[PATCH v2 0/3] perf: fix some tests for musl

[PATCH v2 0/3] perf: fix some tests for musl

[Martin Kaiser]

Fix some perf tests for systems that use musl instead of glibc.

v2:
- add Namhyung's Acked-by for patch 1
- stricter flags check for patch 2
- more generic workload in patch 3

Martin Kaiser (3):
  perf test: fix nanosleep check in the ftrace test
  perf test: fix sys_enter_openat event test for musl
  perf test: fix "trace summary" test for musl-based systems

 tools/perf/tests/openat-syscall-tp-fields.c | 5 ++++-
 tools/perf/tests/shell/ftrace.sh            | 7 ++++---
 tools/perf/tests/shell/trace_summary.sh     | 2 +-
 3 files changed, 9 insertions(+), 5 deletions(-)

-- 
2.43.7

[PATCH v2 1/3] perf test: fix nanosleep check in the ftrace test

[Martin Kaiser]

The perf ftrace test case runs

perf ftrace profile --graph-opts depth=5 sleep 0.1

and checks that the output contains a *clock_nanosleep function with a
count of 1.

This fails on a risc-v system that uses musl as its C library. musl's
nanosleep syscall wrapper uses either the nanosleep or the
clock_nanosleep syscall.

Filter for sys_*nanosleep to allow both syscalls.

Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Acked-by: Namhyung Kim <namhyung@kernel.org>
---
v2:
- add Acked-by

 tools/perf/tests/shell/ftrace.sh | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/perf/tests/shell/ftrace.sh b/tools/perf/tests/shell/ftrace.sh
index 7f8aafcbb761..9f6e590f6437 100755
--- a/tools/perf/tests/shell/ftrace.sh
+++ b/tools/perf/tests/shell/ftrace.sh
@@ -71,9 +71,10 @@ test_ftrace_profile() {
     grep ^# "${output}"
     time_re="[[:space:]]+1[[:digit:]]{5}\.[[:digit:]]{3}"
     # 100283.000 100283.000 100283.000          1   __x64_sys_clock_nanosleep
-    # Check for one *clock_nanosleep line with a Count of just 1 that takes a bit more than 0.1 seconds
-    # Strip the _x64_sys part to work with other architectures
-    grep -E "^${time_re}${time_re}${time_re}[[:space:]]+1[[:space:]]+.*clock_nanosleep" "${output}"
+    # Check for one *sys_*nanosleep line with a Count of just 1 that takes a bit more than 0.1 seconds
+    # Strip the _x64_ part to work with other architectures, strip the clock part to support
+    # C libraries that use the nanosleep syscall instead of clock_nanosleep
+    grep -E "^${time_re}${time_re}${time_re}[[:space:]]+1[[:space:]]+.*sys_.*nanosleep" "${output}"
     echo "perf ftrace profile test  [Success]"
 }
 
-- 
2.43.7

[PATCH v2 2/3] perf test: fix sys_enter_openat event test for musl

[Martin Kaiser]

The "syscalls:sys_enter_openat event fields" test calls

openat(AT_FDCWD, "/etc/passwd", O_RDONLY | O_DIRECTORY)

and verifies that the flags of the captured event are matching.
This fails for musl, where the openat syscall wrapper always adds
O_LARGEFILE.

Update the check to allow for additional flags, the access mode flags
must be unchanged.

Signed-off-by: Martin Kaiser <martin@kaiser.cx>
---
v2:
- check that access mode is unchanged

 tools/perf/tests/openat-syscall-tp-fields.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/tests/openat-syscall-tp-fields.c b/tools/perf/tests/openat-syscall-tp-fields.c
index 2a139d2781a8..5523cf4e9321 100644
--- a/tools/perf/tests/openat-syscall-tp-fields.c
+++ b/tools/perf/tests/openat-syscall-tp-fields.c
@@ -120,7 +120,10 @@ static int test__syscall_openat_tp_fields(struct test_suite *test __maybe_unused
 
 				tp_flags = evsel__intval(evsel, &sample, "flags");
 				perf_sample__exit(&sample);
-				if (flags != tp_flags) {
+				/* C library wrapper may set additional flags,
+				   access mode must be unchanged */
+				if ((tp_flags & O_ACCMODE) != (flags & O_ACCMODE) ||
+				    (tp_flags & flags) != flags) {
 					pr_debug("%s: Expected flags=%#x, got %#x\n",
 						 __func__, flags, tp_flags);
 					goto out_delete_evlist;
-- 
2.43.7

[PATCH v2 3/3] perf test: fix "trace summary" test for musl-based systems

[Martin Kaiser]

The trace summary test calls /bin/true and filters for open, read and
close events. These events are coming from shared library loads.

