Jonathan Rajotte [Thu, 27 Aug 2020 19:07:06 +0000 (15:07 -0400)]
lttng: move parse_userspace_probe_opts to a common util
parse_userspace_probe_opts will be reused for trigger creation.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I377e0a59786794a74a9b0a7787b6f9a221245df1
Jérémie Galarneau [Fri, 12 Feb 2021 18:34:47 +0000 (13:34 -0500)]
Clean-up: lttng: silence bogus warning
clang warns that `name_search_ret` may be used unitiliazed if the domain
is none of {ust, python, log4j, jul}. However, those branches are only
accessible when one of these domains is used.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ibfea5be140a29a60ff8a60f0a994fcd815adb85d
Jérémie Galarneau [Thu, 27 Aug 2020 19:16:29 +0000 (15:16 -0400)]
lttng: move log level name to numerical value conversions to a common util
Move the utils used by the `enable-event` command to convert log level
names (textual) to their values (entries in their respective enums).
Since these utils will be used by other commands in the future, the code
is cleaned-up, notably to maintain type-safety by returning specific
enums rather than casting to `int`.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3f655bed6cd5b12cae1e93421f9d1c41ab1b6247
Jérémie Galarneau [Thu, 11 Feb 2021 16:43:51 +0000 (11:43 -0500)]
sessiond: action-executor: log trigger name rather than its address
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: If262ee1d77bc0ea5d7c1c7dc7de286f5047d35f2
Jérémie Galarneau [Thu, 11 Feb 2021 16:42:10 +0000 (11:42 -0500)]
Fix: action-executor: address of trigger name logged instead of trigger name
The "%p" was left in place when some logging statements of action
executor were changed to log trigger names rather than their address.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ic66327b53dd0add46de574df6f654620e2a7e3e5
Jérémie Galarneau [Thu, 11 Feb 2021 16:37:33 +0000 (11:37 -0500)]
sessiond: docs: document action_executor_enqueue() parameter ownership
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iab668825a7b529e4212aa39080e152628c07a363
Jonathan Rajotte [Fri, 24 Jan 2020 22:37:54 +0000 (17:37 -0500)]
sessiond: defer tracer notification jobs to the action executor
Create an event rule evaluation and enqueue a job to execute the
actions associated with the corresponding trigger on the action executor
work queue.
Clients are notified that a notification was dropped if the action
executor queue is full at the time of invocation.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7b36af2553a4ca9404b58f185e9a12f3900a1d35
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jérémie Galarneau [Thu, 11 Feb 2021 15:28:54 +0000 (10:28 -0500)]
Fix: consumer: timer: uninitialized sigevent fields
Coverity reports:
1445769 Uninitialized scalar variable
The variable will contain an arbitrary value left from earlier computations.
In consumer_channel_timer_start: Use of an uninitialized variable (CWE-457)
CID
1445769 (#1 of 1): Uninitialized scalar variable (UNINIT)7.
uninit_use_in_call: Using uninitialized value sev. Field sev._sigev_un
is uninitialized when calling timer_create.
Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: If9154d8b20699a88bf549c407eaef898dfebcd51
Jérémie Galarneau [Wed, 10 Feb 2021 22:24:21 +0000 (17:24 -0500)]
Fix: sessiond: notification: unreachable code in switch-case
A copy paste error introduced in
bc8daafb causes the initialization
of the session credentials to not be initialized.
Reported-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I50af722a498892bec764e115cfa6e9ca64302e05
Jérémie Galarneau [Wed, 10 Feb 2021 19:33:11 +0000 (14:33 -0500)]
Fix: sessiond: client: leak of trigger object on invalid client message
Coverity reports:
1445927 Resource leak
The system resource will not be reclaimed and reused, reducing the future availability of the resource.
In receive_lttng_trigger: Leak of memory or pointers to system resources (CWE-404)
lttng_trigger_create_from_payload() can return a trigger (positive
return value) that doesn't match the advertised trigger size. In that
case, a trigger reference was still returned and must be released.
Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I5afac029ddcf32dd0a01064c9dc688ed9208f278
Jérémie Galarneau [Wed, 10 Feb 2021 19:25:57 +0000 (14:25 -0500)]
Fix: sessiond: timer: unitiliazed sigevent fields
Coverity reports:
1445773 Uninitialized scalar variable
The variable will contain an arbitrary value left from earlier computations.
In timer_start: Use of an uninitialized variable (CWE-457)
CID
1445773 (#1 of 1): Uninitialized scalar variable (UNINIT)2.
uninit_use_in_call: Using uninitialized value sev. Field sev._sigev_un
is uninitialized when calling timer_create
Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I306ea7e58192223408fac9d354dbbfebe6c0cc70
Jérémie Galarneau [Wed, 10 Feb 2021 19:20:34 +0000 (14:20 -0500)]
Fix: sessiond: ust-registry: dereference of NULL pointer on allocation failure
Coverity reports:
1445771 Dereference after null check
Either the check against null is unnecessary, or there may be a null pointer dereference.
In ust_registry_channel_add: Pointer is checked against null but then dereferenced anyway (CWE-476)
Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3976e0d7b059b47ca233bc9a20585eff2e1290fc
Jérémie Galarneau [Wed, 10 Feb 2021 19:08:55 +0000 (14:08 -0500)]
Fix: sessiond: kernel: leak of notifier rule on failure to init notifier
Coverity reports:
1445768 Resource leak
The system resource will not be reclaimed and reused, reducing the future availability of the resource.
In kernel_create_event_notifier_rule: Leak of memory or pointers to system resources (CWE-404)
Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ica3f414fb7640ecddf9affbcc359121b04a82277
Jérémie Galarneau [Wed, 10 Feb 2021 19:04:09 +0000 (14:04 -0500)]
lttng-ctl: health: remove unreachable condition
Coverity reports:
1442789 Logically dead code
The indicated dead code may have performed some action; that action will never occur.
In lttng_health_create_relayd: Code can never be reached because of a logical contradiction (CWE-561)
`path` is already checked for NULL before reaching the lttng_strncpy
call.
