0400)]
Fix: leak of lttng_elf_shdr in lttng-elf.c
Most lttng_elf_shdr uses in lttng-elf.c end-up leaking the header.
Since those headers instances are always short-lived and relatively
small, these leaks are fixed by not using dynamic allocations and
instead using automatic storage.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 16:44:44 +0000 (12:44 -0400)]
Fix: leaked chunk reference in lttng_consumer_create_trace_chunk
Error paths using the 'end' label leak the reference of the
'created_chunk'. The labels are renamed to error and error_unlock
to follow the convention used in most of the code base.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 16:39:02 +0000 (12:39 -0400)]
Clean-up: remove unused stream file creation and unlink functions
No code should be creating or unlinking stream files without going
through the trace chunk interface. These functions are removed since
there are no more users and no legitimate future users.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 10 Sep 2019 14:27:22 +0000 (10:27 -0400)]
Fix: use the trace chunk to truncate streams on late rotation
A stream's rotation can occur after the reception of data that should
be part of the "next" trace chunk. In those cases, the current stream
file and the next one (belonging to the new trace chunk) need to be
opened. The misplaced data is copied between both files and the
now-old file is closed.
This code was not transitioned to use the trace chunk interface and is
the last user of raw stream file FDs. This patch transitions the
function (rewrites it, really) to use the trace chunk interface.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 16:14:54 +0000 (12:14 -0400)]
Clean-up: format using remaining buffer len rather than total len
end_datetime_suffix is has an extra byte to contain a '-' prefix.
This means that its effective length is ISO8601_STR_LEN and not its
sizeof(). This is not considered a fix as time_to_iso8601_str() does
not write more than ISO8601_STR_LEN (for time being).
The ISO8601_STR_LEN macro is used for the buffer lengths for clarity;
no behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 16:08:21 +0000 (12:08 -0400)]
Fix sessiond: report flush errors on session stop
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 15:22:39 +0000 (11:22 -0400)]
Fix: leak of trace_path on error in ust_app_snapshot_record
trace_path is leaked in some error paths of ust_app_snapshot_record().
Lift trace_path to the function's scope and free it whenever it is not
NULL. This is not clean, but this function should be cleaned-up in a
separate patch.
Moreover, this fixes a use after free in the PER_PID-case as
trace_path is free'd when a channel is not found.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 15:14:47 +0000 (11:14 -0400)]
Fix: uninitialized directory handle finalized on error path
relay_create_trace_chunk() creates a session output directory handle
and assigns it to a trace chunk. Since a handle doesn't have a
dedicated "uninitialized" state, reduce its lifetime and don't
generically finalize it in error paths.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 15:00:30 +0000 (11:00 -0400)]
Fix: release reference to created chunk if it can't be published
The reference to a trace chunk created in
lttng_consumer_create_trace_chunk() should be dropped if it can't be
published. Failing to do so results in a leak.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 14:53:45 +0000 (10:53 -0400)]
Fix: destroy chunk registry on element creation failure
A failure path in trace_chunk_registry_ht_element_create() does not
destroy a newly created chunk registry, resulting in a leak.
The ownership of the registry is transfered to the element being
created as soon as possible. Error paths that release the reference on
the new element will then naturally clean-up the trace chunk registry.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 14:46:21 +0000 (10:46 -0400)]
Fix: return NULL on trace chunk registry failure
A free'd registry object can be returned in the error paths
of lttng_trace_chunk_registry_create(). Right now, this happens
whenever the allocation of the registry's hash table fails.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 14:39:23 +0000 (10:39 -0400)]
Fix: forward fatal error in evaluate_condition_for_client()
A fatal error during the evaluation of a condition on behalf
of a client of a notification channel should be forwarded up
the stack to the notification thread.
Here, 'ret' is silently ignored assuming that a check of "evaluation"
is equivalent, which is not the case.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Mon, 9 Sep 2019 14:22:58 +0000 (10:22 -0400)]
Fix relayd: check for NULL in session_put
The session and relay daemons both define their own "session"
APIs (ltt_session and relay_session) which define a session_put()
function.
Coverity reports that a fair amount of callers now assume that
session_put() assumes that a NULL check is performed (as in the
sessiond).
