Mathieu Desnoyers [Wed, 5 Oct 2016 11:20:32 +0000 (07:20 -0400)]
Fix: work-around upstream Linux timekeeping bug
Linux commit
27727df240c7 ("Avoid taking lock in NMI path with
CONFIG_DEBUG_TIMEKEEPING"), changed the logic to open-code
the timekeeping_get_ns() function, but forgot to include
the unit conversion from cycles to nanoseconds, breaking the
function's output, which impacts LTTng.
The following kernel versions are affected: 4.8, 4.7.4+, 4.4.20+,
4.1.32+
We expect that the upstream fix will reach the master and stable
branches timely before the next releases, so we use 4.8.1, 4.7.7,
4.4.24, and 4.1.34 as upper bounds (exclusive).
Fall-back to the non-NMI-safe trace clock for those kernel versions.
We simply discard events from NMI context with a in_nmi() check,
as we did before Linux 3.17.
Link: http://lkml.kernel.org/r/1475636148-26539-1-git-send-email-john.stultz@linaro.org
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 3 Oct 2016 21:25:13 +0000 (17:25 -0400)]
Documentation: lttng-modules 2.7 supports Linux < 4.8
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 3 Oct 2016 16:15:35 +0000 (12:15 -0400)]
Version 2.7.5
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 1 Sep 2016 22:08:15 +0000 (18:08 -0400)]
Fix: handle large number of pages or subbuffers per buffer
Do no trigger kernel console warnings when we try to allocate too many
pages, or a too large kmalloc area for page array (within a subbuffer),
or a sub-buffer array (within a buffer).
Use vmalloc/vfree for the "pages" local variable used only during
allocation, which is an array of nr_subbuf * nr_pages_per_subbuf
pointers. This ensures we do not limit the overall buffer size due to
kmalloc limitations.
Fixes #1031
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Jonathan Rajotte [Thu, 1 Sep 2016 21:52:09 +0000 (17:52 -0400)]
Fix: unregister cpu hotplug notifier on buffer alloc error
The cpu hotplug notifier needs to be unregistered in the error path of
buffer allocation, else it eventually causes kernel OOPS when the kernel
accesses freed memory of the notifier block.
Fixes #1031
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 20 May 2016 20:55:56 +0000 (16:55 -0400)]
Version 2.7.4
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 20 May 2016 01:32:06 +0000 (21:32 -0400)]
Fix: ring buffer: honor switch parameter type in remote switch
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 20 May 2016 01:31:07 +0000 (21:31 -0400)]
Fix: only flush non-metadata channels
Issue introduced by commit
"Fix: do not generate packet at destroy after stop".
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 20 May 2016 00:30:07 +0000 (20:30 -0400)]
Fix: don't perform extra flush on metadata channel
The metadata channel requires that the LTTng client layer and the ring
buffer keep a notion of the amount of data produced in the channel.
This issue has been introduced recently by commit
"Fix: flush empty packets on snapshot channel".
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 18 May 2016 01:24:46 +0000 (21:24 -0400)]
Fix: flush empty packets on snapshot channel
Snapshot operation on a non-stopped stream should use a "final" flush to
ensure empty packets are flushed, so we gather timestamps at the moment
where the snapshot is taken. This is important for streams that have a
low amount of activity.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 18 May 2016 01:23:24 +0000 (21:23 -0400)]
Fix: do not generate packet at destroy after stop
In the following scenario:
- create, enable events (kernel),
- start
- ...
- stop (await for data_pending to complete)
- destroy
- rm the trace directory
We would expect that the "rm" operation would not conflict with the
consumer daemon trying to output data into the trace files, since the
"stop" operation ensured that there was no data_pending.
However, the "destroy" operation currently generates an extra packet
after the data_pending check. This causes the consumer daemon to try to
perform trace file rotation concurrently with the trace directory
removal in the scenario above, which triggers errors. The main reason
why this empty packet is generated by "destroy" is to deal with trace
start/stop scenario which would otherwise generate a completely empty
stream.