On a musl system, the loader and libc may point to the same file. true
needs only libc, no further shared libraries are loaded at startup. The
test fails since no open, read and close events are captured.

root@host:~# ldd /bin/true
	/lib/ld-musl-riscv64.so.1 (0x3fb8882000)
	libc.so => /lib/ld-musl-riscv64.so.1 (0x3fb8882000)

root@host:~# file /lib/ld-musl-riscv64.so.1
/lib/ld-musl-riscv64.so.1: symbolic link to /usr/lib/libc.so

root@host:~# strace -f /bin/true
execve("/bin/true", ["/bin/true", ...], ... /* 18 vars */) = 1
set_tid_address(0x3fa1f7bf70)           = 330
mprotect(0x2ad6b8e000, 12288, PROT_READ) = 0
exit_group(0)                           = ?
+++ exited with 0 +++

Run "cat /dev/null" instead of "true". This creates the required events
regardless of the C library and it works for cat from busybox or from
coreutils.

Signed-off-by: Martin Kaiser <martin@kaiser.cx>
---
v2:
- use cat /dev/null, this works for busybox and coreutils

 tools/perf/tests/shell/trace_summary.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/tests/shell/trace_summary.sh b/tools/perf/tests/shell/trace_summary.sh
index 22e2651d5919..b80dea77cec6 100755
--- a/tools/perf/tests/shell/trace_summary.sh
+++ b/tools/perf/tests/shell/trace_summary.sh
@@ -14,7 +14,7 @@ OUTPUT=$(mktemp /tmp/perf_trace_test.XXXXX)
 
 test_perf_trace() {
     args=$1
-    workload="true"
+    workload="cat /dev/null"
     search="^\s*(open|read|close).*[0-9]+%$"
 
     echo "testing: perf trace ${args} -- ${workload}"
-- 
2.43.7

Re: [PATCH v2 0/3] perf: fix some tests for musl

[Namhyung Kim]

On Tue, Apr 28, 2026 at 04:38:45PM +0200, Martin Kaiser wrote:
> Fix some perf tests for systems that use musl instead of glibc.
> 
> v2:
> - add Namhyung's Acked-by for patch 1
> - stricter flags check for patch 2
> - more generic workload in patch 3
> 
> Martin Kaiser (3):
>   perf test: fix nanosleep check in the ftrace test
>   perf test: fix sys_enter_openat event test for musl
>   perf test: fix "trace summary" test for musl-based systems

Acked-by: Namhyung Kim <namhyung@kernel.org>

Thanks,
Namhyung

> 
>  tools/perf/tests/openat-syscall-tp-fields.c | 5 ++++-
>  tools/perf/tests/shell/ftrace.sh            | 7 ++++---
>  tools/perf/tests/shell/trace_summary.sh     | 2 +-
>  3 files changed, 9 insertions(+), 5 deletions(-)
> 
> -- 
> 2.43.7
> 

Re: [PATCH v2 2/3] perf test: fix sys_enter_openat event test for musl

[Ian Rogers]

On Tue, Apr 28, 2026 at 7:39 AM Martin Kaiser <martin@kaiser.cx> wrote:
>
> The "syscalls:sys_enter_openat event fields" test calls
>
> openat(AT_FDCWD, "/etc/passwd", O_RDONLY | O_DIRECTORY)
>
> and verifies that the flags of the captured event are matching.
> This fails for musl, where the openat syscall wrapper always adds
> O_LARGEFILE.
>
> Update the check to allow for additional flags, the access mode flags
> must be unchanged.
>
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>
> ---
> v2:
> - check that access mode is unchanged
>
>  tools/perf/tests/openat-syscall-tp-fields.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/tests/openat-syscall-tp-fields.c b/tools/perf/tests/openat-syscall-tp-fields.c
> index 2a139d2781a8..5523cf4e9321 100644
> --- a/tools/perf/tests/openat-syscall-tp-fields.c
> +++ b/tools/perf/tests/openat-syscall-tp-fields.c
> @@ -120,7 +120,10 @@ static int test__syscall_openat_tp_fields(struct test_suite *test __maybe_unused
>
>                                 tp_flags = evsel__intval(evsel, &sample, "flags");
>                                 perf_sample__exit(&sample);
> -                               if (flags != tp_flags) {
> +                               /* C library wrapper may set additional flags,
> +                                  access mode must be unchanged */

I think normal style would be:
```
/*
 * C library ...
 */
```

> +                               if ((tp_flags & O_ACCMODE) != (flags & O_ACCMODE) ||
> +                                   (tp_flags & flags) != flags) {

How can this second condition be true given the first condition?

