Several functions duplicate the logic for handling threshold actions.
When a threshold is reached, these functions stop the trace, perform
configured actions, and restart the trace if --on-threshold continue
is set.
Create common_threshold_handler() to centralize this shared logic and
avoid code duplication. The function executes the configured threshold
actions and restarts the necessary trace instances when appropriate.
Also add should_continue_tracing() helper to encapsulate the check
for whether tracing should continue after a threshold event, improving
code readability at call sites.
In timerlat_top_bpf_main_loop(), use common_params directly instead
of casting through timerlat_params when only common fields are needed.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260309195040.1019085-5-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The actions_parse() function uses open-coded logic to extract arguments
from a string. This includes manual length checks and strncmp() calls,
which can be verbose and error-prone.
To simplify and improve the robustness of argument parsing, introduce a
new extract_arg() helper macro. This macro extracts the value from a
"key=value" pair, making the code more concise and readable.
Also, introduce STRING_LENGTH() and strncmp_static() macros to
perform compile-time calculations of string lengths and safer string
comparisons.
Refactor actions_parse() to use these new helpers, resulting in
cleaner and more maintainable code.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260309195040.1019085-4-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The actions_add_trace_output() and actions_add_shell() functions were
using calloc() followed by strcpy() to allocate and copy a string.
This can be simplified by using strdup(), which allocates memory and
copies the string in a single step.
Replace the calloc() and strcpy() calls with strdup(), making the
code more concise and readable.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260309195040.1019085-3-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Most memory allocations in rtla happen during early initialization
before any resources are acquired that would require cleanup. In these
cases, propagating allocation errors just adds complexity without any
benefit. There's nothing to clean up, and the program must exit anyway.
This patch introduces fatal allocation wrappers (calloc_fatal,
reallocarray_fatal, strdup_fatal) that call fatal() on allocation
failure. These wrappers simplify the code by eliminating unnecessary
error propagation paths.
The patch converts early allocations to use these wrappers in
actions_init() and related action functions, osnoise_context_alloc()
and osnoise_init_tool(), trace_instance_init() and trace event
functions, and parameter structure allocations in main functions.
This simplifies the code while maintaining the same behavior: immediate
exit on allocation failure during initialization. Allocations that
require cleanup, such as those in histogram allocation functions with
goto cleanup paths, are left unchanged and continue to return errors.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260309195040.1019085-2-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>
nr_cpus does not change at runtime, so keeping it in struct members is
unnecessary.
Use the global nr_cpus instead of struct members.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20260306194953.2511960-4-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
nr_cpus does not change at runtime, so passing it through function
arguments is unnecessary.
Use the global nr_cpus instead of propagating it via parameters.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20260306194953.2511960-3-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>
Add unit tests for utility functions in src/utils.c using the Check
testing framework. The tests verify parse_cpu_set(), strtoi(), and
parse_prio() functions.
Unit tests are built conditionally when libcheck is available.
Run tests with 'make unit-test'.
The test framework uses the Check library which provides process
isolation for each test, preventing failures in one test from
affecting others.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20260119105857.797498-3-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
getopt_long() processes short and long options independently.
RTLA, like the majority of applications, uses both short and long
variants for each logical option.
Since the val member of struct option holds the letter of the short
variant, the string of short options can be reconstructed from the
array of long options.
Add getopt_auto() to generate optstring from an array of long options,
eliminating the need to maintain separate short option strings.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260108095011.2115719-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The patch 'Replace atoi() with a robust strtoi()' introduced a bug
in parse_cpu_set(), which relies on partial parsing of the input string.
The function parses CPU specifications like '0-3,5' by incrementing
a pointer through the string. strtoi() rejects strings with trailing
characters, causing parse_cpu_set() to fail on any CPU list with
multiple entries.
Restore the original use of atoi() in parse_cpu_set().
Fixes: 7e9dfccf8f ("rtla: Replace atoi() with a robust strtoi()")
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Link: https://lore.kernel.org/r/20260112192642.212848-2-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Correct the return value documentation for parse_cpu_set() function
in utils.c. The comment incorrectly stated that the function returns
1 on success and 0 on failure, but the actual implementation returns
0 on success and 1 on failure, following the common error-on-nonzero
convention used throughout the codebase.
This documentation fix ensures that developers reading the code
understand the correct return value semantics and prevents potential
misuse of the function's return value in conditional checks.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-18-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Add explicit null termination and buffer initialization for read()
operations in procfs_is_workload_pid() and get_self_cgroup() functions.
The read() system call does not null-terminate the data it reads, and
when the buffer is filled to capacity, subsequent string operations
will read past the buffer boundary searching for a null terminator.
In procfs_is_workload_pid(), explicitly set buffer[MAX_PATH-1] to '\0'
to ensure the buffer is always null-terminated before passing it to
strncmp(). In get_self_cgroup(), use memset() to zero the path buffer
before reading, which ensures null termination when retval is less than
MAX_PATH. Additionally, set path[MAX_PATH-1] to '\0' after the read to
handle the case where the buffer is filled completely.
These defensive buffer handling practices prevent potential buffer
overruns and align with the ongoing buffer safety improvements across
the rtla codebase.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-17-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The stop_tracing global variable is accessed from both the signal
handler context and the main program flow without synchronization.
This creates a potential race condition where compiler optimizations
could cache the variable value in registers, preventing the signal
handler's updates from being visible to other parts of the program.
Add the volatile qualifier to stop_tracing in both common.c and
common.h to ensure all accesses to this variable bypass compiler
optimizations and read directly from memory. This guarantees that
when the signal handler sets stop_tracing, the change is immediately
visible to the main program loop, preventing potential hangs or
delayed shutdown when termination signals are received.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-16-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The rtla tool generates various output files during testing and
execution, including custom trace outputs and histogram data. These
files are artifacts of running the tool with different options and
should not be tracked in version control.
Add gitignore entries for custom_filename.txt, osnoise_irq_noise_hist.txt,
osnoise_trace.txt, and timerlat_trace.txt to prevent accidentally
committing these generated files. This aligns with the existing pattern
of ignoring build artifacts and generated headers like *.skel.h.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-15-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The actions_parse() function uses strtok() to tokenize the trigger
string, but does not check if the returned token is NULL before
passing it to strcmp(). If the trigger parameter is an empty string
or contains only delimiter characters, strtok() returns NULL, causing
strcmp() to dereference a NULL pointer and crash the program.
