Version 2.0.8

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>

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), 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>

Fix: handle writes of length 0

lib_ring_buffer_write(), lib_ring_buffer_memset() and
lib_ring_buffer_copy_from_user_inatomic() 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.

It triggers this warning:

[65356.890016] ------------[ cut here ]------------
[65356.890016] WARNING: at /home/compudj/git/lttng-modules/wrapper/ringbuffer/../../lib/ringbuffer/../../wrapper/ringbuffer/../../lib/ringbuffer/backend.h:110 lttng_event_write+0x118/0x140 [lttng_ring_buffer_client_mmap_overwrite]()
[65356.890016] Hardware name: X7DAL
[65356.890016] Modules linked in: lttng_probe_writeback(O) lttng_probe_workqueue(O) lttng_probe_vmscan(O) lttng_probe_udp(O) lttng_probe_timer(O) lttng_probe_sunrpc(O) lttng_probe_statedump(O) lttng_probe_sock(O) lttng_probe_skb(O) lttng_probe_signal(O) lttng_probe_scsi(O) lttng_probe_sched(O) lttng_probe_rcu(O) lttng_probe_random(O) lttng_probe_printk(O) lttng_probe_power(O) lttng_probe_net(O) lttng_probe_napi(O) lttng_probe_module(O) lttng_probe_kvm(O) lttng_probe_kmem(O) lttng_probe_jbd2(O) lttng_probe_jbd(O) lttng_probe_irq(O) lttng_probe_ext4(O) lttng_probe_ext3(O) lttng_probe_compaction(O) lttng_probe_btrfs(O) lttng_probe_block(O) lttng_types(O) lttng_ring_buffer_metadata_mmap_client(O) lttng_ring_buffer_client_mmap_overwrite(O) lttng_ring_buffer_client_mmap_discard(O) lttng_ring_buffer_metadata_client(O) lttng_ring_buffer_client_overwrite(O) lttng_ring_buffer_client_discard(O) lttng_tracer(O) lttng_kretprobes(O) lttng_ftrace(O) lttng_kprobes(O) lttng_statedump(O) lttng_lib_ring_buffer(O) cpufreq_ondemand loop e1000e kvm_intel kvm ptp pps_core [last unloaded: lttng_lib_ring_buffer]
[65357.287529] Pid: 0, comm: swapper/7 Tainted: G           O 3.9.4-trace-test #143
[65357.309694] Call Trace:
[65357.317022]  <IRQ>  [<ffffffff8103a3ef>] warn_slowpath_common+0x7f/0xc0
[65357.336893]  [<ffffffff8103a44a>] warn_slowpath_null+0x1a/0x20
[65357.354368]  [<ffffffffa0ff17b8>] lttng_event_write+0x118/0x140 [lttng_ring_buffer_client_mmap_overwrite]
[65357.383025]  [<ffffffffa100134f>] __event_probe__block_rq_with_error+0x1bf/0x220 [lttng_probe_block]
[65357.410376]  [<ffffffff812ea134>] blk_update_request+0x324/0x720
[65357.428364]  [<ffffffff812ea561>] blk_update_bidi_request+0x31/0x90
[65357.447136]  [<ffffffff812eb68c>] blk_end_bidi_request+0x2c/0x80
[65357.465127]  [<ffffffff812eb6f0>] blk_end_request+0x10/0x20
[65357.481822]  [<ffffffff81406b7c>] scsi_io_completion+0x9c/0x670
[65357.499555]  [<ffffffff813fe320>] scsi_finish_command+0xb0/0xe0
[65357.517283]  [<ffffffff81406965>] scsi_softirq_done+0xa5/0x140
[65357.534758]  [<ffffffff812f1d30>] blk_done_softirq+0x80/0xa0
[65357.551710]  [<ffffffff81043b00>] __do_softirq+0xe0/0x440
[65357.567881]  [<ffffffff81043ffe>] irq_exit+0x9e/0xb0
[65357.582754]  [<ffffffff81026465>] smp_call_function_single_interrupt+0x35/0x40
[65357.604388]  [<ffffffff8167be2f>] call_function_single_interrupt+0x6f/0x80
[65357.624976]  <EOI>  [<ffffffff8100ac06>] ? default_idle+0x46/0x300
[65357.643541]  [<ffffffff8100ac04>] ? default_idle+0x44/0x300
[65357.660235]  [<ffffffff8100b899>] cpu_idle+0x89/0xe0
[65357.675109]  [<ffffffff81664911>] start_secondary+0x220/0x227

Always from an event that can write a 0-length field as last field of
its payload, and it always happen directly on a sub-buffer boundary.

