Re: [PATCH v1] selftests/sched_ext: Fix rt_stall flaky failure

[Andrea Righi <arighi@nvidia.com>]

Hi Ihor,

On Fri, Feb 13, 2026 at 10:21:36AM -0800, Ihor Solodrai wrote:
> The rt_stall test measures the runtime ratio between an EXT and an RT
> task pinned to the same CPU, verifying that the deadline server prevents
> RT tasks from starving SCHED_EXT tasks. It expects the EXT task to get
> at least 4% of CPU time.
> 
> The test is flaky because sched_stress_test() calls sleep(RUN_TIME)
> immediately after fork(), without waiting for the RT child to complete
> its setup (set_affinity + set_sched). If the RT child experiences
> scheduling latency before completing setup, that delay eats into the
> measurement window: the RT child runs for less than RUN_TIME seconds,
> and the EXT task's measured ratio drops below the 4% threshold.
> 
> For example, in the failing CI run [1]:
>   EXT=0.140s RT=4.750s total=4.890s (expected ~5.0s)
>   ratio=2.86% < 4% → FAIL
> 
> The 110ms gap (5.0 - 4.89) corresponds to the RT child's setup time
> being counted inside the measurement window, during which fewer
> deadline server ticks fire for the EXT task.
> 
> Fix by using pipes to synchronize: each child signals the parent after
> completing its setup, and the parent waits for both signals before
> starting sleep(RUN_TIME). This ensures the measurement window only
> counts time when both tasks are fully configured and competing.
> 
> [1] https://github.com/kernel-patches/bpf/actions/runs/21961895809/job/63442490449
> 
> Fixes: be621a76341c ("selftests/sched_ext: Add test for sched_ext dl_server")
> Assisted-by: claude-opus-4-6-v1
> Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>

The pipe sync makes sense and should make the test more robust, so LGTM.

Reviewed-by: Andrea Righi <arighi@nvidia.com>

Thanks,
-Andrea

> 
> ---
> 
> BPF CI caught the failure, I fed the logs to Claude Code and this is
> what it came up with. I ran this 3 times on CI, and the failure didn't
> happen again. The change makes sense to me, although synchronization
> via pipes might be an overkill?
> 
> Please let me know if this is too sloppy, I'll try to refine.
> 
> Thanks!
> 
> ---
>  tools/testing/selftests/sched_ext/rt_stall.c | 49 ++++++++++++++++++++
>  1 file changed, 49 insertions(+)
> 
> diff --git a/tools/testing/selftests/sched_ext/rt_stall.c b/tools/testing/selftests/sched_ext/rt_stall.c
> index 015200f80f6e..ab772e336f86 100644
> --- a/tools/testing/selftests/sched_ext/rt_stall.c
> +++ b/tools/testing/selftests/sched_ext/rt_stall.c
> @@ -23,6 +23,30 @@
>  #define CORE_ID		0	/* CPU to pin tasks to */
>  #define RUN_TIME        5	/* How long to run the test in seconds */
>  
> +/* Signal the parent that setup is complete by writing to a pipe */
> +static void signal_ready(int fd)
> +{
> +	char c = 1;
> +
> +	if (write(fd, &c, 1) != 1) {
> +		perror("write to ready pipe");
> +		exit(EXIT_FAILURE);
> +	}
> +	close(fd);
> +}
> +
> +/* Wait for a child to signal readiness via a pipe */
> +static void wait_ready(int fd)
> +{
> +	char c;
> +
> +	if (read(fd, &c, 1) != 1) {
> +		perror("read from ready pipe");
> +		exit(EXIT_FAILURE);
> +	}
> +	close(fd);
> +}
> +
>  /* Simple busy-wait function for test tasks */
>  static void process_func(void)
>  {
> @@ -122,14 +146,24 @@ static bool sched_stress_test(bool is_ext)
>  
>  	float ext_runtime, rt_runtime, actual_ratio;
>  	int ext_pid, rt_pid;
> +	int ext_ready[2], rt_ready[2];
>  
>  	ksft_print_header();
>  	ksft_set_plan(1);
>  
> +	if (pipe(ext_ready) || pipe(rt_ready)) {
> +		perror("pipe");
> +		ksft_exit_fail();
> +	}
> +
>  	/* Create and set up a EXT task */
>  	ext_pid = fork();
>  	if (ext_pid == 0) {
> +		close(ext_ready[0]);
> +		close(rt_ready[0]);
> +		close(rt_ready[1]);
>  		set_affinity(CORE_ID);
> +		signal_ready(ext_ready[1]);
>  		process_func();
>  		exit(0);
>  	} else if (ext_pid < 0) {
> @@ -140,8 +174,12 @@ static bool sched_stress_test(bool is_ext)
>  	/* Create an RT task */
>  	rt_pid = fork();
>  	if (rt_pid == 0) {
> +		close(ext_ready[0]);
> +		close(ext_ready[1]);
> +		close(rt_ready[0]);
>  		set_affinity(CORE_ID);
>  		set_sched(SCHED_FIFO, 50);
> +		signal_ready(rt_ready[1]);
>  		process_func();
>  		exit(0);
>  	} else if (rt_pid < 0) {
> @@ -149,6 +187,17 @@ static bool sched_stress_test(bool is_ext)
>  		ksft_exit_fail();
>  	}
>  
> +	/*
> +	 * Wait for both children to complete their setup (affinity and
> +	 * scheduling policy) before starting the measurement window.
> +	 * This prevents flaky failures caused by the RT child's setup
> +	 * time eating into the measurement period.
> +	 */
> +	close(ext_ready[1]);
> +	close(rt_ready[1]);
> +	wait_ready(ext_ready[0]);
> +	wait_ready(rt_ready[0]);
> +
>  	/* Let the processes run for the specified time */
>  	sleep(RUN_TIME);
>  
> -- 
> 2.53.0
>