Since the session daemon's variant checks for NULL, it makes sense to
bring both implementation to parity to fix the problems reported and
make this function less confusing to use. This also allows
simplifications to the error handling paths in the relay daemon
(not included in this patch).
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Geneviève Bastien [Fri, 13 Apr 2018 19:51:28 +0000 (15:51 -0400)]
Fix typo in regenerate statedump test util
Signed-off-by: Geneviève Bastien <gbastien+lttng@versatic.net>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 6 Sep 2019 17:33:18 +0000 (13:33 -0400)]
Fix: time constants already defined on macOS
Include the compatibility time.h header in the common/time.h
to override any platform-defined constants. It seems that the
NSEC_PER_SEC (and other similar definitions) are defined to
nothing on macOS. This was already fixed in the compatibility
header in the past.
This also adds USEC_PER_SEC to the list of undef-initions
on macOS.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 6 Sep 2019 17:09:54 +0000 (13:09 -0400)]
Fix: missing include strings.h for bcopy on Solaris 10/11
bcopy() is defined in strings.h on Solaris 10 and 11.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 6 Sep 2019 16:21:58 +0000 (12:21 -0400)]
Fix: strndup does not exist on Solaris 10
strndup does not exist on Solaris 10. Use the lttng_strndup
compatibility wrapper.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 6 Sep 2019 16:21:14 +0000 (12:21 -0400)]
Fix: strnlen is not defined on Solaris 10
strnlen does not exist on Solaris 10. Use the lttng_strnlen
compatibility wrapper.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 6 Sep 2019 15:56:02 +0000 (11:56 -0400)]
Fix: redefinition of USEC_PER_SEC macro on macOS
USEC_PER_SEC is defined by system headers on macOS, resulting in
a warning that it is redefined in time.h. Define it conditionally.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 6 Sep 2019 15:46:23 +0000 (11:46 -0400)]
Fix: sessiond does not build without lttng-ust support
A syntax error in the stub of ust_registry_session_init()
(extra semi-colon) prevents the build of the lttng-sessiond
from succeeding when configured with the --without-lttng-ust
option.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 5 Sep 2019 19:18:40 +0000 (15:18 -0400)]
Fix: sessiond: handle NULL control output in session descriptor base path getter
Creating a session with "lttng create --live" without specifying any
URL triggers a SEGFAULT (dereferencing a NULL pointer) because the
output is not set when getting the session descriptor base path.
Indeed, the destination output URL will only be set later in
cmd_create_session_from_descriptor(), when setting the default output.
When the default output is used, no base path override is possible,
therefore it is fine to assign the base_path to NULL in the base path
getter.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Thu, 5 Sep 2019 23:24:04 +0000 (19:24 -0400)]
Tests: metadata env scope values
The tests extract information from the metadata env scope of the trace
and reconstruct an lttng directory hierarchy to validate that the
information found in the metadata is correct.
Testing this way simulate a viewer that must reconstruct a lttng directory
hierarchy from metadata information only. This information is mostly there
for this purpose.
While there we validate the value of other env field when possible.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Thu, 5 Sep 2019 22:23:54 +0000 (18:23 -0400)]
Support LTTNG_KERNEL_SESSION_SET_CREATION_DATETIME of lttng-modules
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Thu, 5 Sep 2019 22:23:53 +0000 (18:23 -0400)]
Support LTTNG_KERNEL_SESSION_SET_NAME of lttng-modules
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Thu, 5 Sep 2019 22:23:52 +0000 (18:23 -0400)]
Metadata: add env fields to ease lttng path hierarchy creation for viewer
Add the following fields in the env section of the metadata:
- trace_name
The session name without datetime information.
Hence when the session is an auto-generated one, only print
LTTNG_DEFAULT_NAME.
- trace_creation_datetime:
The datetime at which the session was created.
We use session->creation time for it.
- tracer_buffering_scheme
The buffering scheme used. The value can be uid or pid.
- tracer_buffering_id
The key used by the buffering scheme (uid/pid).
- architecture_bit_width
The bit width of the computer architecture (e.g 32 or 64)
- vpid_datetime
The registration time of the vpid for per-pid mode.
