[PATCH v4 0/2] sched_ext: Update demo schedulers and selftests for deprecated APIs

[PATCH v4 0/2] sched_ext: Update demo schedulers and selftests for deprecated APIs

[Cheng-Yang Chou]

Two sets of sched_ext APIs have been deprecated:

- Direct writes to p->scx.dsq_vtime in favor of 
  scx_bpf_task_set_dsq_vtime()
- ops.cpu_acquire/release() in favor of handling CPU preemption via the
  sched_switch tracepoint, as introduced by commit a3f5d4822253 
  ("sched_ext: Allow scx_bpf_reenqueue_local() to be called from 
  anywhere")

This series updates the demo schedulers (scx_simple, scx_flatcg) and
selftests (select_cpu_vtime, maximal) to use the new APIs, keeping them
in sync with current best practices.

Patch 1 updates scx_simple, scx_flatcg, and select_cpu_vtime to use
scx_bpf_task_set_dsq_vtime() with scale_by_task_weight_inverse(), with
the delta computation moved inside the if block.

Patch 2 replaces the cpu_acquire/release stubs in the maximal selftest
with a minimal sched_switch TP program, and fixes the maximal and
reload_loop tests to properly attach the sched_switch tracepoint via
bpf_map__set_autoattach() and maximal__attach(), as both tests use the
maximal skeleton.

Changes in v4:
- Drop scx_qmap changes from patch 2, already applied in commit 
  3229ac4a5ef5 ("sched_ext: Add SCX_OPS_ALWAYS_ENQ_IMMED ops flag")
- Move delta computation inside the if block (Tejun Heo)
- Link to v3:
  https://lore.kernel.org/all/20260313014944.1234667-1-yphbchou0911@gmail.com/

Changes in v3:
- Use scale_by_task_weight_inverse() instead of scale_by_task_weight()
  (Andrea Righi)
- Link to v2:
  https://lore.kernel.org/all/20260312175527.1220540-1-yphbchou0911@gmail.com/

Changes in v2:
- Use scx_bpf_task_set_dsq_vtime() with scale_by_task_weight instead
  of direct assignment and remove redundant bpf_ksym_exists() logic
  (Andrea Righi)
- Mention commit a3f5d4822253 ("sched_ext: Allow
  scx_bpf_reenqueue_local() to be called from anywhere") in the commit
  message to clarify why ops.cpu_acquire/release() are being deprecated
  (Andrea Righi)
- Fix maximal.c and reload_loop.c to actually attach the sched_switch
  tracepoint by calling bpf_map__set_autoattach() and maximal__attach()
  (Andrea Righi)
- Link to v1:
  https://lore.kernel.org/all/20260312042001.955675-1-yphbchou0911@gmail.com/

Thanks,
Cheng-Yang

---

Cheng-Yang Chou (2):
  sched_ext: Update demo schedulers and selftests to use
    scx_bpf_task_set_dsq_vtime()
  sched_ext: Update selftests to drop ops.cpu_acquire/release()

 tools/sched_ext/scx_flatcg.bpf.c                  | 12 +++++++-----
 tools/sched_ext/scx_simple.bpf.c                  |  6 ++++--
 tools/testing/selftests/sched_ext/maximal.bpf.c   | 15 ++++++---------
 tools/testing/selftests/sched_ext/maximal.c       |  3 +++
 tools/testing/selftests/sched_ext/reload_loop.c   |  3 +++
 .../selftests/sched_ext/select_cpu_vtime.bpf.c    |  7 +++++--
 6 files changed, 28 insertions(+), 18 deletions(-)

-- 
2.48.1

[PATCH v4 1/2] sched_ext: Update demo schedulers and selftests to use scx_bpf_task_set_dsq_vtime()

[Cheng-Yang Chou]

Direct writes to p->scx.dsq_vtime are deprecated in favor of
scx_bpf_task_set_dsq_vtime(). Update scx_simple, scx_flatcg, and
select_cpu_vtime selftest to use the new kfunc with
scale_by_task_weight_inverse().

Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
---
 tools/sched_ext/scx_flatcg.bpf.c                     | 12 +++++++-----
 tools/sched_ext/scx_simple.bpf.c                     |  6 ++++--
 .../selftests/sched_ext/select_cpu_vtime.bpf.c       |  7 +++++--
 3 files changed, 16 insertions(+), 9 deletions(-)

diff --git a/tools/sched_ext/scx_flatcg.bpf.c b/tools/sched_ext/scx_flatcg.bpf.c
index 1351377f64d5..6d3a028d7b59 100644
--- a/tools/sched_ext/scx_flatcg.bpf.c
+++ b/tools/sched_ext/scx_flatcg.bpf.c
@@ -551,9 +551,11 @@ void BPF_STRUCT_OPS(fcg_stopping, struct task_struct *p, bool runnable)
 	 * too much, determine the execution time by taking explicit timestamps
 	 * instead of depending on @p->scx.slice.
 	 */
-	if (!fifo_sched)
-		p->scx.dsq_vtime +=
-			(SCX_SLICE_DFL - p->scx.slice) * 100 / p->scx.weight;
+	if (!fifo_sched) {
+		u64 delta = scale_by_task_weight_inverse(p, SCX_SLICE_DFL - p->scx.slice);
+
+		scx_bpf_task_set_dsq_vtime(p, p->scx.dsq_vtime + delta);
+	}
 
 	taskc = bpf_task_storage_get(&task_ctx, p, 0, 0);
 	if (!taskc) {
@@ -822,7 +824,7 @@ s32 BPF_STRUCT_OPS(fcg_init_task, struct task_struct *p,
 	if (!(cgc = find_cgrp_ctx(args->cgroup)))
 		return -ENOENT;
 
-	p->scx.dsq_vtime = cgc->tvtime_now;
+	scx_bpf_task_set_dsq_vtime(p, cgc->tvtime_now);
 
 	return 0;
 }
@@ -924,7 +926,7 @@ void BPF_STRUCT_OPS(fcg_cgroup_move, struct task_struct *p,
 		return;
 
 	delta = time_delta(p->scx.dsq_vtime, from_cgc->tvtime_now);
-	p->scx.dsq_vtime = to_cgc->tvtime_now + delta;
+	scx_bpf_task_set_dsq_vtime(p, to_cgc->tvtime_now + delta);
 }
 
 s32 BPF_STRUCT_OPS_SLEEPABLE(fcg_init)
diff --git a/tools/sched_ext/scx_simple.bpf.c b/tools/sched_ext/scx_simple.bpf.c
index 9ad6f0949987..cc40552b2b5f 100644
--- a/tools/sched_ext/scx_simple.bpf.c
+++ b/tools/sched_ext/scx_simple.bpf.c
@@ -121,12 +121,14 @@ void BPF_STRUCT_OPS(simple_stopping, struct task_struct *p, bool runnable)
 	 * too much, determine the execution time by taking explicit timestamps
 	 * instead of depending on @p->scx.slice.
 	 */
-	p->scx.dsq_vtime += (SCX_SLICE_DFL - p->scx.slice) * 100 / p->scx.weight;
+	u64 delta = scale_by_task_weight_inverse(p, SCX_SLICE_DFL - p->scx.slice);
+
+	scx_bpf_task_set_dsq_vtime(p, p->scx.dsq_vtime + delta);
 }
 
 void BPF_STRUCT_OPS(simple_enable, struct task_struct *p)
 {
-	p->scx.dsq_vtime = vtime_now;
+	scx_bpf_task_set_dsq_vtime(p, vtime_now);
 }
 
 s32 BPF_STRUCT_OPS_SLEEPABLE(simple_init)
diff --git a/tools/testing/selftests/sched_ext/select_cpu_vtime.bpf.c b/tools/testing/selftests/sched_ext/select_cpu_vtime.bpf.c
index bfcb96cd4954..a2c6be98b81b 100644
--- a/tools/testing/selftests/sched_ext/select_cpu_vtime.bpf.c
+++ b/tools/testing/selftests/sched_ext/select_cpu_vtime.bpf.c
@@ -66,12 +66,15 @@ void BPF_STRUCT_OPS(select_cpu_vtime_running, struct task_struct *p)
 void BPF_STRUCT_OPS(select_cpu_vtime_stopping, struct task_struct *p,
 		    bool runnable)
 {
-	p->scx.dsq_vtime += (SCX_SLICE_DFL - p->scx.slice) * 100 / p->scx.weight;
+	u64 delta = scale_by_task_weight_inverse(p, SCX_SLICE_DFL - p->scx.slice);
+
+	scx_bpf_task_set_dsq_vtime(p, p->scx.dsq_vtime + delta);
+
 }
 
 void BPF_STRUCT_OPS(select_cpu_vtime_enable, struct task_struct *p)
 {
-	p->scx.dsq_vtime = vtime_now;
+	scx_bpf_task_set_dsq_vtime(p, vtime_now);
 }
 
 s32 BPF_STRUCT_OPS_SLEEPABLE(select_cpu_vtime_init)
-- 
2.48.1

[PATCH v4 2/2] sched_ext: Update selftests to drop ops.cpu_acquire/release()

