| Age | Commit message (Collapse) | Author |
|---|
|
Add a new option, -A/--aligned, that enables timerlat thread alignment
implemented on the kernel-side in commit 4245bf4dc58f ("tracing/osnoise:
Add option to align tlat threads"). The option takes an argument,
representing alignment between timerlat threads in microseconds.
The feature is modeled after the option of the same name in the
cyclictest tool.
Link: https://lore.kernel.org/r/20260527144928.2944472-1-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
Instead of using getopt_long() directly to parse the command line
arguments given to an RTLA tool, use libsubcmd's parse_options().
Utilizing libsubcmd for parsing command line arguments has several
benefits:
- A help message is automatically generated by libsubcmd from the
specification, removing the need of writing it by hand.
- Options are sorted into groups based on which part of tracing (CPU,
thread, auto-analysis, tuning, histogram) they relate to.
- Common parsing patterns for numerical and boolean values now share
code, with the target variable being stored in the option array.
To avoid duplication of the option parsing logic, RTLA-specific
macros defining struct option values are created:
- RTLA_OPT_* for options common to all tools
- OSNOISE_OPT_* and TIMERLAT_OPT_* for options specific to
osnoise/timerlat tools
- HIST_OPT_* macros for options specific to histogram-based tools.
Individual *_parse_args() functions then construct an array out of
these macros that is then passed to libsubcmd's parse_options().
All code specific to command line options parsing is moved out of the
individual tool files into a new file, cli.c, which also contains the
contents of the rtla.c file. A private header, cli_p.h, is added
alongside the public header cli.h, so that unit tests are able to test
statically declared option callbacks.
Minor changes:
- The return value of tool-level help option changes to 129, as this is
the value set by libsubcmd; this is reflected in affected test cases.
The implementation of help for command-level and tracer-level help
is set to 129 as well for consistency, and the change is reflected in
exit value documentation.
- Related to the above, {rtla,osnoise,timerlat}_usage() are marked
__noreturn and exit() is removed from after they are called for
cleaner code.
- The error messages for invalid argument for options --dma-latency and
-E/--entries were corrected, fixing off-by-one in the limits.
Note that unsetting options (using --no-<opt> syntax) is currently not
implemented for options that use custom callbacks. For --irq and
--thread, it will never be implemented, as they conflict with already
existing --no-irq and --no-thread with a different meaning.
Assisted-by: Composer:composer-1.5
Link: https://lore.kernel.org/r/20260528103254.2990068-5-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
Before, when rtla got a signal, it stopped the main trace but not the
record trace. With "--on-end trace", this can lead to
save_trace_to_file() failing to keep up, especially on a debug kernel.
Plus, it adds post-stoppage noise to the trace file.
Signed-off-by: Crystal Wood <crwood@redhat.com>
Fixes: c73cab9dbed0 ("rtla/timerlat_hist: Stop timerlat tracer on signal")
Fixes: a4dfce7559d7 ("rtla/timerlat_top: Stop timerlat tracer on signal")
Fixes: 3aadb65db5d6 ("rtla/timerlat: Add action on end feature")
Link: https://lore.kernel.org/r/20260512173731.2151841-1-crwood@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
The code that checks the RTLA_NO_BPF environment variable calls
getenv() twice and uses strncmp() with a length of 2 to compare
against the single-character string "1". This is inefficient and
the comparison length is unnecessarily long.
Store the result of getenv() in a local variable to avoid the
redundant call, and replace strncmp() with strncmp_static() for
the exact match comparison. This follows the same pattern
established in recent commits that improved string comparison
consistency throughout the rtla codebase.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260309195040.1019085-15-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
The argument parsing code in timerlat_main() and osnoise_main() uses
strncmp() with a length of 1 to check if the first argument starts
with a dash, indicating an option flag was passed.
Replace this pattern with str_has_prefix() for consistency with the
rest of the codebase. While character comparison would be slightly
more efficient, using str_has_prefix() provides better readability
and maintains a uniform coding style throughout the rtla tool.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260309195040.1019085-14-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
nr_cpus does not change at runtime, so passing it through the macro
argument is unnecessary.