This issue can be triggered by malformed user input or edge cases in
trigger string parsing. Add a NULL check immediately after the strtok()
call to validate that a token was successfully extracted before using
it. If no token is found, the function now returns -1 to indicate a
parsing error.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-13-wander@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>
The actions struct is allocated using calloc, which already returns
zeroed memory. The subsequent memset call to zero the 'present' member
is therefore redundant.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-10-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The result enum defines custom values for PASSED, ERROR, and FAILED.
These values correspond to standard exit codes EXIT_SUCCESS and
EXIT_FAILURE.
Update the enum to use the standard macros EXIT_SUCCESS and
EXIT_FAILURE to improve readability and adherence to standard C
practices.
The FAILED value is implicitly assigned EXIT_FAILURE + 1, so there
is no need to assign an explicit value.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-9-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The atoi() function does not perform error checking, which can lead to
undefined behavior when parsing invalid or out-of-range strings. This
can cause issues when parsing user-provided numerical inputs, such as
signal numbers, PIDs, or CPU lists.
To address this, introduce a new strtoi() helper function that safely
converts a string to an integer. This function validates the input and
checks for overflows, returning a negative value on failure.
Replace all calls to atoi() with the new strtoi() function and add
proper error handling to make the parsing more robust and prevent
potential issues.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-5-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The for loop to iterate over the list of actions is used in
more than one place. To avoid code duplication and improve
readability, introduce a for_each_action() helper macro.
Replace the open-coded for loops with the new helper.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20260106133655.249887-4-wander@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Both set_pid_cgroup() and set_comm_cgroup() functions contain
identical code for opening the cgroup.procs file.
Extract this common code into a new helper function open_cgroup_procs()
to reduce code duplication and improve maintainability.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251224125058.1771519-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -H/--house-keeping.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-8-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -P/--priority.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-7-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -e/--event.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-6-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -d/--duration.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-5-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -D/--debug.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-4-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -C/--cgroup.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-3-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of -c/--cpus.
Migrate the option parsing from individual tools to the
common_parse_options().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-2-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Each rtla tool duplicates parsing of many common options. This creates
maintenance overhead and risks inconsistencies when updating these
options.
Add common_parse_options() to centralize parsing of options used across
all tools.
Common options to be migrated in future patches.
Changes since v1:
- restore opterr
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251209100047.2692515-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Add -v flag to prove command to also print the names of tests that
succeeded, not only those that failed, to allow easier debugging of the
test suite.
Also, drop printing the option and value to stdout in
check_with_osnoise_options, which was a debugging print that was
accidentally left in the final commit, and which would be otherwise now
visible in make check output, as stdout is no longer suppressed.
Suggested-by: Crystal Wood <crwood@redhat.com>
Reviewed-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20251126144205.331954-6-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Add a test that implements a BPF program writing to a test map, which
is attached to RTLA via --bpf-action to be executed on theshold
overflow.
A combination of --on-threshold shell with bpftool (which is always
present if BPF support is enabled) is used to check whether the BPF
program has executed successfully.
Suggested-by: Crystal Wood <crwood@redhat.com>
Link: https://lore.kernel.org/r/20251126144205.331954-5-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Add an example BPF action program that prints the measured latency to
the tracefs buffer via bpf_printk().
A new Makefile target, "examples", is added to build the example. In
addition, "sample/" subfolder is renamed to "example".
If BPF skeleton support is unavailable or disabled, a warning will be
displayed when building the BPF action program example.
Link: https://lore.kernel.org/r/20251126144205.331954-4-tglozar@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>
Add a map to the rtla-timerlat BPF program that holds a file descriptor
of another BPF program, to be executed on threshold overflow.
timerlat_bpf_set_action() is added as an interface to set the program.
Link: https://lore.kernel.org/r/20251126144205.331954-2-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The rtla tools have significant code quadruplication in their usage
functions. Each tool implements its own version of the same help text
formatting and option descriptions, leading to maintenance overhead and
inconsistencies. Documentation/tools/rtla/common_options.rst lists 14
common options.
Add common_usage() infrastructure to consolidate help formatting.
Subsequent patches will extend this to handle other common options.
The refactored output is almost identical to the original, with the
following changes:
- add square brackets to specify optionality: `usage: [rtla] ...`
- remove `-q` from timerlat hist because hist tools don't support it
- minor spacing
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251124063204.845425-1-costa.shul@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>
Historically four function declarations remain orphaned or duplicated.
Remove them to keep the source clean.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/r/20251012071133.290225-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
- Officially add Tomas Glozar as a maintainer to RTLA tool
- Add for_each_monitored_cpu() helper
In multiple places, RTLA tools iterate over the list of CPUs running
tracer threads.
Use single helper instead of repeating the for/if combination.
- Remove unused variable option_index in argument parsing
RTLA tools use getopt_long() for argument parsing. For its last
argument, an unused variable "option_index" is passed.
Remove the variable and pass NULL to getopt_long() to shorten
the naturally long parsing functions, and make them more readable.
- Fix unassigned nr_cpus after code consolidation
In recent code consolidation, timerlat tool cleanup, previously
implemented separately for each tool, was moved to a common function
timerlat_free().
The cleanup relies on nr_cpus being set. This was not done in the new
function, leaving the variable uninitialized.
Initialize the variable properly, and remove silencing of compiler
warning for uninitialized variables.
- Stop tracing on user latency in BPF mode
Despite the name, rtla-timerlat's -T/--thread option sets timerlat's
stop_tracing_total_us option, which also stops tracing on
return-from-user latency, not only on thread latency.
Implement the same behavior also in BPF sample collection stop tracing
handler to avoid a discrepancy and restore correspondence of behavior
with the equivalent option of cyclictest.
- Fix threshold actions always triggering
A bug in threshold action logic caused the action to execute even
if tracing did not stop because of threshold.
Fix the logic to stop correctly.
- Fix few minor issues in tests
Extend tests that were shown to need it to 5s, fix osnoise test
calling timerlat by mistake, and use new, more reliable output
checking in timerlat's "top stop at failed action" test.
- Do not print usage on argument parsing error
RTLA prints the entire usage message on encountering errors in
argument parsing, like a malformed CPU list.
The usage message has gotten too long. Instead of printing it,
use newly added fatal() helper function to simply exit with
the error message, excluding the usage.
