linux-stable.git/tools/tracing/rtla/src, branch linux-rolling-stable

rtla/actions: Restore continue flag in actions_perform()

2026-07-24T14:19:22+00:00

[ Upstream commit dd520daffbc901f10d49a51a58313547d417b506 ]

Currently, actions_perform() only ever sets the continue flag (when
performing the continue action), but never resets it. That leads to
RTLA continuing tracing even if the continue action was not performed in
the current iteration.

For example, the following command:

$ rtla timerlat hist -T 100 --on-threshold shell,command='
    echo Spike!
    if [ -f /tmp/a ]
    then
      exit 1
    else
      touch /tmp/a
    fi' --on-threshold continue

should print Spike! at most once, because after hitting the threshold
for the first time, /tmp/a exists, the shell action will fail, and the
continue action is not performed. However, unless /tmp/a exists before
the measurement, it will print Spike! until stopped, as the continue
flag stays set.

Set the continue flag to false in the beginning of actions_perform() to
make RTLA continue only if the action was actually performed.

Fixes: 8d933d5c89e8 ("rtla/timerlat: Add continue action")
Link: https://lore.kernel.org/r/20260526102523.2662391-1-tglozar@redhat.com
[ correct Fixes tag to include 12 characters of hash ]
Signed-off-by: Tomas Glozar 
Signed-off-by: Sasha Levin

rtla: Stop the record trace on interrupt

2026-07-24T14:19:14+00:00

[ Upstream commit f03a59f949176ce4312cb466245d1243aaf40389 ]

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 
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 
Signed-off-by: Sasha Levin

tools/rtla: Fix --dump-tasks usage in timerlat

2026-07-24T14:19:14+00:00

[ Upstream commit 704fe8f3d97bba842f5f357e317116f1b88169e8 ]

Fix --dump-task to --dump-tasks in timerlat_hist usage string
and getopt_long table for consistency with timerlat_top.

Add missing --dump-tasks to timerlat_top usage synopsis.

Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Costa Shulyupin 
Fixes: 2091336b9a8b ("rtla/timerlat_hist: Add auto-analysis support")
Link: https://lore.kernel.org/r/20260414185223.65353-1-costa.shul@redhat.com
Signed-off-by: Tomas Glozar 
Signed-off-by: Sasha Levin

rtla: Fix parsing of multi-character short options

2026-06-04T08:53:25+00:00

A bug was reported where the parsing of multi-character short options,
be it a short option with an argument specified without space (e.g.
"-p100") or multiple short options in one argument (e.g. -un), ignores
options specific to individual tools.

Furthermore, if the rest of the option is supposed to be an argument, it
gets reinterpreted as a string of options. For example, -p100 gets
interpreted as -100, which is due to hackish implementation read as
--no-thread --no-irq --no-irq with timerlat hist, causing rtla to error
out:

$ rtla timerlat hist -p100
no-irq and no-thread set, there is nothing to do here

This behavior is caused by getopt_long() being called twice on each
argument, once in common_parse_options(), once in [tool]_parse_args():

- common_parse_options() calls getopt_long() with an array of options
  common for all rtla tools, while suppressing errors (opterr = 0).
- If the option fails to parse, common_parse_options() returns 0.
- If 0 is returned from common_parse_options(), [tool]_parse_args()
  calls getopt_long() again, with its own set of options.

* [tool] means one of {osnoise,timerlat}_{top,hist}

At least in glibc, getopt_long() increments its internal nextchar
variable even if the option is not recognized. That means that in the
case of "-p100", common_parse_options() sets nextchar pointing to '1',
and timerlat_hist_parse_args() sees '1', not 'p'; the same then repeats
for the first and second '0'.

As there is no way to restore the correct internal state of
getopt_long() reliably, fix the issue by merging the common options back
to the longopt array and option string of the [tool]_parse_args()
functions using a macro; only the switch part is left in the original
function, which is renamed to set_common_option().