Reported-by: Coverity Scan
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ifce95c37a6fe7f914a3bbc5443171c511d3c211a
Jérémie Galarneau [Wed, 10 Feb 2021 18:59:48 +0000 (13:59 -0500)]
Fix: sessiond: uninitialized credentials set for unbound trigger
The bound-object credentials should be left unset when the trigger
doesn't have to bind to an object.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I38f75b7c62ed796ff72d4b0692bcb51b361aa81b
Jérémie Galarneau [Tue, 9 Feb 2021 22:09:37 +0000 (17:09 -0500)]
buffer-view: improve logging on creation failure
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iddb902d6a02e69a3d2f02598c99a4a818bb4540c
Jérémie Galarneau [Tue, 9 Feb 2021 22:08:45 +0000 (17:08 -0500)]
Fix: sessiond: leak of trigger on registration error
The caller of cmd_register_trigger assumes that no triggers are returned
on error. This causes a leak of the reference to the returned trigger.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7377f9d4227c85feadba1778cf3c910ac677adac
Jérémie Galarneau [Tue, 9 Feb 2021 21:10:18 +0000 (16:10 -0500)]
sessiond: kernel: misleading log message on event rule creation failure
A logging message referring to "triggers" is emitted when failing to
create an event notifier rule in the kernel domain.
The name trigger was left-over since the terminology used to refer to
those internal objects changed.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I6d655093b83110a97cef21d7e917572a35a8c8d7
Jérémie Galarneau [Tue, 9 Feb 2021 21:04:05 +0000 (16:04 -0500)]
Fix: sessiond: trace-kernel: function return code overwritten
A string copy, performed using `lttng_strncpy` is performed at the end
of trace_kernel_init_event_notifier_from_event_rule() and its return
value is stored in the function's `ret` variable, which is of type `enum
lttng_error_code`.
Since LTTNG_OK is not == 0, the return value is interpreted as an error,
which causes the registration of kernel-domain on-event-hit triggers
to fail.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia0bf6eb3d2a4fca8f9089ecd4fef7a311354560c
Jérémie Galarneau [Tue, 9 Feb 2021 23:36:16 +0000 (18:36 -0500)]
sessiond: convert invalid command return codes to LTTNG_ERR_UNK
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ie68b4f11b28e8cbbae2fc3e4593d468572a8dce1
Jérémie Galarneau [Tue, 9 Feb 2021 23:36:30 +0000 (18:36 -0500)]
lttng-ctl: convert invalid command return codes to LTTNG_ERR_UNK
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I6dd1ff37ea6f889fb6c70282fd774029f80cfc2b
Jérémie Galarneau [Tue, 9 Feb 2021 00:01:15 +0000 (19:01 -0500)]
Clean-up: ust-consumer: replace manual metadata cache buffer allocation
Replace the hand-rolled buffer allocation management code of the
metadata cache by the lttng_dynamic_buffer utility.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I793243f2e8f5f17a4a6c0abdbafb7295ae90cf60
Jérémie Galarneau [Mon, 8 Feb 2021 23:04:20 +0000 (18:04 -0500)]
Clean-up: ust-consumer: pass metadata cache to its write method
The consumer_metadata_cache_write function doesn't need access to the
metadata channel; it can operate on the metadata cache itself and
doesn't need to access its channel.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I4350c034bbd4d89808e4153b21e7ff8a9ca04b2b
Jérémie Galarneau [Mon, 8 Feb 2021 19:40:33 +0000 (14:40 -0500)]
Fix: ust-consumer: metadata thread not woken-up after version change
Issue observed
==============
The metadata regeneration test fails, very rarely, in the "streaming"
case on the CI. The interesting part of the test boils down to:
1) start session
2) launch an app tracing one event
3) stop session
4) delete metadata file
5) start session
6) regenerate metadata
7) stop session
8) destroy session
9) read trace: babeltrace fails on an invalid metadata file.
The problem is hard to capture, but modifying the test allows us to see
that there appears to be a short window between steps 7 and 8 where the
metadata file is empty or doesn't exist.
Cause
=====
When metadata is regenerated, its version is bumped and the metadata
cache is "reset". In some cases, such as in this test, the new metadata
will have exactly the same size as it had prior as nothing happened to
change that (e.g. no new apps/probes were registered).
When this occurs, the metadata thread is not woken-up by
consumer_metadata_cache_write() as it sees that max_offset of the
metadata cache didn't change; the data was replaced but it has the same
size.
The metadata consumption thread also checks for version bumps and
resets the amount of consumed metadata. Hence, if the "cache write"
operation woke up the metadata consumption thread, the stream's
"ust metadata pushed" state would be reset and the new contents would
be consumed.
Solution
========
The metadata stream's "ust metadata pushed" position is directly reset
to zero when a metadata version change is detected by the metadata
cache. The metadata poll thread is also woken up to resume the
consumption of the newly-available data.
It is unclear why the change to the consumption position was only done
on the metadata consumption thread's code path and not directly by the
session daemon command handling.
Note that a session rotation will also result in a reset of the pushed
position and a wake-up of the metadata poll thread from the command
handling thread. I am speculating that this couldn't be done due to the
design of the locking at the time of the original
implementation (I haven't checked).
In implementing this change, the metadata reception code path is
untangled a bit to separate the logic that affects the metadata stream
from the logic that manages the metadata cache. I suspect the original
error stems from a mix-up/confusion between both concerns.
When a metadata version change happens, the metadata cache resets its
'max_offset' (in other words, it's current size) and notifies the
caller. The caller then resets the "ust pushed metadata" position to
zero and wakes-up the metadata thread to consume the new contents of the
metadata cache.
Known drawbacks
===============
None.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I142ef957140d497ac7fc4294ca65a55c12518598
Jérémie Galarneau [Mon, 8 Feb 2021 18:03:20 +0000 (13:03 -0500)]
Fix: ust-consumer: metadata cache lock not taken when sampling max offset
Found by inspecting the code while searching for an unrelated problem.
The 'max_offset' field is probably only accessed by the sessiond
poll thread, but this isn't a documented (nor reasonably maintainable)
guarantee.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I1d2b29e9a483aa8e49c2db589284b853cce65491
Jérémie Galarneau [Mon, 8 Feb 2021 18:00:56 +0000 (13:00 -0500)]
Clean-up: ust-consumer: simplify metadata cache unlock on error path
The metadata cache lock can be released directly after the call to
consumer_metadata_cache_write() since nothing depends on the cache
being locked after.