- Fix unintuitive -C/--cgroup interface
"-C cgroup" and "--cgroup cgroup" are invalid syntax, despite that
being a common way to specify an option with argument. Moreover,
using them fails silently and no cgroup is set.
Create new helper function to unify the handling of all such options
and allow all of:
-Xsomething
-X=something
-X something
as well as the equivalent for the long option.
- Fix -a overriding -t argument filename
Fix a bug where -a following -t custom_file.txt overrides the custom
filename with the default timerlat_trace.txt.
- Stop tracing correctly on multiple events at once
In some race scenarios, RTLA BPF sample collection might send multiple
stop tracing events via the BPF ringbuffer at once.
Compare the number of events for != 0 instead of == 1 to cover for
this scenario and stop tracing properly.
-----BEGIN PGP SIGNATURE-----
iIoEABYKADIWIQRRSw7ePDh/lE+zeZMp5XQQmuv6qgUCaS9bxBQccm9zdGVkdEBn
b29kbWlzLm9yZwAKCRAp5XQQmuv6qhrgAP0a/AtsL9+IFXAK5JK8aO1XWApVyK9n
48FRZWu/jrupuAD7BO+EHazmPEourNaUqYPeuymwxT+4O47RH1Q/aasLQwo=
=RvNH
-----END PGP SIGNATURE-----
Merge tag 'trace-tools-v6.19' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull rtla trace tooling updates from Steven Rostedt:
- Officially add Tomas Glozar as a maintainer to RTLA tool
- Add for_each_monitored_cpu() helper
In multiple places, RTLA tools iterate over the list of CPUs running
tracer threads.
Use single helper instead of repeating the for/if combination.
- Remove unused variable option_index in argument parsing
RTLA tools use getopt_long() for argument parsing. For its last
argument, an unused variable "option_index" is passed.
Remove the variable and pass NULL to getopt_long() to shorten the
naturally long parsing functions, and make them more readable.
- Fix unassigned nr_cpus after code consolidation
In recent code consolidation, timerlat tool cleanup, previously
implemented separately for each tool, was moved to a common function
timerlat_free().
The cleanup relies on nr_cpus being set. This was not done in the new
function, leaving the variable uninitialized.
Initialize the variable properly, and remove silencing of compiler
warning for uninitialized variables.
- Stop tracing on user latency in BPF mode
Despite the name, rtla-timerlat's -T/--thread option sets timerlat's
stop_tracing_total_us option, which also stops tracing on
return-from-user latency, not only on thread latency.
Implement the same behavior also in BPF sample collection stop
tracing handler to avoid a discrepancy and restore correspondence of
behavior with the equivalent option of cyclictest.
- Fix threshold actions always triggering
A bug in threshold action logic caused the action to execute even if
tracing did not stop because of threshold.
Fix the logic to stop correctly.
- Fix few minor issues in tests
Extend tests that were shown to need it to 5s, fix osnoise test
calling timerlat by mistake, and use new, more reliable output
checking in timerlat's "top stop at failed action" test.
- Do not print usage on argument parsing error
RTLA prints the entire usage message on encountering errors in
argument parsing, like a malformed CPU list.
The usage message has gotten too long. Instead of printing it, use
newly added fatal() helper function to simply exit with the error
message, excluding the usage.
- Fix unintuitive -C/--cgroup interface
"-C cgroup" and "--cgroup cgroup" are invalid syntax, despite that
being a common way to specify an option with argument. Moreover,
using them fails silently and no cgroup is set.
Create new helper function to unify the handling of all such options
and allow all of:
-Xsomething
-X=something
-X something
as well as the equivalent for the long option.
- Fix -a overriding -t argument filename
Fix a bug where -a following -t custom_file.txt overrides the custom
filename with the default timerlat_trace.txt.
- Stop tracing correctly on multiple events at once
In some race scenarios, RTLA BPF sample collection might send
multiple stop tracing events via the BPF ringbuffer at once.
Compare the number of events for != 0 instead of == 1 to cover for
this scenario and stop tracing properly.
* tag 'trace-tools-v6.19' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
rtla/timerlat: Exit top main loop on any non-zero wait_retval
rtla/tests: Don't rely on matching ^1ALL
rtla: Fix -a overriding -t argument
rtla: Fix -C/--cgroup interface
tools/rtla: Replace osnoise_hist_usage("...") with fatal("...")
tools/rtla: Replace osnoise_top_usage("...") with fatal("...")
tools/rtla: Replace timerlat_hist_usage("...") with fatal("...")
tools/rtla: Replace timerlat_top_usage("...") with fatal("...")
tools/rtla: Add fatal() and replace error handling pattern
rtla/tests: Fix osnoise test calling timerlat
rtla/tests: Extend action tests to 5s
tools/rtla: Fix --on-threshold always triggering
rtla/timerlat_bpf: Stop tracing on user latency
tools/rtla: Fix unassigned nr_cpus
tools/rtla: Remove unused optional option_index
tools/rtla: Add for_each_monitored_cpu() helper
MAINTAINERS: Add Tomas Glozar as a maintainer to RTLA tool
Comparing to exactly 1 will fail if more than one ring buffer
event was seen since the last call to timerlat_bpf_wait(), which
can happen in some race scenarios.
Signed-off-by: Crystal Wood <crwood@redhat.com>
Link: https://lore.kernel.org/r/20251112152529.956778-5-crwood@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The timerlat "top stop at failed action" test was relying on "ALL" being
printed immediately after the "1" from the threshold action. Besides being
fragile, this depends on stdbuf behavior, which is easy to miss when
recreating the test outside of the framework for debugging purposes.
Instead, use the expected/unexpected text mechanism from the
corresponding osnoise test.
Signed-off-by: Crystal Wood <crwood@redhat.com>
Link: https://lore.kernel.org/r/20251112152529.956778-2-crwood@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
When running rtla as
`rtla <timerlat|osnoise> <top|hist> -t custom_file.txt -a 100`
-a options override trace output filename specified by -t option.
Running the command above will create <timerlat|osnoise>_trace.txt file
instead of custom_file.txt. Fix this by making sure that -a option does
not override trace output filename even if it's passed after trace
output filename is specified.