Fixes: 850cd24cb6d6 ("tools/rtla: Add common_parse_options()")
Reported-by: John Kacur 
Tested-by: John Kacur 
Link: https://lore.kernel.org/r/20260602125506.3325345-1-tglozar@redhat.com
Signed-off-by: Tomas Glozar

Merge tag 'trace-rtla-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace

2026-04-16T00:48:24+00:00

Pull RTLA updates from Steven Rostedt:

- Simplify option parsing

Auto-generate getopt_long() optstring for short options from long
options array, avoiding the need to specify it manually and reducing
the surface for mistakes.

- Add unit tests

Implement unit tests (make unit-tests) using libcheck, next to
existing runtime tests (make check). Currently, three functions from
utils.c are tested.

- Add --stack-format option

In addition to stopping stack pointer decoding (with -s/--stack
option) on first unresolvable pointer, allow also skipping
unresolvable pointers and displaying everything, configurable with a
new option.

- Unify number of CPUs into one global variable

Use one global variable, nr_cpus, to store the number of CPUs instead
of retrieving it and passing it at multiple places.

- Fix behavior in various corner cases

Make RTLA behave correctly in several corner cases: memory allocation
failure, invalid value read from kernel side, thread creation
failure, malformed time value input, and read/write failure or
interruption by signal.

- Improve string handling

Simplify several places in the code that handle strings, including
parsing of action arguments. A few new helper functions and variables
are added for that purpose.

- Get rid of magic numbers

Few places handling paths use a magic number of 1024. Replace it with
MAX_PATH and ARRAY_SIZE() macro.

- Unify threshold handling

Code that handles response to latency threshold is duplicated between
tools, which has led to bugs in the past. Unify it into a new helper
as much as possible.

- Fix segfault on SIGINT during cleanup

The SIGINT handler touches dynamically allocated memory. Detach it
before freeing it during cleanup to prevent segmentation fault and
discarding of output buffers. Also, properly document SIGINT handling
while at it.

* tag 'trace-rtla-v7.1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (28 commits)
Documentation/rtla: Document SIGINT behavior
rtla: Fix segfault on multiple SIGINTs
rtla/utils: Fix loop condition in PID validation
rtla/utils: Fix resource leak in set_comm_sched_attr()
rtla/trace: Fix I/O handling in save_trace_to_file()
rtla/trace: Fix write loop in trace_event_save_hist()
rtla/timerlat: Simplify RTLA_NO_BPF environment variable check
rtla: Use str_has_prefix() for option prefix check
rtla: Enforce exact match for time unit suffixes
rtla: Use str_has_prefix() for prefix checks
rtla: Add str_has_prefix() helper function
rtla: Handle pthread_create() failure properly
rtla/timerlat: Add bounds check for softirq vector
rtla: Simplify code by caching string lengths
rtla: Replace magic number with MAX_PATH
rtla: Introduce common_threshold_handler() helper
rtla/actions: Simplify argument parsing
rtla: Use strdup() to simplify code
rtla: Exit on memory allocation failures during initialization
tools/rtla: Remove unneeded nr_cpus from for_each_monitored_cpu
...

rtla: Fix build without libbpf header

2026-03-30T16:44:48+00:00

rtla supports building without libbpf. However, BPF actions
patchset [1] adds an include of bpf/libbpf.h into timerlat_bpf.h,
which breaks build on systems that don't have libbpf headers
installed.

This is a leftover from a draft version of the patchset where
timerlat_bpf_set_action() (which takes a struct bpf_program * argument)
was defined in the header. timerlat_bpf.c already includes bpf/libbpf.h
via timerlat.skel.h when libbpf is present.

Remove the redundant include to fix build on systems without libbpf
headers.