While we are there, check for length 0 in lib_ring_buffer_read_cstr()
too.

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

Fix: ring buffer: RING_BUFFER_FLUSH ioctl buffer corruption

lib_ring_buffer_switch_slow() clearly states:

 * Note, however, that as a v_cmpxchg is used for some atomic
 * operations, this function must be called from the CPU which owns the
 * buffer for a ACTIVE flush.

But unfortunately, the RING_BUFFER_FLUSH ioctl does not follow these
important directives. Therefore, whenever the consumer daemon or session
daemon explicitly triggers a "flush" on a buffer, it can race with data
being written to the buffer, leading to corruption of the reserve/commit
counters, and therefore corruption of data in the buffer. It triggers
these warnings for overwrite mode buffers:

[65356.890016] WARNING: at
/home/compudj/git/lttng-modules/wrapper/ringbuffer/../../lib/ringbuffer/../../wrapper/ringbuffer/../../lib/ringbuffer/backend.h:110 lttng_event_write+0x118/0x140 [lttng_ring_buffer_client_mmap_overwrite]()

Which indicates that we are trying to write into a sub-buffer for which
we don't have exclusive access. It also causes the following warnings to
show up:

[65335.896208] ring buffer relay-overwrite-mmap, cpu 5: records were lost. Caused by:
[65335.896208]   [ 0 buffer full, 80910 nest buffer wrap-around, 0 event too big ]

Which is caused by corrupted commit counter.

Fix this by sending an IPI to the CPU owning the flushed buffer for
per-cpu synchronization. For global synchronization, no IPI is needed,
since we allow writes from remote CPUs.

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

Version 2.0.7

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

Fix build and load against linux-2.6.33.x

* lttng-event.h declared but did not implement
  lttng_add_perf_counter_to_ctx on kernel >=2.6.33, the implementation
  was in lttng-context-perf-counters.c, which was only included for
  kernel >=2.6.34.  This prevented the module from being loaded.

* on kernel 2.6.33.x, lttng-context-perf-counters.c complains about
  implicit declaration for {get,put}_online_cpus and
  {,un}register_cpu_notifier; so fix header inclusion.

Signed-off-by: Samuel Martin <smartin@aldebaran-robotics.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix check in lttng_strlen_user_inatomic

__copy_from_user_inatomic returns the number of bytes that could not be
copied, not an error code. This fixes the test accordingly.

[ Edit by Mathieu Desnoyers: change "ret" type to unsigned long too. ]

Signed-off-by: Simon Marchi <simon.marchi@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: statedump hang/too early completion due to logic error

The previous "Fix: statedump hang due to incorrect wait/wakeup use" was
not actually fixing the real problem.

The issue is that we should pass the expected condition to wait_event()
rather than its contrary.

This bug has been sitting there for a while. I suspect that a recent
change in the Linux scheduler behavior for newly spawned worker threads
might have contributed to trigger the hang more reliably.

The effects of this bugs are:
- possible hang of the lttng-sessiond (within the kernel) at tracing
  start,
- the statedump end event is traced before all worker threads have
  actually completed, which can confuse LTTng viewer state systems.

Reported-by: Phil Wilshire <sysdcs@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: statedump hang due to incorrect wait/wakeup use

Reported-by: Phil Wilshire <sysdcs@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update version to 2.0.6

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

lib/ringbuffer/frontend_api.h: Include linux/prefetch.h.

prefetch() is used in this file so make sure we include the appropriate
header before using it.

Signed-off-by: Will Newton <will.newton@imgtec.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix warning with 3.6.0 kernel

No actual runtime impact, but neater anyway.

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

Fix: compat layer only on supported architectures

The compat_ulong_t type is only defined on some architectures, so we
need to protect the compat defines with the ifdef CONFIG_COMPAT.