Therefore, introduce the concept of a "quiescent stream". It is
initialized at false on stream creation (first packet is empty). When
tracing is started, it is set to false (for cases of start/stop/start).
When tracing is stopped, if the stream is not quiescent, perform a
"final" flush (which will generate an empty packet if the current packet
was empty), and set quiescent to true. On "destroy" stream: if the
stream is not quiescent, perform a "final" flush, and set the quiescent
state to true.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 18 May 2016 01:16:22 +0000 (21:16 -0400)]
Fix: compat ioctl for flush/get metadata version
Unlike the non-compat version, the compat ioctl lttng ABI code for the
ring buffer flush operation does not invoke
lttng_metadata_output_channel before calling the ring buffer operation.
This could lead to incomplete metadata on 64-bit kernels running with
32-bit lttng userland.
There is also a discrepancy between the non-compat and compat code: the
GET_METADATA_VERSION operation is performed before calling the ring
buffer code. Ensure both non-compat and compat code are alike.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 16 May 2016 20:51:00 +0000 (16:51 -0400)]
Fix: ctf_string() should handle NULL pointers
The regmap instrumentation can send a NULL string (e.g. on ARM32).
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 11 May 2016 14:43:21 +0000 (10:43 -0400)]
Fix: add modules target to base Makefile
This way running 'make modules' will indeed build the modules
like in the upstream build system.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 11 May 2016 14:40:10 +0000 (10:40 -0400)]
Fix: make clean does not include dot-config
Skip the CONFIG_TRACEPOINT test if dot-config has not been included.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 11 May 2016 02:06:18 +0000 (22:06 -0400)]
Fix: trigger build error if CONFIG_TRACEPOINTS is not set
Fixes: #1015
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Thu, 5 May 2016 20:10:36 +0000 (16:10 -0400)]
Fix: mm_page_alloc_extfrag instrumentation for kernel 3.16.35
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 2 May 2016 16:32:48 +0000 (12:32 -0400)]
Fix: integer endianness metadata generation
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 26 Apr 2016 22:41:04 +0000 (18:41 -0400)]
Fix: endianness of integers received by filter
We need to byteswap integers passed to the filter when they are tagged
as being in an endianness which differs from the architecture
endianness, so the integer comparisons make sense in terms of value
rather than raw bytes for those fields.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 26 Apr 2016 20:38:38 +0000 (16:38 -0400)]
Version 2.7.3
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Thu, 21 Apr 2016 20:11:04 +0000 (16:11 -0400)]
Fix: writeback probe on RHEL >= 3.10.0-327.10.1
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Wed, 20 Apr 2016 21:50:05 +0000 (17:50 -0400)]
Fix: RHEL kernel 3.10.0-327.13.1 compat
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 20 Apr 2016 21:58:33 +0000 (17:58 -0400)]
Fix: ctf_user_integer should not trigger page fault
It is not used by the instrumentation currently, but fix it
nevertheless.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 20 Apr 2016 21:56:45 +0000 (17:56 -0400)]
Fix: lib_ring_buffer_copy_from_user_inatomic error handling
It should treat nonzero return value of
lib_ring_buffer_do_copy_from_user_inatomic() as errors, 0 as success.