Adding these fields ensure that the trace itself carry information that
is normally carried via folder hierarchy. e.g
test-
20190417-174951/ <- trace_name, trace_creation_datetime
└── ust <- domain
└── uid <- tracer_buffering_scheme
└── 1000 <- tracer_buffering_id
└── 64-bit <- architecture_bit_width
├── channel0_0
├── index
│ ├── channel0_0.idx
└── metadata
Per-pid buffering is quite similar.
auto-
20190722-174816 <- trace_name, trace_creation_datetime
└── ust <- domain
└── pid <- tracer_buffering_scheme
└── sample-ust-7640-
20190722-174818 <- procname, tracer_buffering_id, vpid_datetime
├── my-channel_0
├── index
│ ├── my-channel_0.idx
├── metadata
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Wed, 4 Sep 2019 22:30:00 +0000 (18:30 -0400)]
Fix: lttng_directory_handle_init fails on opening base relayd output
lttng_directory_handle_init, called from session_set_anonymous_chunk,
fails to open "$LTTNG_HOME/lttng-traces" directory as the folder is
not yet created at this time.
Rename init_session_output_directory_handle to
session_init_output_directory_handle and move it to session.h
For an anonymous chunk, the session->output_path is empty. The
resulting output directory handle is at the "root" node of the
lttng-relayd chosen output.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Wed, 4 Sep 2019 22:29:59 +0000 (18:29 -0400)]
Fix: init_session_output_path is valid for peer >= 2.11 only
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Wed, 4 Sep 2019 22:29:58 +0000 (18:29 -0400)]
Fix: chunk state is not set when relayd does not support trace chunks
Being explicit here enforces that for a lttng-relayd that does not
support chunks, the queried chunk will never exist.
This caused a lttng destroy command to hang during backward
compatibility testing with older lttng-relayd.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Wed, 28 Aug 2019 20:36:03 +0000 (16:36 -0400)]
Fix: check validity of a stream before invoking ust flush command
At the time ustctl_flush_buffer is called the ustream object might have
already been freed on lttng-ust side.
This can happen following a lttng_consumer_cleanup_relayd and concurrent
consumer flush command (lttng stop).
The chain of events goes as follows.
An error on communication with lttng-relayd occurs.
lttng_consumer_cleanup_relayd flags the streams for deletion
(CONSUMER_ENDPOINT_INACTIVE). validate_endpoint_status_data_stream calls
consumer_del_stream.
At the same time the hash table of streams is iterated over in the
flush_channel function following a stop command. The loop is iterating on
a given stream. The current thread is unscheduled before taking the stream
lock.
In the initial thread, the same stream is the current iteration of
cds_lfht_for_each_entry in validate_endpoint_status_data_stream.
consumer_del_stream is called on it. The stream lock is acquired, and
destroy_close_stream is called. lttng_ustconsumer_del_stream is eventually
called and at this point the ustream is freed.
Going back to the iteration in flush_channel. The current stream is still
valid from the point of view of the iteration, ustctl_flush_buffer is then
called on a freed ustream object.
This can lead to unknown behaviour since there is no validation on the
lttng-ust side. The underlying memory of the ustream object is garbage at
this point.
To prevent such scenario, we check for the presence of the node in the
hash table via cds_lfht_is_node_deleted while holding the stream lock.
This is valid because the stream destruction removes the node from
the hash table and frees the ustream object with the stream lock held.
This duplicate similar "validation" check of the stream object. [1][2]
[1] src/common/consumer/consumer.c:consumer_close_channel_streams
[2] src/common/ust-consumer/ust-consumer.c:close_metadata
This issue can be reproduced by the following scenario:
Modify flush_channel to sleep (i.e 10s) before acquiring the lock on
a stream.
Modify lttng-ust ustctl_destroy_stream to set the
ring_buffer_clock_read callback to NULL.