[1] https://lore.kernel.org/linux-trace-kernel/20251126144205.331954-1-tglozar@redhat.com/T/

Cc: John Kacur 
Cc: Luis Goncalves 
Cc: Crystal Wood 
Cc: Costa Shulyupin 
Link: https://patch.msgid.link/20260330091207.16184-1-tglozar@redhat.com
Reported-by: Steven Rostedt (Google) 
Closes: https://lore.kernel.org/linux-trace-kernel/20260329122202.65a8b575@robin/
Fixes: 8cd0f08ac72e ("rtla/timerlat: Support tail call from BPF program")
Signed-off-by: Tomas Glozar 
Reviewed-by: Wander Lairson Costa 
Signed-off-by: Steven Rostedt (Google)

rtla: Fix segfault on multiple SIGINTs

2026-03-20T13:41:36+00:00

Detach stop_trace() from SIGINT/SIGALRM on tool clean-up to prevent it
from crashing RTLA by accessing freed memory.

This prevents a crash when multiple SIGINTs are received.

Fixes: d6899e560366 ("rtla/timerlat_hist: Abort event processing on second signal")
Fixes: 80967b354a76 ("rtla/timerlat_top: Abort event processing on second signal")
Reviewed-by: Wander Lairson Costa 
Link: https://lore.kernel.org/r/20260310160725.144443-1-tglozar@redhat.com
Signed-off-by: Tomas Glozar

rtla/utils: Fix loop condition in PID validation

2026-03-11T14:29:50+00:00

The procfs_is_workload_pid() function iterates through a directory
entry name to validate if it represents a process ID. The loop
condition checks if the pointer t_name is non-NULL, but since
incrementing a pointer never makes it NULL, this condition is always
true within the loop's context. Although the inner isdigit() check
catches the NUL terminator and breaks out of the loop, the condition
is semantically misleading and not idiomatic for C string processing.

Correct the loop condition from checking the pointer (t_name) to
checking the character it points to (*t_name). This ensures the loop
terminates when the NUL terminator is reached, aligning with standard
C string iteration practices. While the original code functioned
correctly due to the existing character validation, this change
improves code clarity and maintainability.

Signed-off-by: Wander Lairson Costa 
Link: https://lore.kernel.org/r/20260309195040.1019085-19-wander@redhat.com
Signed-off-by: Tomas Glozar

rtla/utils: Fix resource leak in set_comm_sched_attr()

2026-03-11T14:29:50+00:00

The set_comm_sched_attr() function opens the /proc directory via
opendir() but fails to call closedir() on its successful exit path.
If the function iterates through all processes without error, it
returns 0 directly, leaking the DIR stream pointer.

Fix this by refactoring the function to use a single exit path. A
retval variable is introduced to track the success or failure status.
All exit points now jump to a unified out label that calls closedir()
before the function returns, ensuring the resource is always freed.

Fixes: dada03db9bb19 ("rtla: Remove procps-ng dependency")
Signed-off-by: Wander Lairson Costa 
Link: https://lore.kernel.org/r/20260309195040.1019085-18-wander@redhat.com
Signed-off-by: Tomas Glozar

rtla/trace: Fix I/O handling in save_trace_to_file()

2026-03-11T14:29:50+00:00

The read/write loop in save_trace_to_file() does not correctly handle
errors from the read() and write() system calls. If either call is
interrupted by a signal, it returns -1 with errno set to EINTR, but
the code treats this as a fatal error and aborts the save operation.
Additionally, write() may perform a partial write, returning fewer
bytes than requested, which the code does not handle.

Fix the I/O loop by introducing proper error handling. The return
value of read() is now stored in a ssize_t variable and checked for
errors, with EINTR causing a retry. For write(), an inner loop ensures
all bytes are written, handling both EINTR and partial writes. Error
messages now include strerror() output for better debugging.

This follows the same pattern established in the previous commit that
fixed trace_event_save_hist(), ensuring consistent and robust I/O
handling throughout the trace saving code.

Signed-off-by: Wander Lairson Costa 
Link: https://lore.kernel.org/r/20260309195040.1019085-17-wander@redhat.com
Signed-off-by: Tomas Glozar