It is currently unused by the instrumentation.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 5 Apr 2016 19:25:20 +0000 (15:25 -0400)]
Fix: remove unused gfpflags.h from btrfs and compaction instrumentation
Its content is unused, and this file disappears in Linux 4.6-rc.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 5 Apr 2016 13:26:51 +0000 (09:26 -0400)]
Fix: Remove dead code from filter interpreter
Detected by Coverity:
** CID
1353794: Control flow issues (DEADCODE)
/lttng-filter-interpreter.c: 123 in stack_strcmp()
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 5 Apr 2016 12:36:44 +0000 (08:36 -0400)]
Fix: x86 kvm mmutrace instrumentation for kernel < 3.6
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 5 Apr 2016 12:09:13 +0000 (08:09 -0400)]
Fix: mm_vmscan instrumentation: remove unused gfpflags.h include
This header disappears in kernel 4.6, but we don't actually use it, so
remove its include.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 5 Apr 2016 12:05:42 +0000 (08:05 -0400)]
Fix: use of uninitialized ret value in lttng_abi_create_event
Fixes the following compiler warning:
/home/jenkins/workspace/lttng-modules_VERSION_param-build/arch/x86-32/src/lttng-modules/lttng-abi.c: In function ‘lttng_abi_create_event’:
/home/jenkins/workspace/lttng-modules_VERSION_param-build/arch/x86-32/src/lttng-modules/lttng-abi.c:987:16: warning: ‘ret’ may be used uninitialized in this function [-Wmaybe-uninitialized]
int event_fd, ret;
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 4 Apr 2016 19:57:24 +0000 (15:57 -0400)]
Fix: kmem instrumentation: remove unused gfpflags.h include
This header disappears in kernel 4.6, but we don't actually use it, so
remove its include.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 4 Apr 2016 19:55:28 +0000 (15:55 -0400)]
Fix: 4.6 kernel: use in_compat_syscall
in_compat_sycall appears in kernel 4.6. Use it instead of
is_compat_task(), but use is_compat_task() as fallback.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 4 Apr 2016 18:39:16 +0000 (14:39 -0400)]
Fix: use BUG_ON instead of accessing array subscript above array bounds
Building lttng-modules lttng-filter-interpreter.c on a 4.6-rc kernel
triggers the following gcc warning:
In file included from /home/compudj/git/lttng-modules/lttng-filter-interpreter.c:25:0:
/home/compudj/git/lttng-modules/lttng-filter-interpreter.c: In function ‘lttng_filter_interpret_bytecode’:
/home/compudj/git/lttng-modules/lttng-filter.h:144:14: warning: array subscript is above array bounds [-Warray-bounds]
&(stack)->e[top]; \
^
/home/compudj/git/lttng-modules/lttng-filter-interpreter.c:714:4: note: in expansion of macro ‘estack_ax’
estack_ax(stack, top)->u.s.str = insn->data;
^
/home/compudj/git/lttng-modules/lttng-filter.h:144:14: warning: array subscript is above array bounds [-Warray-bounds]
&(stack)->e[top]; \
^
/home/compudj/git/lttng-modules/lttng-filter-interpreter.c:715:4: note: in expansion of macro ‘estack_ax’
estack_ax(stack, top)->u.s.seq_len = UINT_MAX;
This is because the bound checking is performed in a prior validation
phase (which the compiler does not know about), and we only
WARN_ON_ONCE() if the interpreter sees values that don't fit in array
range.
Use BUG_ON() in the interpreter instead, which ensures we never, ever
reach the out-of-bound condition from a compiler perspective.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Tue, 29 Mar 2016 23:32:19 +0000 (19:32 -0400)]
Fix: Add granularity to RHEL kernel version checks
Our RHEL kernel version macros were based on the major and minor version
of the distribution version like "7.2" instead of the full kernel version
like "327.4.4". This prevented us from adding compatibility changes with
sufficient granularity.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 10 Mar 2016 01:39:48 +0000 (20:39 -0500)]
Version 2.7.2
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 8 Mar 2016 16:34:01 +0000 (11:34 -0500)]
Fix: filter interpreter with userspace strings
The filter interpreter is populating the "str" register field, but the
string comparison functions expect to use the "user_str" register field
whenever it deals with a userspace string ("user" flag set to 1).
Since we are using __copy_from_user_inatomic(), this is typically
bailing out on anything that requires to read the string (except the '*'
wildcard).
Fixes #943
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 8 Mar 2016 15:49:52 +0000 (10:49 -0500)]
Fix: rename kvm x86 TRACE_SYSTEM to avoid name clash
The lttng-probe-kvm and lttng-probe-kvm-x86 have the same TRACE_SYSTEM
string. Because of that, both modules can't be loaded at the same time.
The second module fails to load silently. Rename the TRACE_SYSTEM of
lttng-probe-kvm-x86 to avoid the clash.