Note: An assert on !cds_lfht_is_node_deleted in flush channel
after acquiring the lock can provide the same information. We are
modifying the callback to simulate the original backtrace from our
customer.
lttng-relayd
lttng-sessiond
lttng create --live
lttng enable-event -u -a
lttng start
Start some applications to generate data.
lttng stop
The stop command force a flush of the channel/streams.
pkill -9 lttng-relayd
Expect assert or segfault
The original customer backtrace:
0 lib_ring_buffer_try_switch_slow (handle=<optimized out>, tsc=<synthetic pointer>, offsets=0x3fffa9b76c80, chan=0x3fff98006e90, buf=<optimized out>,
mode=<optimized out>) at /usr/src/debug/lttng-ust/2.9.1/git/libringbuffer/ring_buffer_frontend.c:1834
1 lib_ring_buffer_switch_slow (buf=0x3fff98016b40, mode=<optimized out>, handle=0x3fff98017670)
at /usr/src/debug/lttng-ust/2.9.1/git/libringbuffer/ring_buffer_frontend.c:1952
2 0x00003fffac680940 in ustctl_flush_buffer (stream=<optimized out>, producer_active=<optimized out>)
at /usr/src/debug/lttng-ust/2.9.1/git/liblttng-ust-ctl/ustctl.c:1568
3 0x0000000010031bc8 in flush_channel (chan_key=<optimized out>) at ust-consumer.c:772
4 lttng_ustconsumer_recv_cmd (ctx=<optimized out>, sock=<optimized out>, consumer_sockpoll=<optimized out>) at ust-consumer.c:1651
5 0x000000001000de50 in lttng_consumer_recv_cmd (ctx=<optimized out>, sock=<optimized out>, consumer_sockpoll=<optimized out>) at consumer.c:2011
6 0x0000000010014208 in consumer_thread_sessiond_poll (data=0x10079430) at consumer.c:3192
7 0x00003fffac608b30 in start_thread (arg=0x3fffa9b7bdb0) at pthread_create.c:462
8 0x00003fffac530d0c in .__clone () at ../sysdeps/unix/sysv/linux/powerpc/powerpc64/clone.S:96
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Michael Jeanson [Mon, 3 Jun 2019 19:25:52 +0000 (15:25 -0400)]
Fix: namespace our gettid wrapper
Since glibc 2.30, a gettid wrapper was added that conflicts with our
static declaration. Namespace our wrapper so there is no conflict,
we'll add support for the glibc provided wrapper in a further commit.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 8 Aug 2019 14:49:23 +0000 (10:49 -0400)]
pprint.m4: add missing copyright header
Add the missing copyright header to pprint.m4, clarifying that it is
GPLv2+ with the special exception applying to Autoconf Macros.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Philippe Proulx <pproulx@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 23 May 2019 19:11:50 +0000 (15:11 -0400)]
Fix: reintroduce lazy kernel modules load, fix empty syscall list
Up and including to lttng 2.10, if the lttng-modules are not available
when the session daemon starts, an attempt is made to load them each
time a command attempts to interact with the kernel domain.
2.11-rc introduces a change in behavior which removes this lazy
initialization. This is an issue for distribution packaging (e.g.
Ubuntu) where there is no strong dependency between lttng-tools and
lttng-modules (because either are optional). So we can be unlucky and
find ourselves in a situation where the modules are not available when
the session daemon is launched at installation, but only afterwards.
Re-introduce the lazy kernel module load behavior, since this is
expected by users.
Also, fix an issue with empty syscall list in those lazy initialization
scenario by moving invocation of syscall_init_table() from main() to
init_kernel_tracer().
While we are there, cleanup the following in session daemon:
- move kernel_tracer_fd from globals.c to kernel.c. It becomes a static
variable,
- move module_proc_lttng from main.c to kernel.c,
- move init_kernel_tracer() from main.c to kernel.c,
- introduce kernel.c:cleanup_kernel_tracer(), invoke it from program
cleanup,
- introduce kernel_tracer_is_initialized() to check the state of
kernel tracer initialization,
- adapt kernel.c functions to use the static kernel_tracer_fd rather
than expect it as parameter,
- update syscall_init_table, invoked from kernel.c, to pass the
kernel_tracer_fd as parameter.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Wed, 22 May 2019 20:49:01 +0000 (16:49 -0400)]
Fix: check for lttng modules presence before testing
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:08:52 +0000 (15:08 -0400)]
sessiond: use epoll()/poll() instead of select()
The select(2) system call is an ancient ABI limited to processes
containing at most FD_SETSIZE file descriptors overall (typically
1024).