This simplifies the unlocking of the metadata cache by removing an
extra "unlock" that is specific to an error path.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ib0151d9640309005dfc83e95efb538e3c4f0df4a
Jérémie Galarneau [Thu, 4 Feb 2021 23:18:32 +0000 (18:18 -0500)]
Fix: sessiond: acquire session list lock when updating event notifiers
Registering triggers with an on-event hit condition affects event
notifiers, imposing a synchronization of enablers with the user space
tracers.
As noted in the comments of session.h, the session list lock protects
those updates and is, ultimately, ill-named. The comment is adjusted to
mention "tracer configurations" rather than "session configurations"
since event notifiers are not part of a session, making the comment
imprecise.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id1bcbcccbdeeafa91176ed3413ddddbcbab10ad2
Jonathan Rajotte [Tue, 11 Feb 2020 20:59:02 +0000 (15:59 -0500)]
sessiond: agent: enable events matching event notifiers
Per event notifier domain agents
--------------------------------
A `struct agent` instance encapsulates the agent events of an agent
domain for a given session. In the context of event notifiers, there is
no session involved to scope the agent instance.
Hence, per-domain instances are maintained to control event-notifiers
enabled by triggers.
Agent event enable count
------------------------
Agents act as a pre-filter on user space tracer events. In order to
honor triggers use event notifiers (on-event-hit conditions), the
register/unregister trigger commands are modified to ensure 'agent'
domain events are created and enabled/disabled suitably for the user
space tracer to be invoked.
Note that since agent events are a "filter" before event enablers, an
event-rule targetting a ring buffer and an identical event-rule
targetting an event notifier can be enabled. The action to take when the
event is hit is completely opaque to the agents. In such cases, the same
agent event instance needs to be enabled and the current implementation
doesn't allow duplicate agent events.
Hence, `struct agent_event`'s enabled state is now a counter which
accounts for all enabled event-rules that require this agent event to be
enabled. The agent event is enabled when one or more event rules
matching it are enabled and it is disabled when that count reached zero.
To ensure no code checks for the agent_event's enabled state by
comparing to '1', a new `AGENT_EVENT_IS_ENABLED` macro is introduced.
The existing code using the `enabled` attribute directly is modified to
use it.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I363db1e4bd7d7d73c75b8576c6323ee41e31aa00
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jérémie Galarneau [Wed, 3 Feb 2021 22:00:06 +0000 (17:00 -0500)]
Fix: sessiond: assertion fails when getting name of trigger
Left-over debug code assumes that a trigger is named and logs
registration/unregistration failures as errors, resulting in a spammy
sessiond output.
Reduce the logging verbodity of those statements to DBG and handle
unnamed triggers.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I74a90150ec2b0914784fb187517025a9a3019cfb
Francis Deslauriers [Thu, 22 Oct 2020 15:32:15 +0000 (11:32 -0400)]
sessiond: client commands: print command enum as a string
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I10ce8911372a6926b7aa234b9d1595a5052d4a0e
Francis Deslauriers [Mon, 14 Dec 2020 22:37:42 +0000 (17:37 -0500)]
Cleanup: add `notification_command_type_str()`
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7a05c8055f2bb2e09e274a7f4d8f9ecd3e5509fc
Francis Deslauriers [Wed, 13 Jan 2021 17:04:49 +0000 (12:04 -0500)]
Fix: lttng-sessiond: Taking sizeof of a pointer
Reported by clang-tidy:
bugprone-sizeof-expression: Suspicious usage of `sizeof(A*)`
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ie3069ad3d81d460646c71923f0f396cf18980540
Mathieu Desnoyers [Wed, 3 Feb 2021 15:26:03 +0000 (10:26 -0500)]
Use MT-safe strtok_r in spawn viewer library helper
Even though its current callers are all in single-threaded context
(lttng view and lttng-crash commands), it is a good practice to use
MT-safe APIs in library functions, in case those are ever used within a
multithreaded context in the future.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: If419fd24b3bf6532f2c75a728a9ec395e365c626
Mathieu Desnoyers [Wed, 3 Feb 2021 15:21:42 +0000 (10:21 -0500)]
Fix: use MT-safe strtok_r in multithreaded context
append_list_to_probes uses the non-multithread-safe "strtok" while in
multithreaded context.
It is called by modprobe_lttng_data(), which is called from
init_kernel_tracer().
init_kernel_tracer is used from main() after other threads are created,
and also from process_client_msg() when a user attempts to interact with
the kernel domain if the kernel tracer has not yet been successfully
initialized.
Fixed by using the MT-safe strtok_r() instead.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ic3f81551b8508451e0f9733c5de56a4d4407b813
Jonathan Rajotte [Fri, 10 Jan 2020 20:29:52 +0000 (15:29 -0500)]
sessiond: synchronize event notifiers for UST tracers and the kernel tracer
Synchronize event notifiers when registering or unregistering a trigger
which has a condition that implies the use of an event notifier.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id2cb2a93c0c6be9f44f9dcd07b9803acbcb6b3c3
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jérémie Galarneau [Thu, 21 Jan 2021 17:30:38 +0000 (12:30 -0500)]
Fix: sessiond: ust-app: account for the event notification pipes fds
The file descriptors of the event notifier group's event pipe are not
accounted by the lttng_fd_*() module, which tracks the use the file
descriptors by the userspace tracer.