We need to use the LTTNG_TRACEPOINT_EVENT_MAP to perform this
namespacing while still hooking onto the kernel tracepoints.
Fixes #895
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 8 Mar 2016 15:30:33 +0000 (10:30 -0500)]
Fix: format address fields as hexadecimal
Fixes #947
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 7 Mar 2016 23:24:25 +0000 (18:24 -0500)]
PowerPC-64 ABIv1: add build error if KALLSYMS_ALL is missing
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 25 Feb 2016 15:46:54 +0000 (10:46 -0500)]
Fix: system call instrumentation overrides
* All architectures
- For mmap, print the "ret" value as an hexadecimal integer (rather than
base 10), which better suits a pointer.
- Add missing "clone" override define, which ensures the clone
system call override is used rather than leaving it as an
"unknown" system call.
* ARM32
- Add missing output return values to arm_fadvise64_64,
sync_file_range2.
- Handle the 64-bit parameters of both arm_fadvise64_64 and
sync_file_range2 correctly on arm32, on big and little endian builds.
- Move the mmap2 system call instrumentation from "pointers" to
"integers" override header, and add the missing return value
(sc_exit).
* PPC32
- Add sync_file_range2 system call instrumentation. Handle the
64-bit parameters on big and little endian builds.
- Move the mmap2 system call instrumentation from "pointers" to
"integers" override header, and add the missing return value
(sc_exit).
* PPC64 compat
- Include compatibility (ppc32) header in ppc64 build to trace compat
system calls on PPC64 big endian.
* x86-32
- Add mmap2 system call instrumentation override.
- Add sync_file_range instrumentation override, correctly combining
the low/high parts of offset and nbytes into 64-bit fields.
Tested-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 4 Mar 2016 01:00:50 +0000 (20:00 -0500)]
Fix: page_alloc wrapper incorrect parenthesis
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 12 Feb 2016 20:11:04 +0000 (15:11 -0500)]
Fix: add missing versions and init call for page_alloc wrapper
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Wed, 10 Feb 2016 16:08:14 +0000 (11:08 -0500)]
Fix: Define lttng-tracer metadata only once
Fixes sysfs trying to create the lttng-tracer module version node
multiple times.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 8 Feb 2016 21:38:46 +0000 (16:38 -0500)]
Fix: handle negative clock offset
In the unlikely situation where a system sets its hardware clock
(CLOCK_REALTIME) to 0 (Epoch) after boot, the difference
monotonic - realtime
becomes negative.
Fixup this situation by returning a 0 offset in this case.
This ensures that trace viewer implementations (e.g. babeltrace) which
currently cannot handle the negative offset (known bug) still work with
the generated traces.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 5 Feb 2016 22:53:07 +0000 (17:53 -0500)]
Fix: sched instrumentation for 4.1-rt11
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 5 Feb 2016 22:53:06 +0000 (17:53 -0500)]
Add RT patch version macros
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 29 Jan 2016 22:46:03 +0000 (17:46 -0500)]
Fix: Ubuntu kernel range for block_rq_complete
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 29 Jan 2016 22:46:02 +0000 (17:46 -0500)]
Fix: Ubuntu kernel range for mm_page_alloc_extfrag
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 25 Jan 2016 19:23:43 +0000 (14:23 -0500)]
Fix: update vmscan instrumentation for kernel 4.5
The following tracepoint changes made their way into Linux 4.5-rc1:
commit
3aa2385111168187f24a6db04697c6fab0fab9b4
Author: yalin wang <yalin.wang2010@gmail.com>
Date: Thu Jan 14 15:18:30 2016 -0800
mm/vmscan.c: change trace_mm_vmscan_writepage() proto type
and
commit
ba5e9579433aefcdccdec207601e124d3bdf2a71
Author: yalin wang <yalin.wang2010@gmail.com>
Date: Thu Jan 14 15:18:48 2016 -0800
mm: change mm_vmscan_lru_shrink_inactive() proto types
Update the lttng-modules instrumentation to match those changes.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 19 Jan 2016 14:51:55 +0000 (09:51 -0500)]
Fix: check reference counts for overflow
Linux kernel CVE-2016-0728 is a use-after-free based on overflow of the
reference counting mechanism.