Remove the argument and use the global nr_cpus instead.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20260306194953.2511960-5-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
sysconf(_SC_NPROCESSORS_CONF) (via get_nprocs_conf) reflects
cpu_possible_mask, which is fixed at boot time, so querying it
repeatedly is unnecessary.
Replace multiple calls to sysconf(_SC_NPROCESSORS_CONF) with a single
global nr_cpus variable initialized once at startup.
`#pragma once` in timerlat_u.h is needed for pre-C23 compilers to avoid
redefinition errors.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20260306194953.2511960-2-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
In the current implementation, the auto-analysis code for printing the
stack captured in the tracefs buffer of the aa instance stops at the
first encountered address that cannot be resolved into a function
symbol.
This is not always the desired behavior on all platforms; sometimes,
there might be resolvable entries after unresolvable ones, and
sometimes, the user might want to inspect the raw pointers for the
unresolvable entries.
Add a new option, --stack-format, with three values:
- truncate: stop at first unresolvable entry. This is the current
behavior, and is kept as the default.
- skip: skip unresolvable entries, but do not stop on them.
- full: print all entries, including unresolvable ones.
To make this work, the "size" field of the stack entry is now also read
and used as the maximum number of entries to print, capped at 64, since
that is the fixed length of the "caller" field.
Link: https://lore.kernel.org/r/20260119115222.744150-1-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
Remove unused includes for <errno.h> and <signal.h> to clean up the
code and reduce unnecessary dependencies.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-12-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
Add option --bpf-action that allows the user to attach an external BPF
program that will be executed via BPF tail call on latency threshold
overflow.
Executing additional BPF code on latency threshold overflow allows doing
low-latency and in-kernel troubleshooting of the cause of the overflow.
The option takes an argument, which is a path to a BPF ELF file
expected to contain a function named "action_handler" in a section named
"tp/timerlat_action" (the section is necessary for libbpf to assign the
correct BPF program type to it).
Link: https://lore.kernel.org/r/20251126144205.331954-3-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
This avoids startup races where one of the instances hit a threshold
before all instances were enabled, and thus tracing stops without
the relevant event. In particular, this is not uncommon with the
tests that set a very tight threshold and then complain if there's
no analysis.
This also ensures that we don't stop tracing during a warmup.
The downside is a small chance of having an event over the threshold
early in the output, without stopping on it, which could cause user
confusion. This should be less likely if the warmup feature is used, but
that doesn't eliminate the race window, just the odds of an unusual spike
right at that moment.
Signed-off-by: Crystal Wood <crwood@redhat.com>
Link: https://lore.kernel.org/r/20251112152529.956778-6-crwood@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
In recently introduced timerlat_free(),
the variable 'nr_cpus' is not assigned.
Assign it with sysconf(_SC_NPROCESSORS_CONF) as done elsewhere.
Remove the culprit: -Wno-maybe-uninitialized. The rest of the
code is clean.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Fixes: 2f3172f9dd58 ("tools/rtla: Consolidate code between osnoise/timerlat and hist/top")
Link: https://lore.kernel.org/r/20251002170846.437888-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
The rtla tools have many instances of iterating over CPUs while
checking if they are monitored.
Add a for_each_monitored_cpu() helper macro to make the code
more readable and reduce code duplication.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251002123553.389467-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
|
|
Currently a lot of code is duplicated between the different rtla tools,
making maintenance more difficult, and encouraging divergence such as
features that are only implemented for certain tools even though they
could be more broadly applicable.
Merge the various main() functions into a common run_tool() with an ops
struct for tool-specific details.
Implement enough support for actions on osnoise to not need to keep the
old params->trace_output path.
Cc: John Kacur <jkacur@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-5-crwood@redhat.com
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Crystal Wood <crwood@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
Merge the common bits of osnoise_apply_config() and
timerlat_apply_config(). Put the result in a new common.c, and move
enough things to common.h so that common.c does not need to include
osnoise.h.
