Version 2.4.0-rc2

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

Fix: baddr_statedump tracepoint registration

Ensure baddr_statedump tracepoint registration is completed prior to
using the tracepoint in lttng_ust_baddr_statedump().

Make liblttng-ust-dl robust for explicit baddr_statedump tracepoint
deregistration in lttng_ust_cleanup() (prevent dlopen/dlclose to get
traced if ust_baddr tracepoints are not available).

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: baddr deadlocks and RCU races

Fix deadlocks wrt internal libc mutexes.

Fix RCU races by ensuring every UST source file using RCU read-side lock
is marked as _LGPL_SOURCE. The dynamic binding to urcu-bp relies on the
dynamic loader, and it seems not to behave well when used from
constructors.

Use Userspace RCU with this commit to completely fix baddr races:

"Fix: urcu-bp interaction with threads vs constructors/destructors"

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

Fix: lock exit mutex when creating threads

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

Fix: baddr deadlock with lttng-ust destructor

Was hanging lttng-tools tests/regression/ust/overlap/test_overlap

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

baddr: include missing header

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

baddr: get session under lock

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

Fix: take the ust lock around session iteration in statedump

This is crafted _very_ carefully: we need to always take the ust lock
_within_ the dynamic loader lock.

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

Fix: baddr_statedump deadlock with JUL tracing

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: client_packet_header() uses wrong packet

It should use lib_ring_buffer_read_offset_address() to get the packet
being read, rather than lib_ring_buffer_offset_address(), which is only
meant to be used when writing to the packet.

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

Cleanup: fix typo in ring buffer backend comment

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

Compile liblttng-ust-baddr c files into liblttng-ust

Move them into liblttng-ust/, where they are now compiled.

This is to ensure we don't rely on statically linked libs (and
associated constructor problems), and that we don't install a shared
object in the system needlessly.

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

Allow suppressing of base-address-state tracing

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Integrate base-address statedump into lttng-ust

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: eliminate timestamp overlap between packets

By using the timestamp sampled at space reservation when the packet is
being filled as "end timestamp" for a packet, we can ensure there is no
overlap between packet timestamp ranges, so that packet timestamp end <=
following packets timestamp begin.

Overlap between consecutive packets becomes an issue when the end
timestamp of a packet is greater than the end timestamp of a following
packet, IOW a packet completely contains the timestamp range of a
following packet.  This kind of situation does not allow trace viewers
to do binary search within the packet timestamps. This kind of situation
will typically never occur if packets are significantly larger than
event size, but this fix ensures it can never even theoretically happen.

The only case where packets can still theoretically overlap is if they
have equal begin and end timestamps, which is valid.

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

Fix: lttng-gen-tp to build out of tree

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

Fix: compile doc/ after JUL/Java

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

JUL: add Hello.java in doc/examples

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

Fix lttng-ust(3) manpage

Describe that TRACEPOINT_CREATE_PROBES is needed.

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

JUL: fix enable all event for delayed Logger

This adds a HashMap containing the enabled Logger names (and wildcard
'*'). This is used to make sure we don't enable twice the same Logger
for the '*' event name (enable all -a).

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

Fix: typo in JNICALL for the JUL tracepoint

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

Fix: move va_end to outermost label

Coverity:

** CID 1021237:  Missing varargs init or cleanup  (VARARGS)
/snprintf/vfprintf.c: 955 in ust_safe_vfprintf()

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

Fix: add missing NULL check after allocation

Found by coverity:

** CID 1021246:  Dereference null return value  (NULL_RETURNS)
/libringbuffer/shm.c: 80 in shm_object_table_create()

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

Fix cleanup: all spaces before flags

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

fix: lttng-gen-tp: add missing spaces around flags

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

Version 2.4.0-rc1

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

doc/examples: error out when a subdir make fails

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

JUL: fix out of tree build

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

Java/jul: fix make dist

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

JUL: install documentation and jar file

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

Add missing copyrights to test scripts

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

Escape minus signs in lttng-ust-cyg-profile manpage

By default, "-" chars are interpreted as hyphens (U+2010) by groff, not
as minus signs (U+002D). Since options to programs use minus signs
(U+002D), this means for example in UTF-8 locales that you cannot cut
and paste options, nor search for them easily.

Source: http://lintian.debian.org/tags/hyphen-used-as-minus-sign.html

Signed-off-by: Jon Bernard <jbernard@tuxion.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix spelling mistake in lttng-ust manpage

Signed-off-by: Jon Bernard <jbernard@tuxion.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix tests/benchmark

Add copyright notice.