Implement a kref wrapper in lttng that validates overflows, and use it
instead of kref_get(). Also check explicitly for overflows on file
fcount counters.
This should not be an issue in practice in lttng-modules because the ABI
is only exposed to root, but let's err on the safe side.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Thu, 14 Jan 2016 19:35:47 +0000 (14:35 -0500)]
Fix: RHEL 7.2 scsi instrumentation
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Thu, 14 Jan 2016 19:35:46 +0000 (14:35 -0500)]
Fix: RHEL 7.2 kvm instrumentation
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 5 Jan 2016 18:34:58 +0000 (13:34 -0500)]
Version 2.7.1
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Antoine Busque [Fri, 4 Dec 2015 20:16:59 +0000 (15:16 -0500)]
Fix: Use MAX_RT_PRIO offset in sched_wakeup_template
All other sched_* events with prio fields offset the task_struct's
prio value by subtracting MAX_RT_PRIO. This way the value corresponds
to that given by the kernel's task_prio function, and thus with the
prio context field.
This adds the same offset to sched_wakeup, sched_wakeup_new, and
sched_waking, via the sched_wakeup_template event class. This way
their value is comparable to other sched_* events, and the prio
context.
Signed-off-by: Antoine Busque <abusque@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 27 Nov 2015 19:57:44 +0000 (14:57 -0500)]
Fix: update sched instrumentation for kernel 4.4.0
Due to upstream Linux kernel commit:
commit
c73464b1c8434ad4cbfd5369c3e724f3e8ffe5a4
Author: Peter Zijlstra <peterz@infradead.org>
Date: Mon Sep 28 18:06:56 2015 +0200
sched/core: Fix trace_sched_switch()
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 10 Nov 2015 18:01:50 +0000 (13:01 -0500)]
Fix: instrumentation: v4lv2 missing include
Starting at least with 4.3 kernel, we need to specifically include
media/videobuf2-core.h.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Sun, 25 Oct 2015 15:23:31 +0000 (11:23 -0400)]
Fix: header size larger than 256 bytes
When adding large context (e.g. callstack), headers larger than 256
bytes cause discrepancy between calculated size and size written into
the trace buffers. This generates a corrupted trace and triggers a
warning in ring buffer backend, which triggers a safety net disabling
tracing for the current channel.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 23 Oct 2015 21:17:43 +0000 (17:17 -0400)]
Fix: tracepoint listing misses last loaded probe
The seqfile iteration listing tracepoints operates on the probe list
directly, without going through the lttng_get_probe_list_head()
accessor which ensures that the lazy list of probes is moved to the
actual list of probes.
This causes an issue when loading a probe and then listing the
tracepoints when no tracing sessions are active: the last probe loaded
is missing from the listing.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 9 Oct 2015 16:24:23 +0000 (12:24 -0400)]
Fix: lttng trace-clock needs to disable preemption
We use a per-cpu data structure, and some contexts (e.g. buffer create)
call this function with preemption enabled. This is the case also for
buffer flush operation.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 24 Sep 2015 20:50:21 +0000 (16:50 -0400)]
Version 2.7.0
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 15 Sep 2015 19:46:01 +0000 (15:46 -0400)]
Fix: writeback instrumentation update for 4.3 Linux kernel
Linux upstream commit:
commit
5634cc2aa9aebc77bc862992e7805469dcf83dac
Author: Tejun Heo <tj@kernel.org>
Date: Tue Aug 18 14:54:56 2015 -0700
writeback: update writeback tracepoints to report cgroup
Updated the tracepoint signatures for writeback probes. Follow this
change. Don't add extra data (cgroup information) yet.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 14 Sep 2015 16:12:43 +0000 (12:12 -0400)]
Fix: update sched wakeup instrumentation for 4.3 kernel
Change introduced in upstream Linux:
commit
fbd705a0c6184580d0e2fbcbd47a37b6e5822511
Author: Peter Zijlstra <peterz@infradead.org>
Date: Tue Jun 9 11:13:36 2015 +0200
sched: Introduce the 'trace_sched_waking' tracepoint
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 11 Sep 2015 20:06:25 +0000 (16:06 -0400)]
Fix: lttng-test build failure for kernels < 3.0
Fixes an undefined reference to kstrtouint_from_user.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 9 Sep 2015 20:43:28 +0000 (16:43 -0400)]
Version 2.7.0-rc2
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Lars Persson [Tue, 8 Sep 2015 13:31:46 +0000 (15:31 +0200)]
Use 3.18 syscall instrumentation for MIPS32
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Lars Persson [Tue, 8 Sep 2015 13:31:45 +0000 (15:31 +0200)]
Update the MIPS32 syscall instrumentation to 3.18
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Lars Persson [Tue, 8 Sep 2015 13:20:15 +0000 (15:20 +0200)]
Fix: Move pipe override to the arch specific pointers_override.h
The MIPS architecture has a non-standard calling convention for pipe()
and cannot use the provided override for the pipe event.