Fixes: 173a3b0148 ("rtla/timerlat: Add the automatic trace option")
Signed-off-by: Ivan Pravdin <ipravdin.official@gmail.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/b6ae60424050b2c1c8709e18759adead6012b971.1762186418.git.ipravdin.official@gmail.com
[ use capital letter in subject, as required by tracing subsystem ]
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Currently, user can only specify cgroup to the tracer's thread the
following ways:
`-C[cgroup]`
`-C[=cgroup]`
`--cgroup[=cgroup]`
If user tries to specify cgroup as `-C [cgroup]` or `--cgroup [cgroup]`,
the parser silently fails and rtla's cgroup is used for the tracer
threads.
To make interface more user-friendly, allow user to specify cgroup in
the aforementioned way, i.e. `-C [cgroup]` and `--cgroup [cgroup]`.
Refactor identical logic between -t/--trace and -C/--cgroup into a
common function.
Change documentation to reflect this user interface change.
Fixes: a957cbc025 ("rtla: Add -C cgroup support")
Signed-off-by: Ivan Pravdin <ipravdin.official@gmail.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/16132f1565cf5142b5fbd179975be370b529ced7.1762186418.git.ipravdin.official@gmail.com
[ use capital letter in subject, as required by tracing subsystem ]
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
A long time ago, when the usage help was short, it was a favor
to the user to show it on error. Now that the usage help has
become very long, it is too noisy to dump the complete help text
for each typo after the error message itself.
Replace osnoise_hist_usage("...") with fatal("...") on errors.
Remove the already unused 'usage' argument from osnoise_hist_usage().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251011082738.173670-6-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
A long time ago, when the usage help was short, it was a favor
to the user to show it on error. Now that the usage help has
become very long, it is too noisy to dump the complete help text
for each typo after the error message itself.
Replace osnoise_top_usage("...") with fatal("...") on errors.
Remove the already unused 'usage' argument from osnoise_top_usage().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251011082738.173670-5-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
A long time ago, when the usage help was short, it was a favor
to the user to show it on error. Now that the usage help has
become very long, it is too noisy to dump the complete help text
for each typo after the error message itself.
Replace timerlat_hist_usage("...\n") with fatal("...") on errors.
Remove the already unused 'usage' argument from timerlat_hist_usage().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251011082738.173670-4-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
A long time ago, when the usage help was short, it was a favor
to the user to show it on error. Now that the usage help has
become very long, it is too noisy to dump the complete help text
for each typo after the error message itself.
Replace timerlat_top_usage("...\n") with fatal("...") on errors.
Remove the already unused 'usage' argument from timerlat_top_usage().
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251011082738.173670-3-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
The code contains some technical debt in error handling,
which complicates the consolidation of duplicated code.
Introduce an fatal() function to replace the common pattern of
err_msg() followed by exit(EXIT_FAILURE), reducing the length of an
already long function.
Further patches using fatal() follow.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251011082738.173670-2-costa.shul@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
osnoise test "top stop at failed action" is calling timerlat instead of
osnoise by mistake.
Fix it so that it calls the correct RTLA subcommand.
Fixes: 05b7e10687 ("tools/rtla: Add remaining support for osnoise actions")
Reviewed-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20251007095341.186923-3-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
In non-BPF mode, it takes up to 1 second for RTLA to notice that tracing
has been stopped. That means that action tests cannot have a 1 second
duration, as the SIGALRM will be racing with the threshold overflow.
Previously, non-BPF mode actions were buggy and always executed
the action, even when stopping on duration or SIGINT, preventing
this issue from manifesting. Now that this has been fixed, the tests
have become flaky, and this has to be adjusted.
Fixes: 4e26f84abf ("rtla/tests: Add tests for actions")
Fixes: 05b7e10687 ("tools/rtla: Add remaining support for osnoise actions")
Reviewed-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20251007095341.186923-2-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Commit 8d933d5c89 ("rtla/timerlat: Add continue action") moved the
code performing on-threshold actions (enabled through --on-threshold
option) to inside the RTLA main loop.
The condition in the loop does not check whether the threshold was
actually exceeded or if stop tracing was requested by the user through
SIGINT or duration. This leads to a bug where on-threshold actions are
always performed, even when the threshold was not hit.
(BPF mode is not affected, since it uses a different condition in the
while loop.)
Add a condition that checks for !stop_tracing before executing the
actions. Also, fix incorrect brackets in hist_main_loop to match the
semantics of top_main_loop.
Fixes: 8d933d5c89 ("rtla/timerlat: Add continue action")
Fixes: 2f3172f9dd ("tools/rtla: Consolidate code between osnoise/timerlat and hist/top")
Reviewed-by: Crystal Wood <crwood@redhat.com>
Reviewed-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20251007095341.186923-1-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
rtla-timerlat allows a *thread* latency threshold to be set via the
-T/--thread option. However, the timerlat tracer calls this *total*
latency (stop_tracing_total_us), and stops tracing also when the
return-to-user latency is over the threshold.
Change the behavior of the timerlat BPF program to reflect what the
timerlat tracer is doing, to avoid discrepancy between stopping
collecting data in the BPF program and stopping tracing in the timerlat
tracer.
Cc: stable@vger.kernel.org
Fixes: e34293ddce ("rtla/timerlat: Add BPF skeleton to collect samples")
Reviewed-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20251006143100.137255-1-tglozar@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: 2f3172f9dd ("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 longindex argument of getopt_long() is optional
and tied to the unused local variable option_index.
Remove it to shorten the four longest functions
and make the code neater.
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Link: https://lore.kernel.org/r/20251002123553.389467-2-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>
The help message incorrectly listed '-t' as the short option for
--threads, but the actual getopt_long configuration uses '-e'.
This mismatch can confuse users and lead to incorrect command-line
usage. This patch updates the usage string to correctly show:
"-e, --threads NRTHR"
to match the implementation.
Note: checkpatch.pl reports a false-positive spelling warning on
'Run', which is intentional.
Link: https://patch.msgid.link/20251106031040.1869-1-zhangchujun@cmss.chinamobile.com
Signed-off-by: Zhang Chujun <zhangchujun@cmss.chinamobile.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
- This update is mostly just consolidating code between osnoise/timerlat
and top/hist for easier maintenance and less future divergence.