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

Fix ring buffer 32/64 bits compat

The type of fields exchanged between kernel and userspace must be
compat_ulong_t instead of unsigned long in case of compat where
userspace is 32 bits and kernel is 64 bits.

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

Fix cleanup: move lttng-tracer-core.h include to lib ringbuffer config.h

This is where it is used. Add a #ifdef check for the LTTNG_TRACER_CORE_H
define to ensure that the header it depends on is always included.

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

Fix ring_buffer_frontend.c: missing include lttng-tracer-core.h

In lib/ringbuffer/ring_buffer_frontend.c, RING_BUFFER_ALIGN is undefined,
leading to no alignment offset being recorded after the call to
config->cb.record_header_size() in lib_ring_buffer_try_reserve_slow().

However, lttng-ring-buffer-client.h does define RING_BUFFER_ALIGN, so
the alignment offset will be produced when the packet header is written
in lttng_write_event_header().

This discrepancy may be observed on architectures that don't set
CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS, such as ARM, with a babeltrace
error such as:

babeltrace: ../../include/babeltrace/ctf/types.h:206: ctf_pos_get_event:
Assertion `pos->offset <= pos->content_size' failed.
Aborted

indicating that the actual content size differs from the calculated one
due to the difference in alignment.  Including the appropriate header
file in ring_buffer_frontend.c solves the problem.

Signed-off-by: Wade Farnsworth <wade_farnsworth@mentor.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

fix timestamps on architectures without CONFIG_KTIME_SCALAR

trace_clock_monotonic_wrapper() should return a u64 representing the
number of nanoseconds since system startup.
ktime_get() provides that value directly within its .tv64 field only
on those architectures defining CONFIG_KTIME_SCALAR, whereas in all
other cases (e.g. PowerPC) a ktime_to_ns() conversion (which
translates back to .tv64 when CONFIG_KTIME_SCALAR is defined)
becomes necessary.

Signed-off-by: Gerlando Falauto <gerlando.falauto@keymile.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update version to 2.0.5

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

Fix: statedump: disable vm maps enumeration

We need the tasklist lock to correctly dump these, which is unavailable
to modules. Disable this feature for now.

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

Fix: ensure userspace accesses are done with _inatomic

Otherwise, triggers scheduling while atomic (might_sleep()) warnings,
since we call those from a tracepoint probe (with preemption disabled).

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

Fix: vppid context should test for current nsproxy

Triggers a NULL pointer exception with tracing scheduling events with
vppid context activated.

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

Revert "trace event: introduce TP_MODULE_NOAUTOLOAD and TP_MODULE_NOINIT"

This reverts commit e4bd62774b66eea5fd4221f2c006586e911a8de4.

This commit belongs to master, not stable-2.0.

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

trace event: introduce TP_MODULE_NOAUTOLOAD and TP_MODULE_NOINIT

In preparation for introducing uevent tracing. (user-space events)

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

Update version to 2.0.4

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

LTTng: probe-statedump: add #include <linux/sched.h>

This fixes the following build error on ARM;

  CC [M]  drivers/staging/lttng/probes/lttng-probe-statedump.o
In file included from drivers/staging/lttng/probes/../instrumentation/events/lttng-module/../../../probes/lttng-events.h:221:0,
                 from drivers/staging/lttng/probes/../instrumentation/events/lttng-module/../../../probes/define_trace.h:117,
                 from drivers/staging/lttng/probes/../instrumentation/events/lttng-module/lttng-statedump.h:162,
                 from drivers/staging/lttng/probes/lttng-probe-statedump.c:41:
drivers/staging/lttng/probes/../instrumentation/events/lttng-module/../../../probes/../instrumentation/events/lttng-module/lttng-statedump.h:29:1: error: 'TASK_COMM_LEN' undeclared here (not in a function)

Signed-off-by: UCHINO Satoshi <satoshi.uchino@toshiba.co.jp>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

fix: signal_generate event should print utf8 for comm field

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

Update version to 2.0.3

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

Fix: free_event_id check should compare unsigned int with -1U

Otherwise, on 32-bit:

lttng-modules/lttng-events.c:295: warning: comparison is always false due
to limited range of data type

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

Fix: update signal instrumentation for 3.4 kernel

Fix #226

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

Update version to 2.0.2

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

Fix: generate header missing echo -e for escape chars

"why is the -e needed ? What was the experienced oddness ?"