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 4 Sep 2015 20:36:22 +0000 (16:36 -0400)]
Implement lttng test module
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 4 Sep 2015 05:29:49 +0000 (01:29 -0400)]
Cleanup: remove logically dead code
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 4 Sep 2015 05:16:52 +0000 (01:16 -0400)]
Fix: filter validator use after free
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 4 Sep 2015 05:12:53 +0000 (01:12 -0400)]
Fix: use after free in channel release
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 2 Sep 2015 02:27:35 +0000 (22:27 -0400)]
Fix: non-enabler events should be disabled by default
The session daemon expects that all events (enablers and events) are
disabled when created, so we can attach a filter before enabling them.
Also fix the kretprobe to ensure we can enable/disable both the entry
and return probe (--function instrumentation).
Fixes: #926
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Mon, 31 Aug 2015 13:57:18 +0000 (09:57 -0400)]
Fix: build failure on 2.6.36
Fixes build failure on the 2.6.36 kernel branch which misses the
vzalloc() function.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Fri, 21 Aug 2015 16:28:44 +0000 (09:28 -0700)]
Cleanup: split bdi_dirty_ratelimit with ifdef into separate macros
Makes the code easier to maintain in the long run.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 21 Aug 2015 02:23:15 +0000 (22:23 -0400)]
Fix: writeback instrumentation for 4.2 kernels
Bandwidth related fields were moved by this commit:
commit
a88a341a73be4ef035ca26170c849f002797da27
Author: Tejun Heo <tj@kernel.org>
Date: Fri, 22 May 2015 17:13:28 -0400
writeback: move bandwidth related fields from backing_dev_info into
bdi_writeback
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 29 Jul 2015 17:00:04 +0000 (13:00 -0400)]
Fix: kmem probe with Ubuntu 3.13 kernels
Fixes an undefined reference to get_pageblock_flags_mask.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Chris J Arges [Fri, 24 Jul 2015 18:18:59 +0000 (13:18 -0500)]
Fix: disable kvm probe if lapic.h isn't found
In a typical distribution if just linux-headers is installed, lapic.h
will not be installed. We should check if that file exists and not build
the module.
Signed-off-by: Chris J Arges <chris.j.arges@canonical.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 16 Jul 2015 17:02:44 +0000 (13:02 -0400)]
Fix: timer instrumentation for 4.2 kernels
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 16 Jul 2015 17:02:23 +0000 (13:02 -0400)]
Fix: ext4 instrumentation for 4.2 kernels
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 16 Jul 2015 17:02:03 +0000 (13:02 -0400)]
Fix: kvm instrumentation for 4.2 kernels
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Fri, 14 Aug 2015 15:37:31 +0000 (11:37 -0400)]
Fix: mm_page_alloc_extfrag instrumentation for kernel 3.18.10
The following commit has been backported from kernel 3.19.2 into
3.18.10, introducing a change to mm_page_alloc_extfrag and therefore
provoking a build failure on a 3.18.10 kernel:
commit
a4f3f96f850db69890f47984c60606a2362bee89
Author: Vlastimil Babka <vbabka@suse.cz>
Date: Wed Feb 11 15:28:15 2015 -0800
mm: when stealing freepages, also take pages created by splitting buddy page
commit
99592d598eca62bdbbf62b59941c189176dfc614 upstream.