-----BEGIN PGP SIGNATURE-----
iIoEABYKADIWIQRRSw7ePDh/lE+zeZMp5XQQmuv6qgUCaN/guhQccm9zdGVkdEBn
b29kbWlzLm9yZwAKCRAp5XQQmuv6qqU9AQCO+u+Qmx678DCfDJo9X1UPDtS/bM5f
r30X1pwYfZ3nNAEA47hbkVFcryFJZbrIPxuTGb0GSM36PHAxmch4QAwBqgs=
=qzZh
-----END PGP SIGNATURE-----
Merge tag 'trace-tools-v6.18' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull tracing tools updates from Steven Rostedt
- This is mostly just consolidating code between osnoise/timerlat and
top/hist for easier maintenance and less future divergence
* tag 'trace-tools-v6.18' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tools/rtla: Add remaining support for osnoise actions
tools/rtla: Add test engine support for unexpected output
tools/rtla: Fix -A option name in test comment
tools/rtla: Consolidate code between osnoise/timerlat and hist/top
tools/rtla: Create common_apply_config()
tools/rtla: Move top/hist params into common struct
tools/rtla: Consolidate common parameters into shared structure
The condition to check if the actions buffer needs to be resized was
incorrect. The check `self->size >= self->len` would evaluate to
true on almost every call to `actions_new()`, causing the buffer to
be reallocated unnecessarily each time an action was added.
Fix the condition to `self->len >= self.size`, ensuring
that the buffer is only resized when it is actually full.
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/20250915181101.52513-1-wander@redhat.com
Fixes: 6ea082b171 ("rtla/timerlat: Add action on threshold feature")
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The basic functionality came with the consolidation; now hook up the
command line options, and add documentation and tests.
Cc: John Kacur <jkacur@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-8-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>
Add a check() parameter to indicate which text must not appear in the
output.
Simplify the code so that we can print failures as they happen rather
than trying to figure out what went wrong after printing "not ok". This
also means that "not ok" gets printed after the info rather than before,
which seems more intuitive anyway.
Cc: John Kacur <jkacur@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-7-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>
This was changed to --on-threshold when the patches were applied.
Cc: John Kacur <jkacur@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-6-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>
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>
The hist members were very similar between timerlat and top, so
just use one common hist struct.
output_divisor, quiet, and pretty printing are pretty generic
concepts that can go in the main struct even if not every
specific tool (currently) uses them.
Cc: John Kacur <jkacur@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250907022325.243930-3-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>
The tool pkg-config used to check libtraceevent and libtracefs, if not
installed, it will report the libs not found, even though they have
already been installed.
Before:
libtraceevent is missing. Please install libtraceevent-dev/libtraceevent-devel
libtracefs is missing. Please install libtracefs-dev/libtracefs-devel
After:
Makefile.config:10: *** Error: pkg-config needed by libtraceevent/libtracefs is missing
on this system, please install it.
Link: https://lore.kernel.org/20250808040527.2036023-2-chen.dylane@linux.dev
Fixes: 01474dc706 ("tools/rtla: Use tools/build makefiles to build rtla")
Signed-off-by: Tao Chen <chen.dylane@linux.dev>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The tool pkg-config used to check libtraceevent and libtracefs, if not
installed, it will report the libs not found, even though they have
already been installed.
Before:
libtraceevent is missing. Please install libtraceevent-dev/libtraceevent-devel
libtracefs is missing. Please install libtracefs-dev/libtracefs-devel
After:
Makefile.config:10: *** Error: pkg-config needed by libtraceevent/libtracefs is missing
on this system, please install it.
Link: https://lore.kernel.org/20250808040527.2036023-1-chen.dylane@linux.dev
Fixes: 9d56c88e52 ("tools/tracing: Use tools/build makefiles on latency-collector")
Signed-off-by: Tao Chen <chen.dylane@linux.dev>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The -P option is used to set priority of osnoise and timerlat threads.
Extend the test for -P with --on-threshold calling a script that looks
for running timerlat threads and checks if their priority is set
correctly.
As --on-threshold is only supported by timerlat at the moment, this is
only implemented there so far.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Chang Yin <cyin@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250725133817.59237-3-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Checking for patterns in rtla output with grep was added to test rtla
actions. Add grep checks also for base tests where applicable.
Also fix trace event histogram trigger check to use the correct syntax
for the command-line option so that the test passes with the grep check.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Chang Yin <cyin@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250725133817.59237-2-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Many of the original rtla tests included durations of 1 minute and 30
seconds. Experience has shown this is unnecessary, since 10 seconds as
waiting time for samples to appear.
Change duration of all rtla tests to at most 10 seconds. This speeds up
testing significantly.
Before:
$ make check
All tests successful.
Files=3, Tests=54, 536 wallclock secs
( 0.03 usr 0.00 sys + 20.31 cusr 22.02 csys = 42.36 CPU)
Result: PASS
After:
$ make check
...
All tests successful.
Files=3, Tests=54, 196 wallclock secs
( 0.03 usr 0.01 sys + 20.28 cusr 20.68 csys = 41.00 CPU)
Result: PASS
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/20250626123405.1496931-9-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Add a bunch of tests covering most of both --on-threshold and --on-end.
Parts sensitive to implementation of hist/top are tested for both.
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/20250626123405.1496931-8-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Add argument to the check command in the test suite that takes a regular
expression that the output of rtla command is checked against. This
allows testing for specific information in rtla output in addition
to checking the return value.
Two minor improvements are included: running rtla with "eval" so that
arguments with spaces can be passed to it via shell quotations, and
the stdout of pushd and popd is suppressed to clean up the test output.
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/20250626123405.1496931-7-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Implement actions on end next to actions on threshold. A new option,
--on-end is added, parallel to --on-threshold. Instead of being
executed whenever a latency threshold is reached, it is executed at the
end of the measurement.
For example:
$ rtla timerlat hist -d 5s --on-end trace
will save the trace output at the end.
All actions supported by --on-threshold are also supported by --on-end,
except for continue, which does nothing with --on-end.
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/20250626123405.1496931-6-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Introduce option to resume tracing after a latency threshold overflow.
The option is implemented as an action named "continue".
Example:
$ rtla timerlat top -q -T 200 -d 1s --on-threshold \
exec,command="echo Threshold" --on-threshold continue
Threshold
Threshold
Threshold
Timer Latency
...
The feature is supported for both hist and top. After the continue
action is executed, processing of the list of actions is stopped and
tracing is resumed.
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/20250626123405.1496931-5-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Currently, rtla-timerlat BPF program uses a global variable stored in a
.bss section to store whether tracing has been stopped.
Move the information to a separate map, so that it is easily writable
from userspace, and add a function that clears the value, resuming
tracing after it has been stopped.