This select call will fail if the target file descriptor is above
FD_SETSIZE in a session daemon containing many file descriptors.
This is unlikely to happen in normal use given than
sessiond_init_thread_quit_pipe() is called early by main(). Odd
scenarios could trigger this, for instance if the parent process leaves
a large number of file descriptors open, or if a library which
allocates file descriptors is LD_PRELOADed with the sessiond.
Never use select, use the lttng epoll/poll wrapper instead.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:08:51 +0000 (15:08 -0400)]
lttng-ctl: notifications: use epoll()/poll() instead of select()
The select(2) system call is an ancient ABI limited to processes
containing at most FD_SETSIZE file descriptors overall (typically
1024).
Those notification APIs will fail if the target file descriptor
is above FD_SETSIZE in a process containing many file descriptors.
Never use select, use the lttng epoll/poll wrapper instead.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:08:50 +0000 (15:08 -0400)]
epoll/poll compat: expose interruptible API
Some use of the epoll/poll wrapper require interruption
by signals to make the poll call return -1, errno EINTR.
Expose a new lttng_poll_wait_interruptible API for this
purpose.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:08:01 +0000 (15:08 -0400)]
tests: invoke full_cleanup from script trap handlers, use modprobe -r
Scripts implementing their own trap handlers override the generic
one provided by utils.sh (full_cleanup). Invoke it at the end of
the handlers to provide the utils cleanup as well.
Moreover, change use of "rmmod" to "modprobe -r", which is better
in trap handlers because it does not print errors if the module
was not loaded yet when the signal occurs.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:08:00 +0000 (15:08 -0400)]
Cleanup: test: don't stop relayd twice
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:07:59 +0000 (15:07 -0400)]
Fix: test: utils.sh: exit from process on full_cleanup
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:07:58 +0000 (15:07 -0400)]
Fix: utils.sh: handle SIGPIPE
perl prove closes its child pipes before giving it a chance to execute
the signal trap handler. This means the child will not be able to
complete execution of the trap handler if that handler writes to stdout
or stderr.
Work-around this situation by redirecting stdin, stdout, and stderr
to /dev/null if a SIGPIPE is caught.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:07:57 +0000 (15:07 -0400)]
Fix: tests: error handling in high throughput limits test (v2)
Each individual call to "tc" should be checked for error, else we
may fail to catch specific tc errors caused, for instance, by a
kernel configuration that only contains some of the required
class modules.
Also, invoke the utils.sh full_cleanup function from the script-specific
interrupt_cleanup trap handler rather than try to perform stopping
of relayd and sessiond within the script.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 16 May 2019 19:07:56 +0000 (15:07 -0400)]
Improve handling of test SIGTERM/SIGINT (v2)
The current state of signal handling for test scripts is: on
SIGTERM/SIGINT of the tests (e.g. a CTRL-C on the console), session
daemon and relay daemon are killed with SIGKILL, thus leaking all their
resources, and leaving lttng kernel modules loaded.
Revamp the "stop" functions to take a signal number and a timeout
as optional parameters. The default signal number is SIGTERM.
The full_cleanup trap handler now tries to nicely kill relayd and
sessiond (if they are present) with SIGTERM, and wait up to the
user-configurable LTTNG_TEST_TEARDOWN_TIMEOUT environment variable
(which has a default of 60s). Then, if there are still either relayd,
sessiond, or consumerd present, it will SIGKILL them and wait for
them to vanish. If it had to kill sessiond with SIGKILL, it will
also explicitly try to unload the lttng modules with modprobe.
