[mainline]selftests/sched_ext: select_cpu_dfl_nodispatch test is failing.

[mainline]selftests/sched_ext: select_cpu_dfl_nodispatch test is failing.

[Samir M]

Hi,


I'm reporting a test failure in the sched_ext selftests on kernel 
version 7.0.0-rc7.


Test Details:
=============
Test Path: linux/tools/testing/selftests/sched_ext
Test Name: select_cpu_dfl_nodispatch
Command: ./runner -t select_cpu_dfl_nodispatch


Test Description:
================
The test is designed to verify the behavior of scx_bpf_select_cpu_dfl() 
in ops.select_cpu()

Failure Details:
================
The test fails with the following output:

./runner -t select_cpu_dfl_nodispatch
===== START =====
TEST: select_cpu_dfl_nodispatch
DESCRIPTION: Verify behavior of scx_bpf_select_cpu_dfl() in ops.select_cpu()
OUTPUT:
ERR: select_cpu_dfl_nodispatch.c:53
Expected skel->bss->saw_local to be true (0)
not ok 1 select_cpu_dfl_nodispatch #
=====  END  =====

RESULTS:
PASSED:  0
SKIPPED: 0
FAILED:  1

Reproduction:
=============
1. Clone the upstream Linux kernel repository
2. Enable CONFIG_SCHED_CLASS_EXT in kernel configuration
3. Build and boot the kernel with sched_ext enabled
4. Navigate to linux/tools/testing/selftests/sched_ext
5. Build the selftests: make
6. Run the test: ./runner -t select_cpu_dfl_nodispatch

Environment:
============
Kernel Version: 7.0.0-rc7
Architecture: ppc64le

Could you please investigate this test failure? I'm happy to provide 
additional
information or test on different configurations if needed.

If you happen to fix the above issue, then please add below tag.
Reported-by: Samir M <samir@linux.ibm.com>

Regards,
Samir.

Re: [mainline]selftests/sched_ext: select_cpu_dfl_nodispatch test is failing.

[Cheng-Yang Chou]

Hi Samir,

On Mon, Apr 13, 2026 at 06:37:08PM +0530, Samir M wrote:
> I'm reporting a test failure in the sched_ext selftests on kernel version
> 7.0.0-rc7.
> 
> 
> Test Details:
> =============
> Test Path: linux/tools/testing/selftests/sched_ext
> Test Name: select_cpu_dfl_nodispatch
> Command: ./runner -t select_cpu_dfl_nodispatch
> 
> 
> Test Description:
> ================
> The test is designed to verify the behavior of scx_bpf_select_cpu_dfl() in
> ops.select_cpu()
> 
> Failure Details:
> ================
> The test fails with the following output:
> 
> ./runner -t select_cpu_dfl_nodispatch
> ===== START =====
> TEST: select_cpu_dfl_nodispatch
> DESCRIPTION: Verify behavior of scx_bpf_select_cpu_dfl() in ops.select_cpu()
> OUTPUT:
> ERR: select_cpu_dfl_nodispatch.c:53
> Expected skel->bss->saw_local to be true (0)

CC sched_ext mailing list,

I suspect this is an arch-specific bug (and perhaps not a sched_ext's bug).
I've verified that on 7.0.0-rc7, all sched_ext selftests pass on x86_64.
FYI, they also pass on kernels 6.14 and 6.18 on x86_64.

Thanks,
Cheng-Yang

> not ok 1 select_cpu_dfl_nodispatch #
> =====  END  =====
> 
> RESULTS:
> PASSED:  0
> SKIPPED: 0
> FAILED:  1
> 
> Reproduction:
> =============
> 1. Clone the upstream Linux kernel repository
> 2. Enable CONFIG_SCHED_CLASS_EXT in kernel configuration
> 3. Build and boot the kernel with sched_ext enabled
> 4. Navigate to linux/tools/testing/selftests/sched_ext
> 5. Build the selftests: make
> 6. Run the test: ./runner -t select_cpu_dfl_nodispatch
> 
> Environment:
> ============
> Kernel Version: 7.0.0-rc7
> Architecture: ppc64le
> 
> Could you please investigate this test failure? I'm happy to provide
> additional
> information or test on different configurations if needed.
> 
> If you happen to fix the above issue, then please add below tag.
> Reported-by: Samir M <samir@linux.ibm.com>

Re: [mainline]selftests/sched_ext: select_cpu_dfl_nodispatch test is failing.

[Cheng-Yang Chou]

On Mon, Apr 13, 2026 at 11:55:27PM +0800, Cheng-Yang Chou wrote:
> 
> CC sched_ext mailing list,
> 
> I suspect this is an arch-specific bug (and perhaps not a sched_ext's bug).
> I've verified that on 7.0.0-rc7, all sched_ext selftests pass on x86_64.
> FYI, they also pass on kernels 6.14 and 6.18 on x86_64.
       ^^^^
       I meant select_cpu_dfl_nodispatch specifically. Sorry for the confusion.

-- 
Thanks,
Cheng-Yang