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/20250626123405.1496931-4-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Extend the functionality provided by the -t/--trace option, which
triggers saving the contents of a tracefs buffer after tracing is
stopped, to support implementing arbitrary actions.
A new option, --on-threshold, is added, taking an argument
that further specifies the action. Actions added in this patch are:
- trace[,file=<filename>]: Saves tracefs buffer, optionally taking a
filename.
- signal,num=<sig>,pid=<pid>: Sends signal to process. "parent" might
be specified instead of number to send signal to parent process.
- shell,command=<command>: Execute shell command.
Multiple actions may be specified and will be executed in order,
including multiple actions of the same type. Trace output requested via
-t and -a now adds a trace action to the end of the list.
If an action fails, the following actions are not executed. For
example, this command:
$ rtla timerlat -T 20 --on-threshold trace \
--on-threshold shell,command="grep ipi_send timerlat_trace.txt" \
--on-threshold signal,num=2,pid=parent
will send signal 2 (SIGINT) to parent process, but only if saved trace
contains the text "ipi_send".
This way, the feature can be used for flexible reactions on latency
spikes, and allows combining rtla with other tooling like perf.
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/20250626123405.1496931-3-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@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>
- Set distinctive value for failed tests
When running "make check" that performs tests on rtla the failure is
checked by examining the output. Instead have the tool return an error
status if it exceeds the threadhold.
- Define __NR_sched_setattr for LoongArch
Define __NR_sched_setattr to allow this to build for LoongArch.
- Define _GNU_SOURCE for timerlat_bpf.c
Due to modifications of struct sched_attr in utils.h when _GNU_SOURCE is
not defined, this can cause errors for timerlat_bpf_init() and breakage in
BPF sample collection mode.
-----BEGIN PGP SIGNATURE-----
iIoEABYKADIWIQRRSw7ePDh/lE+zeZMp5XQQmuv6qgUCaDeTzBQccm9zdGVkdEBn
b29kbWlzLm9yZwAKCRAp5XQQmuv6qokRAP0XJzos+uvQtkGrqiX5SB/rn1s3/tiD
nZagARyiV06BAwEA+NNzqFyx/BLUwMnpx/HFTnIMGXbRVWCVAEeL3t77zgk=
=dx7t
-----END PGP SIGNATURE-----
Merge tag 'trace-tools-v6.16' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull tracing tools updates from Steven Rostedt:
- Set distinctive value for failed tests
When running "make check" that performs tests on rtla the failure is
checked by examining the output. Instead have the tool return an
error status if it exceeds the threadhold.
- Define __NR_sched_setattr for LoongArch
Define __NR_sched_setattr to allow this to build for LoongArch.
- Define _GNU_SOURCE for timerlat_bpf.c
Due to modifications of struct sched_attr in utils.h when _GNU_SOURCE
is not defined, this can cause errors for timerlat_bpf_init() and
breakage in BPF sample collection mode.
* tag 'trace-tools-v6.16' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
rtla: Define _GNU_SOURCE in timerlat_bpf.c
rtla: Define __NR_sched_setattr for LoongArch
rtla: Set distinctive exit value for failed tests
Newer versions of glibc include a definition of struct sched_attr in
bits/sched.h (included through sched.h which is included by rtla).
Commit 0eecee3406 ("tools/rtla: fix collision with glibc
sched_attr/sched_set_attr") has modified the definition of struct
sched_attr in utils.h, so that it is only applied with older versions of
glibc that do not define it, in order to prevent build failure.
The definition in bits/sched.h depends on _GNU_SOURCE.
timerlat_bpf.c does not define _GNU_SOURCE, making it fall back to the
definition in utils.h. The latter has two fields less, leading to
shifted offsets of struct timerlat_params in timerlat_bpf_init.
Because of the shift, timerlat_bpf_init incorrectly reads
params->entries as 0 for timerlat-hist and disables the creation of
histogram maps, causing breakage in BPF sample collection mode:
$ rtla timerlat hist -d 1s
Error pulling BPF data
Fix the issue by also defining _GNU_SOURCE in timerlat_bpf.c.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250430144651.621766-1-tglozar@redhat.com
Fixes: e34293ddce ("rtla/timerlat: Add BPF skeleton to collect samples")
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
When executing "make -C tools/tracing/rtla" on LoongArch, there exists
the following error:
src/utils.c:237:24: error: '__NR_sched_setattr' undeclared
Just define __NR_sched_setattr for LoongArch if not exist.
Link: https://lore.kernel.org/20250422074917.25771-1-yangtiezhu@loongson.cn
Reported-by: Haiyong Sun <sunhaiyong@loongson.cn>
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
A test is considered failed when a sample trace exceeds the threshold.
Failed tests return the same exit code as passed tests, requiring test
frameworks to determine the result by searching for "hit stop tracing"
in the output.
Assign a distinct exit code for failed tests to enable the use of shell
expressions and seamless integration with testing frameworks without the
need to parse output.
Add enum type for return value.
Update `make check`.
Cc: Daniel Bristot de Oliveira <bristot@kernel.org>
Cc: John Kacur <jkacur@redhat.com>
Cc: "Luis Claudio R. Goncalves" <lgoncalv@redhat.com>
Cc: Eder Zulian <ezulian@redhat.com>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Jan Stancek <jstancek@redhat.com>
Link: https://lore.kernel.org/20250417185757.2194541-1-costa.shul@redhat.com
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Add dependencies needed to build rtla with BPF sample collection support
to README, and document both ways of sample collection in the manpages.
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Acked-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Reviewed-by: Luis Claudio R. Goncalves <lgoncalv@redhat.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Link: https://lore.kernel.org/r/20250311114936.148012-5-tglozar@redhat.com
- Allow RTLA to collect data via BPF
The current implementation of rtla uses libtracefs and libtraceevent to
pull sample events generated by the timerlat tracer from the trace
buffer. rtla then processes the sample by updating the histogram and
summary (current, maximum, minimum, and sum values) as well as checks
if tracing has been stopped due to threshold overflow.
In use cases where a large number of samples is being generated, that
is, with measurements running on many CPUs and with a low interval,
this sample processing design causes a significant CPU load on the rtla
side. Furthermore, with >100 CPUs and 100us interval, rtla was reported
as not being able to keep up with the samples and dropping most of them,
leading to it being unusable.
Change the way the timerlat trace processes samples by attaching
a BPF program to the trace event using the BPF skeleton feature of bpftool.