   The \n wasn't getting evaluated to a newline and was being inserted
into the header file as '\n'

   From man echo(1):

       If -e is in effect, the following sequences are recognized:

       \\     backslash

       \a     alert (BEL)

       \b     backspace

       \c     produce no further output

       \e     escape

       \f     form feed

       \n     new line

       \r     carriage return

       \t     horizontal tab

       \v     vertical tab

Suggested-by: Ryan Kyser <Ryan.Kyser@jci.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: add missing uaccess.h include (for ARM)

"What required uaccess.h ?"

   I was getting the following error:
   syscalls_pointers_override.h:6:1: error: implicit declaration of function ?strlen_user?

   I found that strlen_user was defined in asm/uaccess.h, which is
included by linux/uaccess.h.

Suggested-by: Ryan Kyser <Ryan.Kyser@jci.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

README: Document that CONFIG_MODULES is required

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

Fix: README typo

Reported-by: Ryan Kyser <Ryan.Kyser@jci.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: document required and optional kernel config options in README

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

Update version to 2.0.1

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

Fix: is_compat_task !CONFIG_COMPAT compile error on kernels >= 3.3

The Linux kernel already defines a is_compat_task()   (0) when
!CONFIG_COMPAT since kernel 3.3, so if we see it is already defined,
don't redefine it.

Closes #195

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

Update version to 2.0.0 stable

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

Update version to 2.0.0-rc4

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

Update README and add version name place-holder

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

Update version to 2.0.0-rc3

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

Fix clock offset 32-bit multiplication overflow

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

Fix : wrong assign of fd in state dump

lttng_statedump_file_descriptor was using the tgid instead of the fd for
the fd field.

Reported-by: Mathieu Bain <mathieu.bain@polymtl.ca>
Signed-off-by: Julien Desfossez <julien.desfossez@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

License cleanup, ifdef namespace cleanup

- Spell out the LGPL license completely in each file.
- Change LINUX_* ifdefs for LIB_* in lib/ringbuffer header.

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

Fix: ensure power of 2 check handles 64-bit size_t entirely

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

Update version to 2.0.0-rc2

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

Fix: dmesg printout should not print metadata warnings

Metadata channel retry when an event is dropped by the underlying
buffer. We should not print a message showing that the event has been
dropped to the user on dmesg console.

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

Fix: use transport name as channel name

Clarify dmesg printout.

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

Fix: Return -EINVAL instead of print warning if non power of 2 size/num_subbuf

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

Update Version to 2.0.0-rc1

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

Standardize version across toolchain

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

statedump: Use old macro name for kernel 2.6.38

Signed-off-by: Juha Niskanen <juniskane@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update ChangeLog to 2.0-pre15

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

Add timer instrumentation

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

fix: need to undef mainline define

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

fix: Include signal.h instead of irq.h for prototype match check

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

Add signal instrumentation

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

Update ChangeLog to 2.0-pre14

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

syscall tracing: sys_getcpu

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

Add sys_clone x86 instrumentation

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

statedump: fix include circular dep

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

Implement state dump

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

Update Changelog to 2.0-pre13

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

Update README

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

environment: write sysname, release, version, domain to metadata

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

Allow open /proc/lttng for read & write

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

Update Changelog to 2.0-pre12

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

Add x86 32/64 execve syscall instrumentation override

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

Remove unused defines

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

Add padding to ABI

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

Use LTTNG_KERNEL_SYM_NAME_LEN

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

Update version to 1.9.9

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

Add missing double-quotes to clock uuid

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

clock: read bootid as clock monotonic ID

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

Fix comment

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

Cleanup comment

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

clock: output clock description in metadata

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

Properly fix the timekeeping overflow detection

The underlying issue was a mismatch between the ring buffer
configuration description of the number of clock bits (32) saved and the
actual number used (27).

Introduce LTTNG_COMPACT_EVENT_BITS and LTTNG_COMPACT_TSC_BITS across the
code to remove all hardcoded instances of these values to ensure this
kind of mistake does not happen again.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reported-by: Sébastien Barthélémy <barthelemy@crans.org>

Fix init bug

Fix reversed logic bug introduced by commit
80996790fa42072d6525f2adb9a910312077bef6

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

rename lib-ring-buffer to lttng-lib-ring-buffer

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

Remove #warning

Keep a comment about future work, but don't generate compiler warnings.