This patch adds the proper version check to allow building on 3.18.y.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 16 Jul 2015 03:49:15 +0000 (23:49 -0400)]
Fix: allow LTTng to be built within kernel tree
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 15 Jul 2015 18:42:51 +0000 (14:42 -0400)]
Version 2.7.0-rc1
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Martin Leisener [Wed, 10 Jun 2015 13:17:37 +0000 (15:17 +0200)]
Fix: Building the event list fails on fragmented memory
On a small arm imx6 solo with 256MB RAM it often happens that memory
becomes fragmented rather quickly, so that kmalloc will not be able to
get enough consecutive pages (enocuntered for example if you enable all
kernel events: lttng enable-event -k syscall --all).
This patch switches the allocation to vmalloc. Tested for x86 on Ubuntu
12.04 Lts and on imx6 solo 256MB RAM
If this patch is not applied, you can identify low and/or fragmented
memory failures by looking at the kernel ring buffer (please ignore DMA,
it is due to some memory setup misconfiguration, should read Normal):
...
[ 321.993820] lttng-sessiond: page allocation failure: order:4, mode:0x1040d0
...
[ 321.994711] lowmem_reserve[]: 0 0 0
[ 321.994727] DMA: 801*4kB (UEMC) 424*8kB (EMC) 355*16kB (UEMC) 344*32kB (MC) 340*64kB (C) 8*128kB (C) 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB 0*8192kB 0*16384kB 0*32768kB = 46068kB
[ Edit by Mathieu: use vzalloc() rather than vmalloc() + memset 0. ]
Signed-off-by: Martin Leisener <martin@leisener.de>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 25 Jun 2015 13:10:52 +0000 (09:10 -0400)]
Fix: use after free on metadata cache reallocation
When the metadata cache is expanded (reallocated) by
lttng_metadata_printf(), the metadata cache reader
(lttng_metadata_output_channel()) may use freed memory, because the
metadata cache is not protected from concurrent read accesses. The
metadata cache updates are protected from each other by the sessions
mutex, but metadata cache reads do not hold the sessions mutex.
Actually, the comment on top of lttng_metadata_output_channel() stating
"We have exclusive access to our metadata buffer (protected by the
sessions_mutex)" is simply wrong, because this mutex is never held when
calling lttng_metadata_output_channel().
Promote the per-stream lock to the metadata cache used by each of those
metadata streams, thus ensuring mutual exclusion between metadata cache
reallocation and readers.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Michael Jeanson [Wed, 17 Jun 2015 20:28:02 +0000 (16:28 -0400)]
Fix: version checks for kernel 4.0
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Mon, 15 Jun 2015 16:28:25 +0000 (12:28 -0400)]
Fix: Incorrect Ubuntu kernel patchlevel
The kernel version used should be the one in Makefile and generated
version.h, not the one tweaked by Ubuntu as "0" in the built package.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Sat, 13 Jun 2015 09:14:17 +0000 (11:14 +0200)]
Fix: cpu-id context should use int type
cycles_t makes no sense as type for cpu-id context.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 11 Jun 2015 21:40:21 +0000 (23:40 +0200)]
Implement cpu-id context
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 11 Jun 2015 15:39:16 +0000 (17:39 +0200)]
Implement cpu_id context for filtering
cpu_id context can now be used for filtering, e.g.:
lttng enable-event -k sched_switch --filter '$ctx.cpu_id==1'
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 11 Jun 2015 15:38:46 +0000 (17:38 +0200)]
Fix: context printk mismatch on error
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 11 Jun 2015 08:09:49 +0000 (10:09 +0200)]
Cleanup: misleading comment about deferrable timer
Timer in lttng-modules are not deferrable, because we cannot hook into
the Linux kernel CPU idle events, because no notifiers are exposed for
those.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Martin Leisener [Fri, 29 May 2015 20:29:44 +0000 (16:29 -0400)]
Fix: arm-32 build
The current master branch of lttng-modules doesn't seem to compile for
arm-32 builds.