Thanks,
Ian

>                                         pr_debug("%s: Expected flags=%#x, got %#x\n",
>                                                  __func__, flags, tp_flags);
>                                         goto out_delete_evlist;
> --
> 2.43.7
>

Re: [PATCH v2 3/3] perf test: fix "trace summary" test for musl-based systems

[Ian Rogers]

On Tue, Apr 28, 2026 at 7:39 AM Martin Kaiser <martin@kaiser.cx> wrote:
>
> The trace summary test calls /bin/true and filters for open, read and
> close events. These events are coming from shared library loads.
>
> On a musl system, the loader and libc may point to the same file. true
> needs only libc, no further shared libraries are loaded at startup. The
> test fails since no open, read and close events are captured.
>
> root@host:~# ldd /bin/true
>         /lib/ld-musl-riscv64.so.1 (0x3fb8882000)
>         libc.so => /lib/ld-musl-riscv64.so.1 (0x3fb8882000)
>
> root@host:~# file /lib/ld-musl-riscv64.so.1
> /lib/ld-musl-riscv64.so.1: symbolic link to /usr/lib/libc.so
>
> root@host:~# strace -f /bin/true
> execve("/bin/true", ["/bin/true", ...], ... /* 18 vars */) = 1
> set_tid_address(0x3fa1f7bf70)           = 330
> mprotect(0x2ad6b8e000, 12288, PROT_READ) = 0
> exit_group(0)                           = ?
> +++ exited with 0 +++
>
> Run "cat /dev/null" instead of "true". This creates the required events
> regardless of the C library and it works for cat from busybox or from
> coreutils.
>
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>
> ---
> v2:
> - use cat /dev/null, this works for busybox and coreutils
>
>  tools/perf/tests/shell/trace_summary.sh | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/perf/tests/shell/trace_summary.sh b/tools/perf/tests/shell/trace_summary.sh
> index 22e2651d5919..b80dea77cec6 100755
> --- a/tools/perf/tests/shell/trace_summary.sh
> +++ b/tools/perf/tests/shell/trace_summary.sh
> @@ -14,7 +14,7 @@ OUTPUT=$(mktemp /tmp/perf_trace_test.XXXXX)
>
>  test_perf_trace() {
>      args=$1
> -    workload="true"
> +    workload="cat /dev/null"

Could we capture why the unusual workload in a comment, I know it is
in the commit message.

Acked-by: Ian Rogers <irogers@google.com>

Thanks,
Ian

>      search="^\s*(open|read|close).*[0-9]+%$"
>
>      echo "testing: perf trace ${args} -- ${workload}"
> --
> 2.43.7
>

Re: [PATCH v2 2/3] perf test: fix sys_enter_openat event test for musl

[Martin Kaiser]

Thus wrote Ian Rogers (irogers@google.com):

> On Tue, Apr 28, 2026 at 7:39 AM Martin Kaiser <martin@kaiser.cx> wrote:

> > The "syscalls:sys_enter_openat event fields" test calls

> > openat(AT_FDCWD, "/etc/passwd", O_RDONLY | O_DIRECTORY)

> > and verifies that the flags of the captured event are matching.
> > This fails for musl, where the openat syscall wrapper always adds
> > O_LARGEFILE.

> > Update the check to allow for additional flags, the access mode flags
> > must be unchanged.

> > Signed-off-by: Martin Kaiser <martin@kaiser.cx>
> > ---
> > v2:
> > - check that access mode is unchanged

> >  tools/perf/tests/openat-syscall-tp-fields.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)

> > diff --git a/tools/perf/tests/openat-syscall-tp-fields.c b/tools/perf/tests/openat-syscall-tp-fields.c
> > index 2a139d2781a8..5523cf4e9321 100644
> > --- a/tools/perf/tests/openat-syscall-tp-fields.c
> > +++ b/tools/perf/tests/openat-syscall-tp-fields.c
> > @@ -120,7 +120,10 @@ static int test__syscall_openat_tp_fields(struct test_suite *test __maybe_unused

> >                                 tp_flags = evsel__intval(evsel, &sample, "flags");
> >                                 perf_sample__exit(&sample);
> > -                               if (flags != tp_flags) {
> > +                               /* C library wrapper may set additional flags,
> > +                                  access mode must be unchanged */

> I think normal style would be:
> ```
> /*
>  * C library ...
>  */
> ```

Ok, will fix this if we need another version.

> > +                               if ((tp_flags & O_ACCMODE) != (flags & O_ACCMODE) ||
> > +                                   (tp_flags & flags) != flags) {

> How can this second condition be true given the first condition?

Sorry if I am missing something obvious.

The O_ACCMODE check verifies that the two lower bits (access mode flags) in
the syscall and in the recorded event are identical.

The second check tests that all flags from the syscall are also set in the
event. Additional flags in the event are ok.

For tp_flags == 0x0 and flags == 0x4, the first condition is false, the second
one is true.

Best regards,
Martin

> Thanks,
> Ian

> >                                         pr_debug("%s: Expected flags=%#x, got %#x\n",
> >                                                  __func__, flags, tp_flags);
> >                                         goto out_delete_evlist;
> > --
> > 2.43.7