Cc: John Kacur <jkacur@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-4-crwood@redhat.com
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Crystal Wood <crwood@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
timerlat_params and osnoise_params structures contain 15 identical
fields.
Introduce a new header common.h and define a common_params structure to
consolidate shared fields, reduce code duplication, and enhance
maintainability.
Cc: John Kacur <jkacur@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-2-crwood@redhat.com
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Signed-off-by: Crystal Wood <crwood@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
After the introduction of BPF-based sample collection, rtla-timerlat
effectively runs in one of three modes:
- Pure BPF mode, with tracefs only being used to set up the timerlat
tracer. Sample processing and stop on threshold are handled by BPF.
- tracefs mode. BPF is unsupported or kernel is lacking the necessary
trace event (osnoise:timerlat_sample). Stop on theshold is handled by
timerlat tracer stopping tracing in all instances.
- BPF/tracefs mixed mode - BPF is used for sample collection for top or
histogram, tracefs is used for trace output and/or auto-analysis. Stop
on threshold is handled both through BPF program, which stops sample
collection for top/histogram and wakes up rtla, and by timerlat
tracer, which stops tracing for trace output/auto-analysis instances.
Add enum timerlat_tracing_mode, with three values:
- TRACING_MODE_BPF
- TRACING_MODE_TRACEFS
- TRACING_MODE_MIXED
Those represent the modes described above. A field of this type is added
to struct timerlat_params, named "mode", replacing the no_bpf variable.
params->mode is set in timerlat_{top,hist}_parse_args to
TRACING_MODE_BPF or TRACING_MODE_MIXED based on whether trace output
and/or auto-analysis is requested. timerlat_{top,hist}_main then checks
if BPF is not unavailable or disabled, in that case, it sets
params->mode to TRACING_MODE_TRACEFS.
A condition is added to timerlat_apply_config that skips setting
timerlat tracer thresholds if params->mode is TRACING_MODE_BPF (those
are unnecessary, since they only turn off tracing, which is already
turned off in that case, since BPF is used to collect samples).
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Chang Yin <cyin@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Cc: Crystal Wood <crwood@redhat.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Link: https://lore.kernel.org/20250626123405.1496931-2-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
rtla currently only sets tracer options that are explicitly set by the
user, with the exception of OSNOISE_WORKLOAD.
This leads to improper behavior in case rtla is run with those options
not set to the default value. rtla does reset them to the original
value upon exiting, but that does not protect it from starting with
non-default values set either by an improperly exited rtla or by another
user of the tracers.
For example, after running this command:
$ echo 1 > /sys/kernel/tracing/osnoise/stop_tracing_us
all runs of rtla will stop at the 1us threshold, even if not requested
by the user:
$ rtla osnoise hist
Index CPU-000 CPU-001
1 8 5
2 5 9
3 1 2
4 6 1
5 2 1
6 0 1
8 1 1
12 0 1
14 1 0
15 1 0
over: 0 0
count: 25 21
min: 1 1
avg: 3.68 3.05
max: 15 12
rtla osnoise hit stop tracing
Fix the problem by setting the default value for all tracer options if
the user has not provided their own value.
For most of the options, it's enough to just drop the if clause checking
for the value being set. For cpus, "all" is used as the default value,
and for osnoise default period and runtime, default values of
the osnoise_data variable in trace_osnoise.c are used.
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250320092500.101385-5-tglozar@redhat.com
Fixes: 1eceb2fc2ca5 ("rtla/osnoise: Add osnoise top mode")
Fixes: 829a6c0b5698 ("rtla/osnoise: Add the hist mode")
Fixes: a828cd18bc4a ("rtla: Add timerlat tool and timelart top mode")
Fixes: 1eeb6328e8b3 ("rtla/timerlat: Add timerlat hist mode")
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Reviewed-by: John Kacur <jkacur@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
The functions osnoise_top_apply_config and osnoise_hist_apply_config, as
well as timerlat_top_apply_config and timerlat_hist_apply_config, are
mostly the same.
Move common part from them into separate functions osnoise_apply_config
and timerlat_apply_config.