Add the necessary lttng_fd_*() uses to account for the pipe's file
descriptors over their lifetime.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ib24b4c2934655573965f06fa7998859335053083
Jérémie Galarneau [Thu, 21 Jan 2021 17:28:59 +0000 (12:28 -0500)]
sessiond: ust-app: close the event notifier group's pipe write-end
The write-end of the event notifier group's pipe can be closed after
it is passed to the application. This reduces the amount of file
descriptors used by the session daemon on a per-application basis.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ic8f016bc4cdd3dce2d7ea8ac9cccb53e3b69c5b0
Jonathan Rajotte [Tue, 31 Mar 2020 18:33:18 +0000 (14:33 -0400)]
sessiond: notification: receive incoming notifications from tracers
Receive and discard notifications received by tracers. This simply puts
in place the basic tracer notification reception mechanisms.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3d94431051015c69785ec256095135746e3972ce
Jérémie Galarneau [Tue, 26 Jan 2021 23:15:15 +0000 (18:15 -0500)]
sessiond: unregister triggers during clean-up on shutdown
Triggers should be unregistered on sessiond clean-up. This allows the
kernel and user space tracer control code to assume that all event
notifiers have already been destroyed on tear-down and that any
remaining event notifier is the result of a leak.
Moreover, it simplifies the tear down as it is simply handled in the
way a regular trigger unregistration would be.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ie821c5162c2fbe9c3d6705f464224c50ae9df1c8
Jérémie Galarneau [Tue, 26 Jan 2021 23:13:11 +0000 (18:13 -0500)]
sessiond: trigger: decouple reception of triggers from commands
The trigger to register/unregister is received in the body of the
register and unregister commands.
Isolate the reception of the trigger from the command payload in
a util in client.c (receive_lttng_trigger()) which allows the commands
to only operate on fully-formed objects.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iaee6c54590e259ce053d368494ff0c8b356a36a1
Jérémie Galarneau [Tue, 26 Jan 2021 21:14:33 +0000 (16:14 -0500)]
sessiond: clean-up: trigger to unregister can be 'const'
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ic2a731b36467f898402ff5e19e329e5734e2da69
Jérémie Galarneau [Thu, 28 Jan 2021 19:39:37 +0000 (14:39 -0500)]
Fix: sessiond: double free on duplicate removal of tracer source
An unrelated bug (fixed in a separate commit) can cause an event source
to be removed from the notification thread's monitored sources twice.
The event source removal starts by searching for the source to remove
based on the source pipe's read-end fd number and assumes that it will
always be found. After iterating on the list, an assertion that
`source_element` is not NULL is done in the assumption that NULL would
mean that the source was not found.
This is incorrect since, if the source is not found, `source_element`
will simply point to the last element of the list, causing the assertion
to succeed.
Then, the last source in the list is torn down, but not removed from the
list. This causes that event source to be free'd twice when it is
actually removed later on.
The assumption that an event source can always be found does not hold
for the moment. For instance, when an application can exit, closing its
end of the notification pipe, the notification thread could wake-up
before the application management thread.
In that case, the notification thread will react to the event by
removing the application's source from its monitored sources. Then, when
the application management thread wakes up, it will ask the notification
thread to (again) remove the event source, which will fail as it will
not be found.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7b5ebf90b868faded47a4e9675e01e1fb2b77a70
Jonathan Rajotte [Mon, 23 Mar 2020 21:09:18 +0000 (17:09 -0400)]
sessiond: kernel triggers: add infrastructure to create event notifiers
Add the infrastructure to initialize the kernel tracer event notifier
group and individual event notifiers from event rules issued from
triggers.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I20127f655018260e45566d09d11c2852cd3b3f97
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jonathan Rajotte [Mon, 23 Mar 2020 21:03:37 +0000 (17:03 -0400)]
kernel: event notifier: kernel-ctl interface
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Idd3983c594ffecff0a20b71b7cd3d297e77446a1
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jonathan Rajotte [Mon, 3 Feb 2020 19:03:25 +0000 (14:03 -0500)]
kernel: load lttng-ring-buffer-event-notifier-client module
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I9449e895b1eb88eb6db47dc3f8eb2864a2eb816d
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jérémie Galarneau [Tue, 19 Jan 2021 19:56:49 +0000 (14:56 -0500)]
sessiond: kernel: make modules required/optional property per-module
Modules are considered required or optional based on their
category (control or data probes). Make the load policy per-probe since
optional control probes will be introduced in a follow-up change.
No change in behaviour is intended by this change.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I0048b60bee3969d2fa2b9ed94b6fb24d3b5ae659
Francis Deslauriers [Mon, 14 Dec 2020 22:30:12 +0000 (17:30 -0500)]
Fix: add rcu_barrier() after sessiond_cleanup()
This is to ensure that tracer event source (event notifier socket) are
removed from the notification thread list.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2ca8f72c023132c341193bf626c0dac20b89e1f2
Jonathan Rajotte [Mon, 13 Jan 2020 18:59:39 +0000 (13:59 -0500)]
ust-app: implement event notifier support
Event notifier support mostly resemble how it is done for regular event.
We end up implementing ust_app_synchronize_event_notifier_rules which is
used in a similar fashion to ust_app_synchronize minus the dependency on
a ltt_ust_session session.
The lttng_event_rule_generate_bytecode interface is modified to return a
status code since it could fail (return NULL) for reasons other than not
having exclusions (e.g. an allocation failure).
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2cde2b3d2530e2114bff99b1b26ac6d83f575ad9
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jérémie Galarneau [Tue, 12 Jan 2021 22:41:54 +0000 (17:41 -0500)]
Fix: liblttng-ctl: unreported truncations when copying strings
gcc 10.2 reports a large number of string truncation warning in
liblttng-ctl. Replace the uses of lttng_ctl_copy_string() util by
lttng_strncpy() (handling the null source case when applicable) and
report the truncations when they occur.
Example gcc warning:
lttng-ctl.c:86:3: warning: ‘strncpy’ output may be truncated copying 254 bytes from a string of length 254 [-Wstringop-truncation]
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Icca5f4c2490c6796b451999d7694db8597bae719
Jérémie Galarneau [Tue, 12 Jan 2021 22:08:56 +0000 (17:08 -0500)]
Fix: sessiond: event name truncation during listing
The use of strncpy can lead to silently-truncated event names. Replace
its use by the internal lttng_strncpy which fails on truncation.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I65d6bd46208dc7b62a83e4600a52a6669fd99d55
Jérémie Galarneau [Tue, 12 Jan 2021 20:36:00 +0000 (15:36 -0500)]
Clean-up: replace erroneous of empty parameter list by void
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I951feeed92b346e79e34bec45a14f8b226283ae4
Jonathan Rajotte [Fri, 10 Jan 2020 21:05:48 +0000 (16:05 -0500)]
sessiond: setup event notifier group for registering app
Create a pipe for each application and setup an event notifier group
associated with that pipe. Transfer the write side to the app, and
transfer the read side to the notification thread as an application
event source ([...]_ADD_TRACER_EVENT_SOURCE)
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3e4aab84e3270ddef1f50f72f946a4d80b3f36e0
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jérémie Galarneau [Mon, 11 Jan 2021 23:11:46 +0000 (18:11 -0500)]
Fix: configure: support Autoconf 2.70
The newly-released autoconf 2.70 introduces a number of breaking
changes [1] and is being rolled-out by some distros.