This approach is inspired from sysv init script shutdown behavior.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 5 Sep 2019 12:47:40 +0000 (08:47 -0400)]
Fix: incorrect conversion specifier used with size and padding
packet>size and padding_len, of type size_t, are logged using a
%PRIu64 specifier resulting in a warning being emitted on 32-bit
architectures.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 5 Sep 2019 12:42:40 +0000 (08:42 -0400)]
Fix: incorrect conversion specifier used with packet size
packet_size, of type size_t, is logged using a %PRIu64 specifier
resulting in a warning being emitted on 32-bit architectures.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 5 Sep 2019 00:22:20 +0000 (20:22 -0400)]
Fix: mismatch of ust_app_get_size_one_more_packet_per_stream prototype
The stub of ust_app_get_size_one_more_packet_per_stream() used when
building without lttng-ust support does not match the actual prototype
defined in ust-app.h. The ltt_ust_session pointer parameter must be
marked as const.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Thu, 5 Sep 2019 00:20:00 +0000 (20:20 -0400)]
Fix: mismatch of ust_app_snapshot_record prototype
The stub of ust_app_snapshot_record() used when building without
lttng-ust support does not match the actual prototype defined
in ust-app.h. The consumer_output pointer parameter must be
marked as const.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 4 Sep 2019 23:54:16 +0000 (19:54 -0400)]
Fix: lttcomm_consumer_close_trace_chunk_reply undefined without UST
lttcomm_consumer_close_trace_chunk_reply is not defined when
lttng-tools is build without lttng-ust support. Move its declaration
as part of the common section of sessiond-comm.h.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Sat, 24 Aug 2019 23:02:18 +0000 (16:02 -0700)]
Fix: kernel/ust snapshot backward compat for 2.10 relay
Ensure the snapshots are saved in the right subdirectory on older relay
daemons (2.10 and before).
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Fri, 23 Aug 2019 00:59:49 +0000 (17:59 -0700)]
relayd protocol: reply path for close chunk and create session 2.11
Since the relay daemon is expected to be able to move the target
destination of the trace, reply the chunk and session path so session
can have relevant data rather than guessing their location.
The session daemon now use this information to send the path back to
the client for rotation and destroy commands, as well as for
rotation completion notifications.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Sun, 18 Aug 2019 18:53:33 +0000 (14:53 -0400)]
Fix: streaming and snapshot backward compat for relayd < 2.11
Fix backward compatibility of session daemon 2.11 with relayd versions
prior to 2.11.
Session daemon and consumer daemon 2.11 use the "chunk" object to
represent the current output directory. However, both sessiond and
consumerd need to ensure they do not send chunk and rotation commands
to a relayd older than 2.11, and tweak the stream paths accordingly.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 22 Aug 2019 17:06:22 +0000 (10:06 -0700)]
Fix: relayd: create_index_file error handling
Fix mixup between status and 0/-1 return value;
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 3 Sep 2019 22:20:01 +0000 (18:20 -0400)]
Fix: don't use newlines in logging message
Using explicit newlines ('\n') in a logging messages breaks the
logging output and makes it unnecessarily harder to read and parse the
LTTng daemons' logs.
Moreover, it is unnecessary to escape multi-line string literals.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 22 Aug 2019 00:36:29 +0000 (17:36 -0700)]
Fix: relayd outputs traces of legacy sessionds to home dir
The relay daemon's session output path creation fallback, which is
used when interacting with legacy (< 2.11) session daemons, does not
append the DEFAULT_TRACE_DIR_NAME to the user's home directory
(or LTTNG_HOME, if specified).
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Sun, 18 Aug 2019 19:12:02 +0000 (15:12 -0400)]
relayd comm: add base path to create session
Add base path and session_name_contains_creation_time flag to
session creation 2.11 message.
Use it to ensure the "auto-<ts>" session names don't get
duplicate timestamps for their output paths.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 15 Aug 2019 19:34:48 +0000 (15:34 -0400)]
tests: add base-path tests
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Wed, 21 Aug 2019 23:24:39 +0000 (16:24 -0700)]
test utils: support URI in lttng_snapshot_add_output
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Thu, 15 Aug 2019 19:36:08 +0000 (15:36 -0400)]
Fix: close consumer sockets before waiting for them
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Wed, 14 Aug 2019 20:32:58 +0000 (16:32 -0400)]
Fix: consumer: put each chunk on teardown
Trace chunks in the registry may still exist because those are top-level
entities for which we may have received a "create" command from session
daemon without pairing "destroy". Iterate on the entire chunk registry
and put the sessiond reference for each chunk on consumerd teardown
Fixes the following assert on teardown caused by non-empty registry on
SIGINT:
lttng-consumerd: trace-chunk.c:1109: lttng_trace_chunk_registry_destroy: Assertion `!ret' failed.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 30 Aug 2019 22:10:56 +0000 (18:10 -0400)]
Fix: validate that session, host and basepath are legal
Ensure that session name, hostname and the session's base path
do not contain dots ('.') to safeguard against malformed names
that could be used to walk-up the relay daemon output path
hierarchy.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Wed, 14 Aug 2019 17:30:28 +0000 (13:30 -0400)]
Fix: honor base path for network URIs
Starting from 2.11, the following use-case stopped working as expected:
lttng create --set-url=net://localhost/my/path/custom
lttng enable-event -u -a
lttng start
[ run app ]
The output trace is expected in the following directory on the relayd
(no timestamp suffix):
$LTTNG_HOME/lttng-traces/[hostname]/my/path/custom
Add the base_path information to the session object in the session
daemon, extracted from the network URI, and pass that information to the
relay daemon through the create session (2.11) command.