Remove extern C guard with respect to #597.

Signed-off-by: Zifei Tong <soariez@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Implement base-address-state tracing

Dump the base-address state (executable and shared objects) into session
on session-enable (per-session events).

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Base-address tracing for dlopen and dlclose

Provide an LD_PRELOAD library to allow tracing of calls to dlopen and
dlclose.

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add a check against excluders

When matching enablers with events, first check against all
excluders of the enabler. If the event matches with any of the excluders,
then the event does not match with the enabler.

[ Edit by Mathieu Desnoyers: apply coding style changes. ]

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add handler for LTTNG_UST_EXCLUSION in UST ABI

Add message handler for the LTTNG_UST_EXCLUSION command in
UST ABI. Copy the exclusion data into a lttng_ust_excluder_node
structure and pass it to the enabler command handler.

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Store exclusions to enablers

Implement a function that adds exclusions to the list in the enabler.
Call this function in the enabler command handler when the
LTTNG_UST_EXCLUSION command is received.

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add excluders to enabler structure

Define a structure that holds excluders and can be used in lists.
Add a list holding these structures to the enabler structure.
Initialize and destroy the list when the enabler is initialized
and destroyed.

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Send exclusion data through ustcomm

Define a function ustctl_set_exclusion that sends the
exclusion data through ustcomm with the LTTNG_UST_EXCLUSION
command message.

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Define exclusion structure for UST ABI

Define a structure for holding exception data. The last data item
is a flexible array that contains a variable number of
exclusions names.

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Define a new command LTTNG_UST_EXCLUSION

Define a new command id for attaching exclusions to enablers.
Define the structure for passing the command.

Signed-off-by: JP Ikaheimonen <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: package of LTTngUst.h was renamed without renaming target in makefile

Signed-off-by: Michael Jeanson <mjeanson@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: check for java .class in package subfolders

Signed-off-by: Michael Jeanson <mjeanson@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: java library to use user define JDK at compile time

Signed-off-by: David Goulet <dgoulet@efficios.com>

Add liblttng-ust-jul for JUL support

The build system creates a jar file named liblttng-ust-jul.jar to be
linked with the Java application. A public library named
liblttng-ust-jul.so is also created and must be in the library path of
the Java application in order for the LTTngAgent to use it for the
tracer's JNI call.

A unit test is also added and integrated with the "make check" command.

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

Fix: application SIGBUS when starting in parallel with sessiond

There is a race between application startup and sessiond startup, where
there is an intermediate state where applications can SIGBUS if they see
a zero-sized shm, if the shm has been created, but not ftruncated yet.

On the UST side, fix this by ensuring that UST can read the shared
memory file descriptor with a read() system call before they try
accessing it through a memory map (which triggers the SIGBUS if the
access goes beyond the file size).

On the sessiond side, another commit needs to ensure that the shared
memory is writeable by applications as long as its size is 0, which
allow applications to perform ftruncate and extend its size.

Fixes #623

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

Fix: Minor corrections to the lttng-ust man page

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Documentation cleanup: Overhaul of the lttng-gen-tp man page

Fix a couple of typos, mistakes and out-of date informations.

[Mathieu: added extra fix recommended by Daniel Thibault]

Acked-by: Yannick Brosseau <yannick.brosseau@gmail.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Warning cleanup: add missing padding initializer

Reported-by: Marcus Tomlinson <themarcustomlinson@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: socket connect hang on heavy loads

We need to perform both connect and sending registration message before
doing the next connect otherwise we may reach unix socket connect queue
max limits and block on the 2nd connect while the session daemon is
awaiting the first connect registration message.

This happens in scenarios where unix socket connect queues are nearly
full.

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

Cleanup: fix typo

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

Fix: ust-comm recvmsg should handle partial receive

Handles cases where unix socket buffer is full. Without this fix, the
session daemon kicks out application that happen to have their
registration message split into multiple recvmsg on the sessiond side.

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

gcc warning fix: -Wextra

For the "ordered comparison of pointer with integer zero" warning, fix
this by comparing (type) -1 against (type) 0 instead of just 0, so if
"type" is a pointer type, this pointer type will be applied to the right
operand too, thus fixing the warning.

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

Add usage reference count for tracepoints

Keep track of how many libraries use a tracepoint, and disable the
tracepoint when the number of users drops to zero.

A new reference counter is added to tracepoint_entry. This keeps track
of how many callsites use that tracepoint.

When you have libraries and/or executables sharing tracepoints, you
cannot just disable your tracepoints when the library is unregistered.
You must check that the tracepoint is not used by any other libraries
before you disable it.