Amongst those changes, the AC_PROG_CC_STDC macro is marked as obsolete
and was merged into AC_PROG_CC, which we already use. On 2.70, this
results in a warning which we handle as an error.
A version check is added to invoke the AC_PROG_CC_STDC macro only when
running a pre-2.70 version of autoconf, fixing the issue.
A single use of the AC_HELP_STRING macro is replaced by AS_HELP_STRING
as the former was marked as obsolete.
The AC_PROG_LEX now takes an argument, and the argument-less version is
marked as obsolete. The macro is invoked with the `noyywrap` option, as
recommended in the documentation.
Also, the AX_PTHREAD macro makes use of the $as_echo built-in shell
variable which no longer exists in 2.70. A patch was submitted to the
GNU Autoconf archive in March, but there have been no signs of life
given since then [2].
As such, our local copy is updated to the latest version and the patch
(which looks fairly straight-forward / safe) is applied. This should
minimize changes once we go back to an "official" version of the macro.
[1] https://lwn.net/Articles/839395/
[2] https://savannah.gnu.org/patch/?9906
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-by: Michael Jeanson <michael.jeanson@efficios.com>
Change-Id: Ie949de73442770f60cbef55300265205527731c6
Michael Jeanson [Wed, 16 Dec 2020 17:40:53 +0000 (12:40 -0500)]
Fix: different pthread_getname_np signature() on macOS causes build failure
macOS likes to be special so it has pthread_setname_np() without a
thread id parameter, but a pthread_getname_np() with it. Split the
detection macro in two and modifiy the compat layer to handle it.
Change-Id: I8034c54057d68eef59546960c75afe8fbe07f5ad
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Francis Deslauriers [Fri, 18 Dec 2020 22:00:40 +0000 (17:00 -0500)]
lttng-ust abi: sync _UST_CMD() values
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ieacee6ecab41855cebae7113e7c512d4d684eb98
Francis Deslauriers [Fri, 4 Dec 2020 18:47:32 +0000 (13:47 -0500)]
hashtable: silence -fsanitize=address warning for `hashlittle()` function
Issue
=====
The code of this function triggers the following heap-buffer-overflow
warning when compiled with `-fsanitize=address` in specific situation:
==247225==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x602000001310 at pc 0x5559db6c575a bp 0x7f193e6faeb0 sp 0x7f193e6faea0
READ of size 4 at 0x602000001310 thread T4 (Notification)
#0 0x5559db6c5759 in hashlittle /home/frdeso/projets/lttng/tools/src/common/hashtable/utils.c:315
#1 0x5559db6c6df4 in hash_key_str /home/frdeso/projets/lttng/tools/src/common/hashtable/utils.c:490
#2 0x5559db5e3282 in hash_trigger_by_name_uid /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/notification-thread-events.c:378
#3 0x5559db5ecbe3 in trigger_name_taken /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/notification-thread-events.c:2333
#4 0x5559db5ecd7c in generate_trigger_name /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/notification-thread-events.c:2362
#5 0x5559db5ed6e0 in handle_notification_thread_command_register_trigger /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/notification-thread-events.c:2491
#6 0x5559db5ef967 in handle_notification_thread_command /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/notification-thread-events.c:2927
#7 0x5559db5ddbb7 in thread_notification /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/notification-thread.c:693
#8 0x5559db60e56d in launch_thread /home/frdeso/projets/lttng/tools/src/bin/lttng-sessiond/thread.c:66
#9 0x7f19456ec608 in start_thread /build/glibc-ZN95T4/glibc-2.31/nptl/pthread_create.c:477
#10 0x7f1945602292 in __clone (/lib/x86_64-linux-gnu/libc.so.6+0x122292)
Given that the `k` pointer used in this loop is a `uint32_t *` we might
read bytes outside of the allocated key if the key is less than 4 bytes
long. As the comment about Valgrind explains, this is not a real problem
because memory protections are typically word bounded.
I tried to use the `__SANITIZE_ADDRESS__` define to select the
Valgrind implementation of this code when building with AddressSanitizer
but that still triggers the same head-buffer-overflow warning.
Why wasn't that a problem before?
=======================================
The trigger feature will use small default names like "T0".
Workaround
==========
Exclude this function from the sanitizing using the compiler attribute
"no_sanitize_address".
Drawback
========
This removes our sanitizing coverage for this function.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I82d0d3539916ed889faa93871f9b700064f2c52a
Francis Deslauriers [Wed, 9 Dec 2020 03:05:22 +0000 (22:05 -0500)]
Tests: Fail test if sessiond is not running when it should
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I9b39bfe6bfb9f404fe2a32c27de0276386a36212
Francis Deslauriers [Thu, 10 Dec 2020 20:42:22 +0000 (15:42 -0500)]
Cleanup: erroneous use of CDS_INIT_LIST_HEAD() on node
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I6caf957af4d3e325e9f2086441d0552d64a77db5
Jonathan Rajotte [Fri, 10 Jan 2020 22:03:05 +0000 (17:03 -0500)]
UST: update ABI for event notifier
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia3088ebdf0fe64e57e93c2bec02625176460ffc9
Francis Deslauriers [Wed, 22 Jan 2020 16:15:10 +0000 (11:15 -0500)]
userspace-probe: Decouple `userspace_probe_add_callsite()` from event and session
Currently this function takes event and session pointers:
- The event is used to get the location type of the probe,
- the session is used to get the uid and gid of the user to used them
with the `run_as_*()` functions.