[Cheng-Yang Chou]

ops.cpu_acquire/release() are deprecated by commit a3f5d4822253
("sched_ext: Allow scx_bpf_reenqueue_local() to be called from
anywhere") in favor of handling CPU preemption via the sched_switch
tracepoint.

In the maximal selftest, replace the cpu_acquire/release stubs with a
minimal sched_switch TP program. Attach all non-struct_ops programs
(including the new TP) via maximal__attach() after disabling auto-attach
for the maximal_ops struct_ops map, which is managed manually in run().

Apply the same fix to reload_loop, which also uses the maximal skeleton.

Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
---
 tools/testing/selftests/sched_ext/maximal.bpf.c | 15 ++++++---------
 tools/testing/selftests/sched_ext/maximal.c     |  3 +++
 tools/testing/selftests/sched_ext/reload_loop.c |  3 +++
 3 files changed, 12 insertions(+), 9 deletions(-)

diff --git a/tools/testing/selftests/sched_ext/maximal.bpf.c b/tools/testing/selftests/sched_ext/maximal.bpf.c
index 01cf4f3da4e0..5858f64313e9 100644
--- a/tools/testing/selftests/sched_ext/maximal.bpf.c
+++ b/tools/testing/selftests/sched_ext/maximal.bpf.c
@@ -67,13 +67,12 @@ void BPF_STRUCT_OPS(maximal_set_cpumask, struct task_struct *p,
 void BPF_STRUCT_OPS(maximal_update_idle, s32 cpu, bool idle)
 {}
 
-void BPF_STRUCT_OPS(maximal_cpu_acquire, s32 cpu,
-		    struct scx_cpu_acquire_args *args)
-{}
-
-void BPF_STRUCT_OPS(maximal_cpu_release, s32 cpu,
-		    struct scx_cpu_release_args *args)
-{}
+SEC("tp_btf/sched_switch")
+int BPF_PROG(maximal_sched_switch, bool preempt, struct task_struct *prev,
+	     struct task_struct *next, unsigned long prev_state)
+{
+	return 0;
+}
 
 void BPF_STRUCT_OPS(maximal_cpu_online, s32 cpu)
 {}
@@ -150,8 +149,6 @@ struct sched_ext_ops maximal_ops = {
 	.set_weight		= (void *) maximal_set_weight,
 	.set_cpumask		= (void *) maximal_set_cpumask,
 	.update_idle		= (void *) maximal_update_idle,
-	.cpu_acquire		= (void *) maximal_cpu_acquire,
-	.cpu_release		= (void *) maximal_cpu_release,
 	.cpu_online		= (void *) maximal_cpu_online,
 	.cpu_offline		= (void *) maximal_cpu_offline,
 	.init_task		= (void *) maximal_init_task,
diff --git a/tools/testing/selftests/sched_ext/maximal.c b/tools/testing/selftests/sched_ext/maximal.c
index c6be50a9941d..1dc369224670 100644
--- a/tools/testing/selftests/sched_ext/maximal.c
+++ b/tools/testing/selftests/sched_ext/maximal.c
@@ -19,6 +19,9 @@ static enum scx_test_status setup(void **ctx)
 	SCX_ENUM_INIT(skel);
 	SCX_FAIL_IF(maximal__load(skel), "Failed to load skel");
 
+	bpf_map__set_autoattach(skel->maps.maximal_ops, false);
+	SCX_FAIL_IF(maximal__attach(skel), "Failed to attach skel");
+
 	*ctx = skel;
 
 	return SCX_TEST_PASS;
diff --git a/tools/testing/selftests/sched_ext/reload_loop.c b/tools/testing/selftests/sched_ext/reload_loop.c
index 308211d80436..49297b83d748 100644
--- a/tools/testing/selftests/sched_ext/reload_loop.c
+++ b/tools/testing/selftests/sched_ext/reload_loop.c
@@ -23,6 +23,9 @@ static enum scx_test_status setup(void **ctx)
 	SCX_ENUM_INIT(skel);
 	SCX_FAIL_IF(maximal__load(skel), "Failed to load skel");
 