Function lib_disable_tracepoints becomes unnecessary, and is removed.

Signed-off-by: Ikaheimonen, JP <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

tracepoint.c: Move add_callsite/remove_callsite further down in file

In preparation for calling disable_tracepoint() from those functions.

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

Store the callsites into the library callsite list

Fix the issue where the callsites are registered but never
properly unregistered.

[ Edit by Mathieu Desnoyers: minor coding style fix: remove parenthesis. ]

Signed-off-by: Ikaheimonen, JP <jp_ikaheimonen@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix linking warning

lttng_static_ctx should be extern in this header to avoid getting a
warning when --warn-common is used.

PS: Thanks to Yannick Brosseau for his help.

Signed-off-by: Mohamad Gebai <mohamad.gebai@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Per-stream ioctl to get the current timestamp

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

LTTng ringbuffer ABI calls for index generation

These new calls export the data required for the consumer to
generate the index while tracing :
- timestamp begin
- timestamp end
- events discarded
- context size
- packet size
- stream id

Signed-off-by: Julien Desfossez <jdesfossez@efficios.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Specialize lttng_ust_lib_ring_buffer_client_cb

Prepare the ring-buffer config to have custom callbacks. These custom
callbacks are not related to the ring-buffer operations but allow
applications to add custom functions.
No additional feature or change in behaviour in this patch.

Signed-off-by: Julien Desfossez <jdesfossez@efficios.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add tracing instrumentation for pthread mutex lock functions

Signed-off-by: Stefan Seefeld <stefan_seefeld@mentor.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Version 2.3.0

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

Version 2.3.0-rc2

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

doc/examples/gen-tp: pass automake CPPFLAGS/CFLAGS/LDFLAGS

Fix cross-build.

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

tools/lttng-gen-tp: honor CPPFLAGS and LDFLAGS

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

Fix: doc/examples cross-build

We need to pass automake's CPPFLAGS and LDFLAGS to examples. Add CFLAGS
too for completeness.

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

Fix: liblttng_ust.la should depend on liblttng-ust-tracepoint.la

Changing -llttng-ust-tracepoint to liblttng-ust-tracepoint.la, so it
appears in the build dependencies.

Fixes #325

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

Introduce configure --with-lttng-system-rundir

Allow overriding the path where the system-wide sessiond is expected to
keep its runtime files.

It does _not_ allow multiple instances of system-wide sessiond to run in
parallel though, because the wait shm files still requires having only
one single system-wide sessiond at any given time.

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

examples: tracepoint probes don't need extern C

Fix all UST demo applications under doc/examples/

Fixes #597

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

Tracepoint probes don't need extern C

tracepoint.h has the proper extern C guards, so every liblttng-ust API
called from the probe has the proper linkage.

The rest of the symbols are only meaningful within the probe, so we
don't care if they are mangled or not.

This fixes the two examples, and lttng-gen-tp.

Fixes #597

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

Add trace support for memalign and posix_memalign

Signed-off-by: Stefan Seefeld <stefan_seefeld@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

malloc instrumentation: remove dependency on pthread

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

Add trace support for calloc and realloc.

Signed-off-by: Stefan Seefeld <stefan_seefeld@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: allow make check to run in VPATH build

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

Update .gitignore

Signed-off-by: Zifei Tong <soariez@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update obsolete benchmark test

Update the test to use lttng-ust 2.x API and output result to TAP
format.

[ Update by Mathieu Desnoyers: README is EXTRA_DIST within Makefile.am ]

Signed-off-by: Zifei Tong <soariez@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup tests (2nd commit)

Move TAP related code to tests/utils.
Remove scripts only used by obsolete tests.

[ Mathieu : my previous commit was missing some files. ]

Signed-off-by: Zifei Tong <soariez@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup tests

Move TAP related code to tests/utils.
Remove scripts only used by obsolete tests.

Signed-off-by: Zifei Tong <soariez@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: doc/examples VPATH build

Fixes #577

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

Version 2.3.0-rc1

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

Fix: Unchecked asprintf/vasprintf return values

strp·may·be·modified·even·in·case·of·failure·according·to·the
ASPRINTF(3) man page.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Missing NULL pointer init in tap.c

Could lead to free() of unitialized pointer.

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

Disable doc/examples build

This breaks VPATH build.

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

Add python3 support to lttng-gen-tp

Fixes #585

Tested-by: Christian Babeux <christian.babeux@efficios.com>
Signed-off-by: Zifei Tong <soariez@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Bump ABI major to 5

Bump major version between UST-sessiond-consumerd.