For rtla-timerlat, also unify params->user_hist and params->user_top
into one field called params->user_data, and move several fields used
only by timerlat-top into the top-only section of struct
timerlat_params.
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250320092500.101385-3-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Reviewed-by: John Kacur <jkacur@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
rtla_usage(), osnoise_usage() and timerlat_usage() all exit with an
error status.
However when these are called from help, they should exit with a
non-error status.
Fix this by passing the exit status to the functions.
Note, although we remove the subsequent call to exit after calling
usage, we leave it in at the end of a function to suppress the compiler
warning "control reaches end of a non-void function".
Link: https://lkml.kernel.org/r/20221107144313.22470-1-jkacur@redhat.com
Signed-off-by: John Kacur <jkacur@redhat.com>
Acked-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
The rtla hist hist mode displays a histogram of each tracer event
occurrence, both for IRQ and timer latencies. The tool also allows
many configurations of the timerlat tracer and the collection of
the tracer output.
Here is one example of the rtla timerlat hist mode output:
---------- %< ----------
[root@alien ~]# rtla timerlat hist -c 0-3 -d 1M
# RTLA timerlat histogram
# Time unit is microseconds (us)
# Duration: 0 00:01:00
Index IRQ-000 Thr-000 IRQ-001 Thr-001 IRQ-002 Thr-002 IRQ-003 Thr-003
0 58572 0 59373 0 58691 0 58895 0
1 1422 57021 628 57241 1310 56160 1102 56805
2 6 2931 0 2695 0 3567 4 3031
3 1 40 0 53 0 260 0 142
4 0 7 0 5 0 6 0 17
5 0 2 0 5 0 7 0 4
6 0 0 0 2 0 1 0 1
8 0 0 0 0 0 0 0 1
over: 0 0 0 0 0 0 0 0
count: 60001 60001 60001 60001 60001 60001 60001 60001
min: 0 1 0 1 0 1 0 1
avg: 0 1 0 1 0 1 0 1
max: 3 5 1 6 1 6 2 8
---------- >% ----------
Running
- rtla timerlat hist --help
provides information about the available options.
Link: https://lkml.kernel.org/r/7049ed3c46b7d6aceab18ffe7770003dfc4ddceb.1639158831.git.bristot@kernel.org
Cc: Tao Zhou <tao.zhou@linux.dev>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Daniel Bristot de Oliveira <bristot@kernel.org>
Cc: linux-rt-users@vger.kernel.org
Cc: linux-trace-devel@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
The rtla timerlat tool is an interface for the timerlat tracer.
The timerlat tracer dispatches a kernel thread per-cpu. These threads set a
periodic timer to wake themselves up and go back to sleep. After the
wakeup, they collect and generate useful information for the debugging of
operating system timer latency.
The timerlat tracer outputs information in two ways. It periodically
prints the timer latency at the timer IRQ handler and the Thread handler.
It also provides information for each noise via the osnoise tracepoints.
The rtla timerlat top mode displays a summary of the periodic output from
the timerlat tracer.
Here is one example of the rtla timerlat tool output:
---------- %< ----------
[root@alien ~]# rtla timerlat top -c 0-3 -d 1m
Timer Latency
0 00:01:00 | IRQ Timer Latency (us) | Thread Timer Latency (us)
CPU COUNT | cur min avg max | cur min avg max
0 #60001 | 0 0 0 3 | 1 1 1 6
1 #60001 | 0 0 0 3 | 2 1 1 5
2 #60001 | 0 0 1 6 | 1 1 2 7
3 #60001 | 0 0 0 7 | 1 1 1 11
---------- >% ----------
Running:
# rtla timerlat --help
# rtla timerlat top --help
provides information about the available options.
Link: https://lkml.kernel.org/r/e95032e20c2b88c962195bf7693bb53c9ebcced8.1639158831.git.bristot@kernel.org
Cc: Tao Zhou <tao.zhou@linux.dev>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Daniel Bristot de Oliveira <bristot@kernel.org>
Cc: linux-rt-users@vger.kernel.org
Cc: linux-trace-devel@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