It also fixes the use-case:
lttng create --snapshot --set-url=net://localhost/my/path/custom
lttng enable-event -u -a
lttng start
[ run app ]
lttng snapshot record
which is expected to record the snapshot under:
$LTTNG_HOME/lttng-traces/[hostname]/my/path/custom/snapshot-[N]-[timestamp]
Similar situation for:
lttng create --snapshot
lttng enable-event -u -a
lttng start
[ run app ]
lttng snapshot record net://localhost/my/path/custom
which is expected to record the snapshot under:
$LTTNG_HOME/lttng-traces/[hostname]/my/path/custom/snapshot-[N]-[timestamp]
Note that specifying the base_path on the snapshot record command
overrides the base path specified at create.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Tue, 3 Sep 2019 17:01:09 +0000 (13:01 -0400)]
Fix: bindings import segfaults on missing hash_key_u64
A mishandled error in SWIG-generated code causes the Python3
interpreter to segfault when a missing symbol is reported during
the linking (at runtime) against liblttng-ctl.
libcommon makes use of the internal libhashtable.la since the
addition of the lttng_trace_chunk interface. This introduces
a transitive dependency to libhashtable.la in liblttng-ctl.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Wed, 14 Aug 2019 14:52:49 +0000 (10:52 -0400)]
Cleanup: typo: overriden -> overridden
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 13 Aug 2019 20:38:47 +0000 (16:38 -0400)]
Fix: relay: relay_rotate_session_streams uninitialized return value
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 13 Aug 2019 20:27:51 +0000 (16:27 -0400)]
Fix: test_kernel rotation: expect 60 rather than 64
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Tue, 13 Aug 2019 20:25:16 +0000 (16:25 -0400)]
Save registration time for app
Reuse the registration time for path generation.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 13 Aug 2019 19:52:52 +0000 (15:52 -0400)]
Fix: tests: update rotation ust number of tests
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 21:54:00 +0000 (17:54 -0400)]
Clean-up: missing word `lock` in comment
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 21:03:45 +0000 (17:03 -0400)]
Use case-scope error labels handling LTTNG_CONSUMER_ROTATE_CHANNEL
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_ROTATE_CHANNEL command. This is done in order to allow a
follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 21:01:16 +0000 (17:01 -0400)]
Use case-scope error labels handling LTTNG_CONSUMER_PUSH_METADATA
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_PUSH_METADATA command. This is done in order to allow a
follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 20:57:29 +0000 (16:57 -0400)]
Use case-scope error labels when handling LTTNG_CONSUMER_GET_CHANNEL
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_GET_CHANNEL command. This is done in order to allow a
follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended beyond additional calls to
health_code_update() in some code paths which should have no impact.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 20:49:38 +0000 (16:49 -0400)]
Use case-scope error labels handling LTTNG_CONSUMER_ROTATE_CHANNEL
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_ROTATE_CHANNEL command. This is done in order to allow
a follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 20:46:36 +0000 (16:46 -0400)]
Use case-scope error labels handling LTTNG_CONSUMER_DESTROY_CHANNEL
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_DESTROY_CHANNEL command. This is done in order to allow
a follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 20:37:26 +0000 (16:37 -0400)]
Use case-scope error labels handling LTTNG_CONSUMER_STREAMS_SENT
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_STREAMS_SENT command. This is done in order to allow a
follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 27 Aug 2019 20:19:34 +0000 (16:19 -0400)]
Use case-scope error labels when handling LTTNG_CONSUMER_ADD_STREAM
This cleans-up the error handling of the case handling the
LTTNG_CONSUMER_ADD_STREAM command. This is done in order to allow a
follow-up commit to perform some common clean-up operations on error
while introducing as few changes as possible.