Unlike the current implementation, the BPF implementation does not check
whether tracing is stopped (in BPF mode, tracing is always off to improve
performance), but waits for a write to a BPF ringbuffer instead. This allows
rtla to exit immediately when a threshold is violated, without waiting
for the next iteration of the while loop.
If the requirements for the BPF implementation are not met, either at
build time or at run time, the current implementation is used as
fallback. Which implementation is being used can be seen when running
rtla timerlat with "-D" option. rtla can be forced to run in non-BPF
mode by setting the RTLA_NO_BPF option to 1, for debugging purposes.
- Fix LD_FLAGS from being dropped in build
- Refactor code to remove duplication of save_trace_to_file
- Always set options and do not rely on default settings
Do not rely on the default kernel settings of the tracers when
starting. They could have been changed by the user which gives
inconsistent results. Always set the options that rtla expects.
- Add creation of ctags and TAGS for traversing code
-----BEGIN PGP SIGNATURE-----
iIoEABYIADIWIQRRSw7ePDh/lE+zeZMp5XQQmuv6qgUCZ+WBgRQccm9zdGVkdEBn
b29kbWlzLm9yZwAKCRAp5XQQmuv6qg54AQDCOChaSSBiUkD0VoPKIeDMlPfvO5Qz
Xvrst5gtopfKFgEA12/9Lll/sh1eoc4saeGBooNY48HBUMjmX3KNFB14PQg=
=aHFu
-----END PGP SIGNATURE-----
Merge tag 'trace-tools-v6.15' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull tracing tooling updates from Steven Rostedt:
- Allow RTLA to collect data via BPF
The current implementation of rtla uses libtracefs and libtraceevent
to pull sample events generated by the timerlat tracer from the trace
buffer. rtla then processes the sample by updating the histogram and
summary (current, maximum, minimum, and sum values) as well as checks
if tracing has been stopped due to threshold overflow.
In use cases where a large number of samples is being generated, that
is, with measurements running on many CPUs and with a low interval,
this sample processing design causes a significant CPU load on the
rtla side. Furthermore, with >100 CPUs and 100us interval, rtla was
reported as not being able to keep up with the samples and dropping
most of them, leading to it being unusable.
Change the way the timerlat trace processes samples by attaching a
BPF program to the trace event using the BPF skeleton feature of
bpftool. Unlike the current implementation, the BPF implementation
does not check whether tracing is stopped (in BPF mode, tracing is
always off to improve performance), but waits for a write to a BPF
ringbuffer instead. This allows rtla to exit immediately when a
threshold is violated, without waiting for the next iteration of the
while loop.
If the requirements for the BPF implementation are not met, either at
build time or at run time, the current implementation is used as
fallback. Which implementation is being used can be seen when running
rtla timerlat with "-D" option. rtla can be forced to run in non-BPF
mode by setting the RTLA_NO_BPF option to 1, for debugging purposes.
- Fix LD_FLAGS from being dropped in build
- Refactor code to remove duplication of save_trace_to_file
- Always set options and do not rely on default settings
Do not rely on the default kernel settings of the tracers when
starting. They could have been changed by the user which gives
inconsistent results. Always set the options that rtla expects.
- Add creation of ctags and TAGS for traversing code
* tag 'trace-tools-v6.15' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
rtla: Add the ability to create ctags and etags
rtla/tests: Test setting default options
rtla/tests: Reset osnoise options before check
rtla: Always set all tracer options
rtla/osnoise: Set OSNOISE_WORKLOAD to true
rtla: Unify apply_config between top and hist
rtla/osnoise: Unify params struct
rtla: Fix segfault in save_trace_to_file call
tools/build: Use SYSTEM_BPFTOOL for system bpftool
rtla: Refactor save_trace_to_file
tools/rv: Keep user LDFLAGS in build
rtla/timerlat: Test BPF mode
rtla/timerlat_top: Use BPF to collect samples
rtla/timerlat_top: Move divisor to update
rtla/timerlat_hist: Use BPF to collect samples
rtla/timerlat: Add BPF skeleton to collect samples
rtla: Add optional dependency on BPF tooling
tools/build: Add bpftool-skeletons feature test
rtla/timerlat: Unify params struct
- Add the ability to create and remove ctags and etags, using the following
make tags
make TAGS
make tags_clean
- fix a comment in Makefile.rtla with the correct spelling and don't
imply that the ability to create an rtla tarball will be removed
Cc: Tomas Glozar <tglozar@redhat.com>
Cc: "Luis Claudio R . Goncalves" <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250321175053.29048-1-jkacur@redhat.com
Signed-off-by: John Kacur <jkacur@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Add function to test engine to test with pre-set osnoise options, and
use it to test whether osnoise period (as an example) is set correctly.
The test works by pre-setting a high period of 10 minutes and stop on
threshold. Thus, it is easy to check whether rtla is properly resetting
the period to default: if it is, the test will complete on time, since
the first sample will overflow the threshold. If not, it will time out.
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250320092500.101385-7-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>
Remove any dangling tracing instances from previous improperly exited
runs of rtla, and reset osnoise options to default before running a test
case.
This ensures that the test results are deterministic. Specific test
cases checked that rtla behaves correctly even when the tracer state is
not clean will be added later.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250320092500.101385-6-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: 1eceb2fc2c ("rtla/osnoise: Add osnoise top mode")
Fixes: 829a6c0b56 ("rtla/osnoise: Add the hist mode")
Fixes: a828cd18bc ("rtla: Add timerlat tool and timelart top mode")
Fixes: 1eeb6328e8 ("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>
If running rtla osnoise with NO_OSNOISE_WORKLOAD, it reports no samples:
$ echo NO_OSNOISE_WORKLOAD > /sys/kernel/tracing/osnoise/options
$ rtla osnoise hist -d 10s
Index
over: 0
count: 0
min: 0
avg: 0
max: 0
This situation can also happen when running rtla-osnoise after an
improperly exited rtla-timerlat run.
Set OSNOISE_WORKLOAD in rtla-osnoise, too, similarly to what we
already did for timerlat in commit 217f0b1e99 ("rtla/timerlat_top: Set
OSNOISE_WORKLOAD for kernel threads") and commit d8d866171a
("rtla/timerlat_hist: Set OSNOISE_WORKLOAD for kernel threads").
Note that there is no user workload mode for rtla-osnoise yet, so
OSNOISE_WORKLOAD is always set to true.