With the incoming trigger support, we want to reuse this function to add
trigger userspace-probe callsites.
This commit extracts what will be common in both event and trigger
implementations by creating a specialized
`userspace_probe_event_add_callsite()` function that uses a generalized
`userspace_probe_add_callsite()`.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia3b90050a7bd227a30af0c11395dcdf5aca13583
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479
Jonathan Rajotte [Mon, 23 Mar 2020 15:56:05 +0000 (11:56 -0400)]
Generalize disable_ust_event to support multiple types of ust object
This will allow us to pass a trigger object later.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2d2632c80c9bcb2d2ca6966e080d0fa5d3422796
Jonathan Rajotte [Fri, 10 Jan 2020 19:47:30 +0000 (14:47 -0500)]
Generalize enable_ust_event to support multiple types of ust object
This will allow us to pass a trigger object later.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I594c3eb0437345406e3a24fcbfbf4b7a8162908b
Jonathan Rajotte [Mon, 16 Dec 2019 21:01:03 +0000 (16:01 -0500)]
Generalize set_ust_event_exclusion to support multiple types of ust object
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ibe8f6b2c459afc698971b23b3f1a72d0a45e036f
Jonathan Rajotte [Mon, 16 Dec 2019 20:48:40 +0000 (15:48 -0500)]
Generalize set_ust_event_filter to support multiple types of ust object
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iab5bc94b8895b6470c4c3339a691f71c5e9a2e3c
Jérémie Galarneau [Fri, 18 Dec 2020 21:15:18 +0000 (16:15 -0500)]
notification: mark tracer source element as out of poll set
Mark the tracer source element as being out of the notification thread's
poll set once it has been removed. This has no effect right now, but it
is less error-prone considering future changes to this function.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia4d08dabd6b07ec455fe3120b7188e414232536e
Jonathan Rajotte [Mon, 13 Jan 2020 18:52:51 +0000 (13:52 -0500)]
Introduce trigger hash table with tracer token as key
This will allow easy lookup on reception of the tracer token coming
from the tracer.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iee42539f0a664ead5ca03534549c6bbd5e505953
Jonathan Rajotte [Wed, 25 Mar 2020 22:49:32 +0000 (18:49 -0400)]
notification: add/remove tracer event source
The notification thread will be responsible of consuming the tracer
notification event coming from the UST tracers and kernel tracer.
On a 'add' operation, the tracer event source (i.e read side of a pipe)
is added to the notification poll set. Book-keeping is also done via a
list for later lookup.
On 'remove', the event source is removed from the pollset and from the
list.
On cleanup (notification_thread_handle_destroy), it is expected that all
added tracer event sources be removed by their respective "adder". No
bulk cleanup is performed.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I23679922a58849c9bc86f30b2aae17b39fa2e222
Jonathan Rajotte [Mon, 17 Aug 2020 22:24:35 +0000 (18:24 -0400)]
DBG: add debug statement for trigger not bound to any object
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I2c21c8c702017daf38648bcb835e711335a8fd77
Francis Deslauriers [Wed, 9 Sep 2020 21:36:12 +0000 (17:36 -0400)]
sessiond: Extract condition hashing functions
Extract these functions so it can be used by other files.
The lttng_condition hashing code is kept in this (rather than
common/condition/condition.c) since it makes use of GPLv2 code
(hashtable utils), which we don't want to link in liblttng-ctl.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iaafe1402b2d198a00920d939502004038e78fff0
Francis Deslauriers [Wed, 9 Dec 2020 13:43:58 +0000 (08:43 -0500)]
Cleanup: misplaced white space in `ERR()` statement
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I44291907c9973394c3edaf9b470230c59bb75eec
Jonathan Rajotte [Mon, 17 Aug 2020 22:23:27 +0000 (18:23 -0400)]
Add base support for event rule hit
Add some of the scafolding to support event-rule hit conditions.
This includes the hashing of event rule conditions and, consequently,
of event rules and the various probe location types.
The kernel module ABI is checked to verity that the kernel tracer
supports event notifiers.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iab4db4fc7e9f0c5a7206106fa6a4781b6b95d306
Jérémie Galarneau [Thu, 26 Nov 2020 20:36:03 +0000 (15:36 -0500)]
sessiond: return 'invalid protocol' error on reception error
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I15758543ae51dd2ce30b40d88a05ef0492ce0e51
Jonathan Rajotte [Mon, 17 Aug 2020 22:19:47 +0000 (18:19 -0400)]
Only perform notification related unregistering when action is notify
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id89eaf1c8f550e20adbd0b0d82462f6bf0b8ba21
Jonathan Rajotte [Mon, 13 Jan 2020 18:40:12 +0000 (13:40 -0500)]
Use lttng_trigger_is_equal when iterating over the trigger ht
Since a trigger can now have other type of actions then the notify one,
we must account for it. We use lttng_trigger_equal to perform that task.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3300b0fff66e760152c1f7065d8fbfb945cce48e
Jonathan Rajotte [Mon, 23 Mar 2020 21:26:47 +0000 (17:26 -0400)]
Generate bytecodes related to the trigger on reception
The compositing objects of a trigger might need to generate internal
bytecode. Doing it at the registration step allows an early validation
of the filter expressions.
There is no need to generate it for the unregister command since
bytecodes are not used for comparison and are for internal use only.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia1282d55f028e6b056e8ff3877790894c582acdb
Francis Deslauriers [Fri, 13 Nov 2020 21:27:59 +0000 (16:27 -0500)]
kernel: Add token field to `struct lttng_kernel_event`
This field will be used by event notifier and counters features.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I48d71a240150355d80b5a325717ca695467c5451
Francis Deslauriers [Wed, 25 Nov 2020 17:06:38 +0000 (12:06 -0500)]
kernel: Add `struct lttng_kernel_syscall` to ABI
This struct is now used by the kernel tracer to allow to selectively
turn on and off syscalls event firing.
This way, the sessiond can decide to turn on only syscall entries,
exits, or both.
This will be used by the upcoming event notifier features to only
generate a notification on syscall entry.