+	bpf_map__set_autoattach(skel->maps.maximal_ops, false);
+	SCX_FAIL_IF(maximal__attach(skel), "Failed to attach skel");
+
 	return SCX_TEST_PASS;
 }
 
-- 
2.48.1

Re: [PATCH v4 0/2] sched_ext: Update demo schedulers and selftests for deprecated APIs

[Tejun Heo]

Hello,

Patch 1 looks fine. For patch 2, ops.cpu_acquire/release() are already
removed from the tree, so this patch doesn't apply as-is.

Also, a couple minor things:

- select_cpu_vtime.bpf.c: There's an extra trailing blank line before
  the closing brace in select_cpu_vtime_stopping().

- maximal.bpf.c: The sched_switch tracepoint's prev_state parameter is
  typed as unsigned long but the TP_PROTO declares unsigned int.

Can you rebase and fix these up?

Thanks.
--
tejun

[PATCH v5 0/2] sched_ext: Update demo schedulers and selftests for deprecated APIs

[Cheng-Yang Chou]

Two sets of sched_ext APIs have been deprecated:

- Direct writes to p->scx.dsq_vtime in favor of
  scx_bpf_task_set_dsq_vtime()
- ops.cpu_acquire/release() in favor of handling CPU preemption via the
  sched_switch tracepoint, as introduced by commit a3f5d4822253
  ("sched_ext: Allow scx_bpf_reenqueue_local() to be called from
  anywhere")

This series updates the demo schedulers (scx_simple, scx_flatcg) and
selftests (select_cpu_vtime, maximal) to use the new APIs, keeping them
in sync with current best practices.

Patch 1 updates scx_simple, scx_flatcg, and select_cpu_vtime to use
scx_bpf_task_set_dsq_vtime() with scale_by_task_weight_inverse(), with
the delta computation moved inside the if block.

Patch 2 replaces the cpu_acquire/release stubs in the maximal selftest
with a minimal sched_switch TP program, and fixes the maximal and
reload_loop tests to properly attach the sched_switch tracepoint via
bpf_map__set_autoattach() and maximal__attach(), as both tests use the
maximal skeleton.

Changes in v5:
- Fix trailing blank line in select_cpu_vtime_stopping() (Tejun Heo)
- Fix prev_state type: unsigned long -> unsigned int in
  maximal_sched_switch (Tejun Heo)
- Link to v4:
  https://lore.kernel.org/all/20260314033708.26483-1-yphbchou0911@gmail.com/

Changes in v4:
- Drop scx_qmap changes from patch 2, already applied in commit
  3229ac4a5ef5 ("sched_ext: Add SCX_OPS_ALWAYS_ENQ_IMMED ops flag")
- Move delta computation inside the if block (Tejun)
- Link to v3:
  https://lore.kernel.org/all/20260313014944.1234667-1-yphbchou0911@gmail.com/

Changes in v3:
- Use scale_by_task_weight_inverse() instead of scale_by_task_weight()
  (Andrea Righi)
- Link to v2:
  https://lore.kernel.org/all/20260312175527.1220540-1-yphbchou0911@gmail.com/

Changes in v2:
- Use scx_bpf_task_set_dsq_vtime() with scale_by_task_weight instead
  of direct assignment and remove redundant bpf_ksym_exists() logic
  (Andrea Righi)
- Mention commit a3f5d4822253 ("sched_ext: Allow
  scx_bpf_reenqueue_local() to be called from anywhere") in the commit
  message to clarify why ops.cpu_acquire/release() are being deprecated
  (Andrea Righi)
- Fix maximal.c and reload_loop.c to actually attach the sched_switch
  tracepoint by calling bpf_map__set_autoattach() and maximal__attach()
  (Andrea Righi)
- Link to v1:
  https://lore.kernel.org/all/20260312042001.955675-1-yphbchou0911@gmail.com/