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

Mass rename: ltt_*/ltt-* to LTTNG_*/LTTNG-*

For files, defines and symbols.

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

Update TODO

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

Update TODO

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

Remove debugfs file (keep only proc file)

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

Rename lttng-debugfs-abi files to lttng-abi

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

Update ChangeLog for 2.0-pre11

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

Fix OOPS caused by reference of config pointer

The config pointer points to data that may vanish across the life-time
of the ring buffer stream references. It's only kept valid for the
duration between the channel create and destroy, but the streams can
keep an extra reference on the channel, and thus need the config to stay
valid.

Perform a copy of the config rather than keeping a pointer to it.

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

Gather detailed info from x86 64 32-bit compat syscall instrumentation

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

lttng lib: ring buffer move null pointer check to open

Let's move the NULL buf check to the file "open", where it belongs.  The
"open" file operation is the actual interface between lib ring buffer
and the modules using it.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng lib: ring buffer remove duplicate null pointer

This check is performed at mapping setup time in
lib_ring_buffer_mmap_buf() already, so we can safely remove this
duplicata.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng lib: ring buffer: remove stale null-pointer

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng wrapper: add missing include to kallsyms wrapper

Needed to keep bissectability.

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

lttng: cleanup one-bit signed bitfields

* Dan Carpenter <dan.carpenter@oracle.com> wrote:
> Sparse complains that these signed bitfields look "dubious".  The
> problem is that instead of being either 0 or 1 like people would expect,
> signed one bit variables like this are either 0 or -1.  It doesn't cause
> a problem in this case but it's ugly so lets fix them.

* walter harms (wharms@bfs.de) wrote:
> hi,
> This patch looks ok to me but this design is ugly by itself.
> It should be replaced by an uchar uint whatever or use a
> real bool (obviously not preferred by this programmes).

bool :1, uchar :1 or uint :1 could make sense. uchar:1/bool:1 won't save
any space here, because the surrounding fields are either uint or
pointers, so alignment will just add padding.

I try to use int/uint whenever possible because x86 CPUs tend to get
less register false-dependencies when using instructions modifying the
whole register (generated by using int/uint types) rather than only part
of it (uchar/char/bool). I only use char/uchar/bool when there is a
clear wanted space gain.

The reason why I never use the bool type within a structure when I want
a compact representation is that bool takes a whole byte just to
represent one bit:

struct usebitfield {
int a;
unsigned int f:1, g:1, h:1, i:1, j:1;
int b;
};

struct usebool {
int a;
bool f, g, h, i, j;
int b;
};

struct useboolbf {
int a;
bool f:1, g:1, h:1, i:1, j:1;
int b;
};

int main()
{
printf("bitfield %d bytes, bool %d bytes, boolbitfield %d bytes\n",
sizeof(struct usebitfield), sizeof(struct usebool),
sizeof(struct useboolbf));
}

result:

bitfield 12 bytes, bool 16 bytes, boolbitfield 12 bytes

This is because each bool takes one byte, while the bitfields are put in
units of "unsigned int" (or bool for the 3rd struct). So in this
example, we need 5 bytes + 3 bytes alignment for the bool, but only 4
bytes to hold the "unsigned int" unit for the bitfields.

The choice between bool and bitfields must also take into account the
frequency of access to the variable, because bitfields require mask
operations to access the selected bit(s). You will notice that none of
these bitfields are accessed on the tracing fast-path: only in
slow-paths. Therefore, space gain is more important than speed here.

One might argue that I have so few of these fields here that it does not
make an actual difference to go for bitfield or bool. I am just trying
to choose types best suited for their intended purpose, ensuring they
are future-proof and will allow simply adding more fields using the same
type, as needed.

So I guess I'll go for unsigned int :1 (rather than uint:1, so we
keep source-level compatibility with user-space LTTng-UST).

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add TODO file

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

Update symbol name length max size to 256

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

Fix last modifications to string_from_user operations

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

Document that depmod needs to be executed by hand

External modules don't seem to actually execute depmod on make install,
even though the console output says so. Document that it needs to be
manually executed in the README file.

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