No behaviour change is intended beyond additional calls to
health_code_update() in some code paths which should have no impact.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sat, 24 Aug 2019 23:58:48 +0000 (16:58 -0700)]
Clean-up: set stream's channel pointer to NULL after releasing ref
A stream's "chan" pointer to its parent channel remains set after
the reference to the channel has been released. This can lead to
accidental uses after the release of the channel through the
stream object.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sat, 24 Aug 2019 23:08:33 +0000 (16:08 -0700)]
Fix: mark consumer channels as logically deleted during deletion
A consumer channel that is "logically" deleted (but not yet reclaimed
with regards to RCU) can be iterated-on when creating a trace chunk.
Such a deleted channel will have released its reference to its
previous trace chunk, but the trace chunk creation operation will
set its current trace chunk to the new trace chunk, thus acquiring
a new reference to the trace chunk.
The clean-up performed by the RCU thread already assumes that a
logical deletion already took place and will not re-release the
reference to the channel's current trace chunk.
In effect, the reference to the trace chunk is leaked and will
prevent any rotation out of the trace chunk from completing.
Reported-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Sat, 24 Aug 2019 22:14:11 +0000 (15:14 -0700)]
Cleanup: mark utils_get_home_dir as returning a const string
utils_get_home_dir() returns a string obtained from getenv() that
should not be modified nor free'd. However, utils_get_user_home_dir()
returns the path as a copy.
The return parameter of utils_get_home_dir() is marked as const
to underline this difference and prevent mix-ups in the use of
those functions.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jonathan Rajotte [Mon, 12 Aug 2019 21:30:45 +0000 (17:30 -0400)]
Fix: use utils_get_home_dir for anonymous trace chunk path
The default behaviour of lttng_directory_handle_init will set
the current working directory if opt_output_path is null.
Instead we want to honour LTTNG_HOME/HOME.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 13 Aug 2019 19:29:29 +0000 (15:29 -0400)]
Fix: tests: typo in rotation util
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Tue, 13 Aug 2019 19:28:34 +0000 (15:28 -0400)]
Fix: keep ust/kernel session items around for destroy notifier
Split the destruction and release operations for ust and kernel
inner-sessions as they may be accessed by session destruction
notifiers registered against their parent ltt_session.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Fri, 23 Aug 2019 00:35:17 +0000 (17:35 -0700)]
Tests: add a test for the "skip non-empy" rmdir flag
Add tests for the newly introduced "skip non-empty" flag that
was added to the lttng_directory_handle interface's
"remove_subdirectory_recursive" function.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Mon, 12 Aug 2019 21:25:19 +0000 (17:25 -0400)]
Fix: rmdir recursive: skip non-empty directories with flag
Restore the behavior of LTTng 2.10 with a flag.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Wed, 21 Aug 2019 22:24:12 +0000 (15:24 -0700)]
Fix: uninitialized variable returned by relay_rotate_session_streams
`ret` can be returned uninitialized when relay_rotate_session_streams
succeeds. Initialize it to 0 on the successful code path.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Mon, 12 Aug 2019 16:16:18 +0000 (12:16 -0400)]
Fix: allow quiet rotation at destroy of live session
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Mon, 12 Aug 2019 15:47:34 +0000 (11:47 -0400)]
Fix: relayd: live: read index file header
The header of index files opened in "read" mode are not validated
when the file is being opened. This check should be performed to
restore the previous behaviour.
Moreover, the rest of the code using index files in "read" mode
assumes that the file header has already been skipped when reading
index entries.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Mathieu Desnoyers [Mon, 12 Aug 2019 15:46:34 +0000 (11:46 -0400)]
Fix: index: use parenthesis around define
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
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>
This page took 0.046561 seconds and 4 git commands to generate.