Thanks,
Cheng-Yang

---

Cheng-Yang Chou (2):
  sched_ext: Update demo schedulers and selftests to use
    scx_bpf_task_set_dsq_vtime()
  sched_ext: Update selftests to drop ops.cpu_acquire/release()

 tools/sched_ext/scx_flatcg.bpf.c                  | 12 +++++++-----
 tools/sched_ext/scx_simple.bpf.c                  |  6 ++++--
 tools/testing/selftests/sched_ext/maximal.bpf.c   | 15 ++++++---------
 tools/testing/selftests/sched_ext/maximal.c       |  3 +++
 tools/testing/selftests/sched_ext/reload_loop.c   |  3 +++
 .../selftests/sched_ext/select_cpu_vtime.bpf.c    |  6 ++++--
 6 files changed, 27 insertions(+), 18 deletions(-)

-- 
2.48.1

[PATCH v4 0/2] sched_ext: Update demo schedulers and selftests for deprecated APIs

[Cheng-Yang Chou]

Two sets of sched_ext APIs have been deprecated:

- Direct writes to p->scx.dsq_vtime in favor of 
  scx_bpf_task_set_dsq_vtime()
- ops.cpu_acquire/release() in favor of handling CPU preemption via the
  sched_switch tracepoint, as introduced by commit a3f5d4822253 
  ("sched_ext: Allow scx_bpf_reenqueue_local() to be called from 
  anywhere")

This series updates the demo schedulers (scx_simple, scx_flatcg) and
selftests (select_cpu_vtime, maximal) to use the new APIs, keeping them
in sync with current best practices.

Patch 1 updates scx_simple, scx_flatcg, and select_cpu_vtime to use
scx_bpf_task_set_dsq_vtime() with scale_by_task_weight_inverse(), with
the delta computation moved inside the if block.

Patch 2 replaces the cpu_acquire/release stubs in the maximal selftest
with a minimal sched_switch TP program, and fixes the maximal and
reload_loop tests to properly attach the sched_switch tracepoint via
bpf_map__set_autoattach() and maximal__attach(), as both tests use the
maximal skeleton.

Changes in v4:
- Drop scx_qmap changes from patch 2, already applied in commit 
  3229ac4a5ef5 ("sched_ext: Add SCX_OPS_ALWAYS_ENQ_IMMED ops flag")
- Move delta computation inside the if block (Tejun Heo)
- Link to v3:
  https://lore.kernel.org/all/20260313014944.1234667-1-yphbchou0911@gmail.com/

Changes in v3:
- Use scale_by_task_weight_inverse() instead of scale_by_task_weight()
  (Andrea Righi)
- Link to v2:
  https://lore.kernel.org/all/20260312175527.1220540-1-yphbchou0911@gmail.com/

Changes in v2:
- Use scx_bpf_task_set_dsq_vtime() with scale_by_task_weight instead
  of direct assignment and remove redundant bpf_ksym_exists() logic
  (Andrea Righi)
- Mention commit a3f5d4822253 ("sched_ext: Allow
  scx_bpf_reenqueue_local() to be called from anywhere") in the commit
  message to clarify why ops.cpu_acquire/release() are being deprecated
  (Andrea Righi)
- Fix maximal.c and reload_loop.c to actually attach the sched_switch
  tracepoint by calling bpf_map__set_autoattach() and maximal__attach()
  (Andrea Righi)
- Link to v1:
  https://lore.kernel.org/all/20260312042001.955675-1-yphbchou0911@gmail.com/

Thanks,
Cheng-Yang

---

Cheng-Yang Chou (2):
  sched_ext: Update demo schedulers and selftests to use
    scx_bpf_task_set_dsq_vtime()
  sched_ext: Update selftests to drop ops.cpu_acquire/release()

 tools/sched_ext/scx_flatcg.bpf.c                  | 12 +++++++-----
 tools/sched_ext/scx_simple.bpf.c                  |  6 ++++--
 tools/testing/selftests/sched_ext/maximal.bpf.c   | 15 ++++++---------
 tools/testing/selftests/sched_ext/maximal.c       |  3 +++
 tools/testing/selftests/sched_ext/reload_loop.c   |  3 +++
 .../selftests/sched_ext/select_cpu_vtime.bpf.c    |  7 +++++--
 6 files changed, 28 insertions(+), 18 deletions(-)

-- 
2.48.1

Re: [PATCH v4 0/2] sched_ext: Update demo schedulers and selftests for deprecated APIs

[Cheng-Yang Chou]

Please disregard this outdated cover letter, as I accidentally sent it.
My apologies for the mess!

-- 
Thanks,
Cheng-Yang