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250320092500.101385-4-tglozar@redhat.com
Fixes: 1eceb2fc2c ("rtla/osnoise: Add osnoise top mode")
Fixes: 829a6c0b56 ("rtla/osnoise: Add the 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>
Instead of having separate structs osnoise_top_params and
osnoise_hist_params, use one struct osnoise_params for both.
This allows code using the structs to be shared between osnoise-top and
osnoise-hist.
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250320092500.101385-2-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>
Running rtla with exit on threshold, but without saving trace leads to a
segmenetation fault:
$ rtla timerlat hist -T 10
...
Max timerlat IRQ latency from idle: 4.29 us in cpu 0
Segmentation fault
This is caused by null pointer deference in the call of
save_trace_to_file, which attempts to dereference an uninitialized
osnoise_tool variable:
save_trace_to_file(record->trace.inst, params->trace_output);
^ this is uninitialized if params->trace_output is
not set
Fix this by not attempting to dereference "record" if it is NULL and
passing NULL instead. As a safety measure, the first field is also
checked for NULL inside save_trace_to_file.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Costa Shulyupin <costa.shul@redhat.com>
Link: https://lore.kernel.org/20250313141034.299117-1-tglozar@redhat.com
Fixes: dc4d4e7c72 ("rtla: Refactor save_trace_to_file")
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The feature test for system bpftool uses BPFTOOL as the variable to set
its path, defaulting to just "bpftool" if not set by the user.
This conflicts with selftests and a few other utilities, which expect
BPFTOOL to be set to the in-tree bpftool path by default. For example,
bpftool selftests fail to build:
$ make -C tools/testing/selftests/bpf/
make: Entering directory '/home/tglozar/dev/linux/tools/testing/selftests/bpf'
make: *** No rule to make target 'bpftool', needed by '/home/tglozar/dev/linux/tools/testing/selftests/bpf/tools/include/vmlinux.h'. Stop.
make: Leaving directory '/home/tglozar/dev/linux/tools/testing/selftests/bpf'
Fix the problem by renaming the variable used for system bpftool from
BPFTOOL to SYSTEM_BPFTOOL, so that the new usage does not conflict with
the existing one of BPFTOOL.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Link: https://lore.kernel.org/20250326004018.248357-1-tglozar@redhat.com
Fixes: 8a635c3856 ("tools/build: Add bpftool-skeletons feature test")
Closes: https://lore.kernel.org/linux-kernel/5df6968a-2e5f-468e-b457-fc201535dd4c@linux.ibm.com/
Reported-by: Venkat Rao Bagalkote <venkat88@linux.ibm.com>
Tested-by: Venkat Rao Bagalkote <venkat88@linux.ibm.com>
Suggested-by: Quentin Monnet <qmo@kernel.org>
Acked-by: Quentin Monnet <qmo@kernel.org>
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
The functions osnoise_hist_main(), osnoise_top_main(),
timerlat_hist_main(), and timerlat_top_main() are lengthy and contain
duplicated code.
Refactor by consolidating the duplicate lines into the
save_trace_to_file() function.
Cc: Daniel Bristot de Oliveira <bristot@kernel.org>
Cc: John Kacur <jkacur@redhat.com>
Cc: "Luis Claudio R. Goncalves" <lgoncalv@redhat.com>
Cc: Eder Zulian <ezulian@redhat.com>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Link: https://lore.kernel.org/20250219115138.406075-1-costa.shul@redhat.com
Signed-off-by: Costa Shulyupin <costa.shul@redhat.com>
Reviewed-by: Tomas Glozar <tglozar@redhat.com>
Tested-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Using the RTLA_NO_BPF environmental variable, execute rtla-timerlat
tests both with and without BPF support to cover both paths.
If rtla is built without BPF or the osnoise:timerlat_sample trace event
is not available, test only the non-BPF path.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Link: https://lore.kernel.org/20250218145859.27762-9-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Collect samples using BPF program instead of pulling them from tracefs.
If the osnoise:timerlat_sample tracepoint is unavailable or the BPF
program fails to load for whatever reason, rtla falls back to the old
implementation.
The collection of samples using the BPF program is fully self-contained
and requires no activity of the userspace part of rtla during the
measurement. Thus, rtla only pulls the summary from the BPF map and
displays it every second, improving the performance.
In --aa-only mode, the BPF program does not collect any data and only
signalizes the end of tracing to userspace. An optimization that re-used
the main trace instance for auto-analysis in aa-only mode was dropped, as
rtla no longer turns tracing on in the main trace instance, making it
useless for auto-analysis.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Link: https://lore.kernel.org/20250218145859.27762-8-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Unlike timerlat-hist, timerlat-top applies the output divisor used to
set ns/us mode when printing results instead of applying it when
collecting the samples.
Move the application of the divisor from timerlat_top_print into
timerlat_top_update to make it consistent with timerlat-hist.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Link: https://lore.kernel.org/20250218145859.27762-7-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Collect samples using BPF program instead of pulling them from tracefs.
If the osnoise:timerlat_sample tracepoint is unavailable or the BPF
program fails to load for whatever reason, rtla falls back to the old
implementation.
The collection of samples using the BPF program is fully self-contained
and requires no activity of the userspace part of rtla during the
measurement. Thus, instead of waking up every second to collect samples,
rtla simply sleeps until woken up by a signal or threshold overflow.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Link: https://lore.kernel.org/20250218145859.27762-6-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Add BPF program that attaches to the osnoise:timerlat_sample tracepoint
and collects both the summary and the histogram (if requested) into BPF
maps (one map of each kind per context).
The program is designed to be used for both timerlat-top and
timerlat-hist. If using with timerlat-top, the "entries" parameter is
set to zero, which prevents the BPF program from recording histogram
entries. In that case, the maps for histograms do not have to be
created, as the BPF verifier will identify the code using them as
unreachable.
An IRQ or thread latency threshold might be supplied to stop recording
if hit, similar to the timerlat tracer threshold, which stops ftrace
tracing if hit. A BPF ringbuffer is used to signal threshold overflow to
userspace. In aa-only mode, this is the only function of the BPF
program.
Cc: John Kacur <jkacur@redhat.com>
Cc: Luis Goncalves <lgoncalv@redhat.com>
Cc: Gabriele Monaco <gmonaco@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Link: https://lore.kernel.org/20250218145859.27762-5-tglozar@redhat.com
Signed-off-by: Tomas Glozar <tglozar@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