The patch below repairs the build for arm (and still builds on a x86)
(tested on Ubuntu 12.04 LTS x86 + imx6 solo)
Signed-off-by: Martin Leisener <martin@leisener.de>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Alexandre Montplaisir [Tue, 26 May 2015 20:12:00 +0000 (16:12 -0400)]
Fix: Update kmem event for Ubuntu's 3.16 kernel
The upstream patch in question in commit
72906e1 was cherry-picked into
Ubuntu's 3.16 kernel backport branch (called lts-utopic). The version
check should also include this branch.
The other backport branch, lts-vivid, and all following ones will be
covered by the standard
LINUX_VERSION_CODE >= KERNEL_VERSION(3,19,2)
check, so should not require other workarounds.
References #889.
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 27 May 2015 16:56:05 +0000 (12:56 -0400)]
Fix: RHEL 7.1 block instrumentation
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 27 May 2015 16:19:01 +0000 (12:19 -0400)]
Fix: add RHEL version macros
RHEL 7.1 has backported upstream patches that change the kvm.h
tracepoint prototype. Add version check specific to RHEL to handle this
case.
RHEL 7.x has the required fixes backported that allow removing the 3.10
blacklist check for them, so take this into account.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Wed, 27 May 2015 16:13:32 +0000 (12:13 -0400)]
Fix: add missing parenthesis in ubuntu version check
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Tue, 26 May 2015 19:22:01 +0000 (15:22 -0400)]
Fix: remove regmap instrumentation for kernels < 4.1
Modifications to regmap.h instrumentation in mainline Linux kernel has
been no less than erratic. First, the public instrumentation header
regmap.h include a private driver header in Linux 4.0. Then, regmap.h
is moved to the private driver directory in Linux 4.1. To make things
worse, the 4.0 commit has been picked into stable branches of the Linux
kernel.
Since this does not appear to be an instrumentation of utmost
importance, only build this instrumentation probe if the private header
is found in the driver directory (need full kernel sources).
This removes regmap instrumentation for older kernels.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Alexandre Montplaisir [Fri, 22 May 2015 18:36:53 +0000 (14:36 -0400)]
Fix: Update kmem event for recent Ubuntu kernel
Ubuntu backported a patch into their 3.13 (Trusty) branch which
changes kmem event definition.
Upstream commit
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=
99592d598eca62bdbbf62b59941c189176dfc614
first appears in 4.0 (and was later backported to 3.19.2).
We already have the proper check for this.
This corresponds to Ubuntu commit
http://kernel.ubuntu.com/git/ubuntu/ubuntu-trusty.git/commit/?id=
d938e4bff1089a1f4757ed94a5cace803c6a363b
which was first released with 3.13.0-50.82.
This required adding a new check specifically for the
Ubuntu version.
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 14 May 2015 22:29:58 +0000 (18:29 -0400)]
Fix: missing statedump end event when block device listing available
Incorrect handling of lttng_enumerate_block_devices() return value of 0
(which means "ok"), causing do_lttng_statedump() to return early.
To causes missing "lttng_statedump_end" in traces where block device
enumeration is available (CONFIG_KALLSYMS=y).
Reported-by: Antoine Busque <abusque@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Mathieu Desnoyers [Thu, 14 May 2015 15:33:47 +0000 (11:33 -0400)]
Fix: don't generate incorrect macros for debian/ubuntu version check
When the debian/ubuntu macros are unset, we end up generating macros
that have no right hand operand to "+", which seems to break compilation
on some compilers, even though it will always be optimized away.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
This page took 0.044456 seconds and 4 git commands to generate.