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

callsite: add "ip" context

Add caller's instruction pointer context.

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

Allow environment variable LTTNG_HOME to override HOME

Patch functionality - If LTTNG_HOME environment variable exists, it is
used instead of HOME.  Reason for patch - We are trying to deploy LTTng
on a system where $HOME is on a filesystem mounted read-only, but cannot
afford to run lttng as a different user and cannot move the home
directories of users to writeable locations.

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

Fix: libs order in Makefile

Since commit 16c96fc0224145192f803fecbdb20ec05fdaf5e2, the examples
don't build well on Ubuntu. The gcc -l argument needs to be the last
one.

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

Fix: ring buffer: get_subbuf() checks should be performed on "consumed" parameter

This triggers lots of false-positive -EAGAIN errors in flight recorder
snapshots.

Reported-by: Julien Desfossez <jdesfossez@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Introduce ustctl_write_one_packet_to_channel

This function allows the consumer to write at most one packet of data to
the channel, that way we can push data to the ring buffer without
risking to block on subbuffer switch.

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

build: Fix out-of-tree build

To allow out-of-tree build, we need to include top_buildir in include
directories or the generated config header won't be found.

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: SWITCH_FLUSH new sub-buffer checks

The SWITCH_FLUSH, when performed on a completely empty sub-buffer, was
missing some checks (imported from space reservation).

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

Fix: ring buffer: handle concurrent update in nested buffer wrap around check

With stress-test loads that trigger sub-buffer switch very frequently
(small 4kB sub-buffers, frequent flush) in lttng-modules, we currently
observe this kind of warnings once every few minutes:

[65335.896208] ring buffer relay-overwrite-mmap, cpu 5: records were lost. Caused by:
[65335.896208]   [ 0 buffer full, 1 nest buffer wrap-around, 0 event too big ]

It appears that the check for nested buffer wrap-around does not take
into account that a concurrent execution contexts (either nested for
per-cpu buffers, or from another CPU or nested for global buffers) can
update the commit_count value concurrently.

What we really want to do with this check is to ensure that if we enter
a sub-buffer that had an unbalanced reserve/commit count, assuming there
is no hope that this gets rebalanced promptly, we detect this and drop
the current event. However, in the case where the commit counter has
been concurrently updated by another reserve or a switch, we want to
retry the entire reserve operation.

One way to detect this is to sample the reserve offset twice, around the
commit counter read, along with the appropriate memory barriers.
Therefore, we can detect if the mismatch between reserve and commit
counter is actually caused by a concurrent update, which necessarily has
updated the reserve counter.

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

Cleanup: lib_ring_buffer_switch_new_end() only calls subbuffer_set_data_size()

lib_ring_buffer_switch_new_end() is always called when an event exactly
fills a sub-buffer, which makes padding_size always 0. However, there is
one side-effect that lib_ring_buffer_switch_new_end() needs to have: it
calls subbuffer_set_data_size() to update the size of the data to be
read from the sub-buffer.

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

Fix: doc/examples: gen-tp path

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

Revert "Cleanup: ring buffer: remove lib_ring_buffer_switch_new_end()"

This reverts commit 019fbcb65ec7eec2c86019ad1afce8fa47dc1269.

Breaks regression tests of lttng-tools.

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

Fix: handle writes of length 0

lib_ring_buffer_write() could be passed a length of 0. This typically
has no side-effect as far as writing into the buffers is concerned,
except for one detail: in overwrite mode, there is a check to make sure
the sub-buffer can be written into. This check is performed even if
length is 0. In the case where this would fall exactly at the end of a
sub-buffer, the check would fail, because the offset would fall exactly
at the beginning of the next sub-buffer.

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

Cleanup: ring buffer: remove lib_ring_buffer_switch_new_end()

lib_ring_buffer_switch_new_end() is a leftover from the days where an
event that would exactly fill the current sub-buffer would automatically
trigger a sub-buffer switch into the next sub-buffer.

Even before the ring buffer code has been moved into lttng-modules, this
behavior had been changed: an event that exactly fills a sub-buffer only
fills this current sub-buffer, and does not need to switch into the
next one to populate the sub-buffer header. This change had been done so
periodical timer switch, which shares the same semantic as an event
exactly filling a sub-buffer, would not create tons of empty
sub-buffers.

However, when doing this change, lib_ring_buffer_switch_new_end() has
not been removed, but clearly should have been. Its job is now performed
by the event "commit".

lib_ring_buffer_switch_new_end() has no effect, since padding_size is
always 0.

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