This new struct doesn't change the layout of the `lttng_kernel_event`
structure.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I478de30b91b415f517e9d0ac0686f3130f79d86b
Anders Wallin via lttng-dev [Wed, 25 Nov 2020 08:31:40 +0000 (09:31 +0100)]
Extras: Perl 5.26 requires { to be escaped by \
Unescaped literal "{" characters in regular expression patterns are no
longer permissible
Signed-off-by: Anders Wallin <wallinux@gmail.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Jérémie Galarneau [Tue, 1 Dec 2020 21:51:23 +0000 (16:51 -0500)]
Fix: sessiond: metadata not created on app unregistration during start
Issue observed
==============
A test for an incoming feature (trigger actions on on-event conditions)
hangs. While this problem was discovered using this test, it exercises a
scenario that is problematic as of this fix.
The destruction of a session can hang if a single application being
traced unregisters (dies) during the 'start' of a session.
Cause
=====
When a per-uid session is started, its buffers (channels and streams)
are allocated only if an instrumented application is registered to the
session daemon at that moment.
For historical reasons, the 'data' and 'metadata' buffers are allocated
in separate code paths. The 'data' buffers are allocated in
ust_app_synchronize() and the 'metadata' buffers are allocated in
ust_app_start_trace(). Both functions perform their own look-up for an
application session and will gracefully fail if an application session
can't be found; it typically means the application has exited.
This leaves a race window open where ust_app_synchronize() can succeed
in looking-up the application session, and ust_app_start_trace() can
fail following the death of the application.
When this occurs, the session is left with 'data' buffers allocated and
unallocated ''metadata' buffers. This is an unexpected state and results
in the rotation code attempting to rotate a partially initialized
metadata stream.
The rotation of this partially initialized metadata stream never
completes which, in turn, never allows the session to complete its
implicit rotation on destruction.
This race window is fairly narrow, but can be reproduced by sleep()-ing
at the beginning of ust_app_start_trace() and killing an application
that is being traced during the sleep period.
Solution
========
The creation of the metadata channel is performed as part of
ust_app_synchronize() if the application look-up succeeds. When it
fails, both 'data' and 'metadata' streams will fail to be created
resulting in an expected and valid state.
Known drawbacks
===============
None.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ice0ec16734a39a6bb885986d3ad70d20cd2618e0
Jonathan Rajotte [Mon, 30 Nov 2020 18:54:14 +0000 (13:54 -0500)]
test: utils: lttng_pgrep performs lookup on non-existing pid
Observed issue
==============
# Killing (signal SIGTERM) lttng-sessiond and lt-lttng-sessiond pids: 20962 20963
./tests/regression/tools/trigger/start-stop//../../../../utils/utils.sh: line 103: /proc/20963/cmdline: No such file or directory
Cause
=====
lttng_pgrep performs a two step search/validation for the pattern. Since
lttng_pgrep is used during tear-down of process (staged termination
signalling) a process returned by pgrep might exit before the second
check.
Solution
========
Simply silence the error. The code flow already acknowledges the
possibility of failure here.
Known drawbacks
=========
None
References
==========
Fixes: #1292
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I82cb9fd4754c10a5104af495a8a959f4fbd92664
Francis Deslauriers [Mon, 30 Nov 2020 19:54:18 +0000 (14:54 -0500)]
Fix: missing `_mutex_lock()` before signaling a condition variable
According to the PTHREAD_COND(3) man page, a condition variable
signaling and broadcast should alway be protected with a mutex.
This commit fixes two calls to `pthread_cond_signal()` function without
holding the right lock.
This commit also adds an assertion right before two calls to
`pthread_cond_broadcast()` where it's less obvious from the surrounding
code that the mutex is held. This documents the code and may be useful
for future debugging.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Iebf5a8b2e4251bd1ff4cd462e548cd3486c6cb75
Francis Deslauriers [Tue, 15 Sep 2020 16:10:18 +0000 (12:10 -0400)]
Cleanup: use `modprobe --remove` rather than `rmmod`
Background
==========
According to the rmmod(8) man page:
rmmod is a trivial program to remove a module (when module unloading
support is provided) from the kernel. Most users will want to use
modprobe(8) with the -r option instead.
`rmmod` simply unloads the provided module and decrements the refcount
of the modules it depended on but doesn't unload those dependencies if
their refcount is zero.
Issue
=====
With the following scenario we can end up if modules with a zero
refcount still loaded in the kernel:
modprobe lttng-test
lttng-sessiond
... (test case) ...
ctrl+c sessiond
rmmod lttng-test
When we teardown the lttng-sessiond, some modules are kept in the kernel
because the `lttng-test` module depends on them. So unloading
`lttng-test` using `rmmod` keeps those dependencies in the kernel.
Solution
========
Use `modprobe --remove` to unload modules and their now unused
dependencies.
From the modprobe(8) man page:
-r, --remove
This option causes modprobe to remove rather than insert a module.
If the modules it depends on are also unused, modprobe will try to
remove them too. Unlike insertion, more than one module can be
specified on the command line
Note
====
This commit also replaces existing uses of `modprobe -r` to `modprobe
--remove` for consistency.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I7be83a645097e1eddd478cfbb717906b971f04ea
Jonathan Rajotte [Mon, 10 Feb 2020 01:33:55 +0000 (20:33 -0500)]
trigger: consider domain on register and unregister
This allows the sessiond to inform the client if a trigger that requires a
particular domain (event rule based condition, for example) is at all
valid.
This is useful to fail early when a trigger being registered requires an
unavailable tracer.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I660937e64b294f6239ba15faeef705438a93a41a
Jonathan Rajotte [Wed, 25 Mar 2020 14:41:17 +0000 (10:41 -0400)]
trigger: lttng_trigger_get_underlying_domain_type_restriction
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I5fe156a09e4e4c833f84a0fe9027c838b73fe728
Jonathan Rajotte [Thu, 24 Sep 2020 19:36:43 +0000 (15:36 -0400)]
action-executor: missing include of internal event-rule header
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: If541bb203f1d851750ee485fe9bd1a12d9963774
Jonathan Rajotte [Wed, 4 Dec 2019 19:30:38 +0000 (14:30 -0500)]
Tests: unit: lttng_condition_event_rule
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I281df3b2267d6ddf3b0591d181b7f276802d8963
Jonathan Rajotte [Tue, 3 Dec 2019 20:57:08 +0000 (15:57 -0500)]
condition: implement event rule based condition
An event rule condition is met when a tracer hit an event matching the
associated event rule.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I550903c231d83cb3852e8ef8aee2abafe9069b10
Jonathan Rajotte [Tue, 3 Dec 2019 21:07:34 +0000 (16:07 -0500)]
Move conditions source files to src/common/conditions directory
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I43165eacf82a1bf537e7187313664e32ca2833a9
Jonathan Rajotte [Thu, 23 Jan 2020 19:13:11 +0000 (14:13 -0500)]
trigger: implement listing of registered trigger
Each client have visibility over triggers matching its user id (uid).
The root user have visibility over all registered triggers.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3e5ae75939214ed85c376bea12f1e4b307d78976
Jonathan Rajotte [Tue, 4 Feb 2020 20:14:34 +0000 (15:14 -0500)]
Apply policy on channel sampling
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Id755b73c1f976a5a7d7a188656a1de21bd703143
Jonathan Rajotte [Thu, 23 Jan 2020 19:14:14 +0000 (14:14 -0500)]
trigger: introduce firing policies
A firing policy controls the rate of firing of a trigger.
Two firing policy mode are implemented:
LTTNG_TRIGGER_FIRING_POLICY_FIRE_EVERY_N
The triggers's actions are executed every N times the
condition occurs.
LTTNG_TRIGGER_FIRING_POLICY_ONCE_AFTER_N
The triggers's actions are executed once the condition was met N
times.
Firing policies will be moved to the specific `action` objects
in a follow-up commit as not all actions can implement the firing
policies.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ifaeeaaec7b6f2bed57d0d5f4ed8546762ec02e8d
Francis Deslauriers [Mon, 16 Nov 2020 21:50:41 +0000 (16:50 -0500)]
Fix: lttng-ctl: deserialize on orderly shutdown of sessiond
Issue
=====
The `recv_data_sessiond()` function may return zero if the socket peer
has shutdown orderly. This happens if the session daemon is killed while
the client is blocked on the `recv_data_sessiond()` call. Currently,
when this happens, the client simply goes on to decode the uninitialized
reply buffer.
This bug was witnessed while developing the upcoming event-notifier
feature where complex objects are received from sessiond and attempts to
deserialize these objects resulted in segmentation faults.
Solution
========
Return -LTTNG_ERR_NO_SESSIOND when `recvmsg()` returns zero. This way,
the client can simply tell the user that the session daemon is no longer
available.
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ib2387526c4101e3bae706e38181bfeb25da26fa3
Jérémie Galarneau [Wed, 18 Nov 2020 22:12:45 +0000 (17:12 -0500)]
Fix: trigger: erroneous check for success of trigger creation
6808ef55e added a check for `ret == 0` to determine if a trigger
could be created from a payload. The function returns >= 0 on
success, leading to crashes when a trigger is de-serialized.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Icd769dcb04f7637aa1877436e9a5570e7f20b63b
Jérémie Galarneau [Wed, 18 Nov 2020 19:14:02 +0000 (14:14 -0500)]
Fix: trigger: leak of trigger on failure to set name
lttng_trigger_create_from_payload() leaks its newly-created
trigger when it fails to set the trigger's name. Drop
the reference to the new trigger whenever the function fails.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I9dbf91d404fd67e4b79f2af550f3768680d6d4ec
Jérémie Galarneau [Wed, 18 Nov 2020 19:04:03 +0000 (14:04 -0500)]
Clean-up: trigger: use condition and action put
Use the internal *_put() functions to discard condition and
action references rather than the public *_destroy() functions
as they may cause confusion.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Idfdfda3ea2289315408245074f7cc0de6541167a
Jérémie Galarneau [Wed, 18 Nov 2020 16:55:24 +0000 (11:55 -0500)]
Docs: payload/buffer view: validate is missing an argument description
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I3d575dcda6c3e6820e911ab3c4e28b18d29f045c
Jérémie Galarneau [Sat, 14 Nov 2020 02:39:36 +0000 (21:39 -0500)]
Fix: unchecked buffer size for communication header
A number of object de-serialization functions rely on a
fixed-size communication header to create an object from
a payload.
A large number of those functions assume that the initial
header fits in the provided buffer or payload view. Also,
the functions that do validate that the header fits do so
in different ways:
- checking the view's size,
- creating a new fixed-size view and checking the 'data' pointer.
To harmonize all of those checks, the following utils are added:
- lttng_buffer_view_is_valid()
- lttng_payload_view_is_valid()
These functions should be used whenever a fixed-size view is
created (not passing -1 as the length parameter).
The checks are added and/or harmonized to:
- create a new 'header' view,
- validate it with the corresponding *_is_valid() function,
- initialize the header pointer using the header view.
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I763946feac714ecef4fc5bd427dab2d3fe5dc1a4
Jérémie Galarneau [Mon, 16 Nov 2020 21:10:09 +0000 (16:10 -0500)]
relayd: logging of `trace chunk exists` command refers to the wrong command
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I6e2bf4eee379f4e1d42333779dfeaf8f087d8217
Jonathan Rajotte [Tue, 21 Jan 2020 19:22:37 +0000 (14:22 -0500)]
trigger: lttng_triggers: implement a container for multiple triggers
This container is exposed for the listing of triggers.
We also plan on using it internally in the sessiond for inter-thread
communication.
The current implementation is backed by a lttng_dynamic_pointer_array.
Caller of lttng_triggers_add is responsible for managing ownership via
ref-counting of the lttng_trigger object.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ib541027a6d7d856daa746de5aa49f0002bbe036f
Jonathan Rajotte [Wed, 23 Sep 2020 20:13:37 +0000 (16:13 -0400)]
action-executor: evaluated object credentials are optional
Use the is_set member instead of the LTTNG_OPTIONAL_GET_PTR macro
which asserts whenever an optional member is unset.
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ia00e4a7f5f9b8198061a742bf6bd750c721908cf
This page took 0.056307 seconds and 4 git commands to generate.