* [PATCH 01/46] sched_ext: Use rhashtable_lookup() instead of rhashtable_lookup_fast()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 02/46] sched_ext: Improve SCX_KF_DISPATCH comment Tejun Heo
` (44 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The find_user_dsq() function is called from contexts that are already
under RCU read lock protection. Switch from rhashtable_lookup_fast() to
rhashtable_lookup() to avoid redundant RCU locking.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index f5873f8ed669..df433f6fab4b 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -207,7 +207,7 @@ static struct scx_dispatch_q *find_global_dsq(struct task_struct *p)
static struct scx_dispatch_q *find_user_dsq(struct scx_sched *sch, u64 dsq_id)
{
- return rhashtable_lookup_fast(&sch->dsq_hash, &dsq_id, dsq_hash_params);
+ return rhashtable_lookup(&sch->dsq_hash, &dsq_id, dsq_hash_params);
}
/*
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 02/46] sched_ext: Improve SCX_KF_DISPATCH comment
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
2025-09-20 0:58 ` [PATCH 01/46] sched_ext: Use rhashtable_lookup() instead of rhashtable_lookup_fast() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 03/46] sched_ext: Fix stray scx_root usage in task_can_run_on_remote_rq() Tejun Heo
` (43 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The comment for SCX_KF_DISPATCH was incomplete and didn't explain that
ops.dispatch() may temporarily release the rq lock, allowing ENQUEUE and
SELECT_CPU operations to be nested inside DISPATCH contexts.
Update the comment to clarify this nesting behavior and provide better
context for when these operations can occur within dispatch.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/sched/ext.h | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 7047101dbf58..d82b7a9b0658 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -108,7 +108,11 @@ enum scx_kf_mask {
SCX_KF_UNLOCKED = 0, /* sleepable and not rq locked */
/* ENQUEUE and DISPATCH may be nested inside CPU_RELEASE */
SCX_KF_CPU_RELEASE = 1 << 0, /* ops.cpu_release() */
- /* ops.dequeue (in REST) may be nested inside DISPATCH */
+ /*
+ * ops.dispatch() may release rq lock temporarily and thus ENQUEUE and
+ * SELECT_CPU may be nested inside. ops.dequeue (in REST) may also be
+ * nested inside DISPATCH.
+ */
SCX_KF_DISPATCH = 1 << 1, /* ops.dispatch() */
SCX_KF_ENQUEUE = 1 << 2, /* ops.enqueue() and ops.select_cpu() */
SCX_KF_SELECT_CPU = 1 << 3, /* ops.select_cpu() */
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 03/46] sched_ext: Fix stray scx_root usage in task_can_run_on_remote_rq()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
2025-09-20 0:58 ` [PATCH 01/46] sched_ext: Use rhashtable_lookup() instead of rhashtable_lookup_fast() Tejun Heo
2025-09-20 0:58 ` [PATCH 02/46] sched_ext: Improve SCX_KF_DISPATCH comment Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 04/46] sched_ext: Use bitfields for boolean warning flags Tejun Heo
` (42 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
task_can_run_on_remote_rq() takes @sch but it is using scx_root when
incrementing SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE, which is inconsistent and
gets in the way of implementing multiple scheduler support. Use @sch
instead. As currently scx_root is the only possible scheduler instance, this
doesn't cause any behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index df433f6fab4b..5801ac676d59 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -1622,8 +1622,7 @@ static bool task_can_run_on_remote_rq(struct scx_sched *sch,
if (!scx_rq_online(rq)) {
if (enforce)
- __scx_add_event(scx_root,
- SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE, 1);
+ __scx_add_event(sch, SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE, 1);
return false;
}
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 04/46] sched_ext: Use bitfields for boolean warning flags
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (2 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 03/46] sched_ext: Fix stray scx_root usage in task_can_run_on_remote_rq() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 05/46] sched_ext: Add SCX_EFLAG_INITIALIZED to indicate successful ops.init() Tejun Heo
` (41 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Convert warned_zero_slice and warned_deprecated_rq in scx_sched struct to
single-bit bitfields to reduce struct size.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext_internal.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 2e289931e567..1a80d01b1f0c 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -871,8 +871,8 @@ struct scx_sched {
struct scx_dispatch_q **global_dsqs;
struct scx_sched_pcpu __percpu *pcpu;
- bool warned_zero_slice;
- bool warned_deprecated_rq;
+ bool warned_zero_slice:1;
+ bool warned_deprecated_rq:1;
atomic_t exit_kind;
struct scx_exit_info *exit_info;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 05/46] sched_ext: Add SCX_EFLAG_INITIALIZED to indicate successful ops.init()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (3 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 04/46] sched_ext: Use bitfields for boolean warning flags Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 06/46] sched_ext: Make qmap dump operation non-destructive Tejun Heo
` (40 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
ops.exit() may be called even if the loading failed before ops.init()
finishes successfully. This is because ops.exit() allows rich exit info
communication. Add SCX_EFLAG_INITIALIZED flag to scx_exit_info.flags to
indicate whether ops.init() finished successfully.
This enables BPF schedulers to distinguish between exit scenarios and
handle cleanup appropriately based on initialization state.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 1 +
kernel/sched/ext_internal.h | 13 +++++++++++++
2 files changed, 14 insertions(+)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 5801ac676d59..d131e98156ac 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4554,6 +4554,7 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
scx_error(sch, "ops.init() failed (%d)", ret);
goto err_disable;
}
+ sch->exit_info->flags |= SCX_EFLAG_INITIALIZED;
}
for (i = SCX_OPI_CPU_HOTPLUG_BEGIN; i < SCX_OPI_CPU_HOTPLUG_END; i++)
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 1a80d01b1f0c..b3617abed510 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -62,6 +62,16 @@ enum scx_exit_code {
SCX_ECODE_ACT_RESTART = 1LLU << 48,
};
+enum scx_exit_flags {
+ /*
+ * ops.exit() may be called even if the loading failed before ops.init()
+ * finishes successfully. This is because ops.exit() allows rich exit
+ * info communication. The following flag indicates whether ops.init()
+ * finished successfully.
+ */
+ SCX_EFLAG_INITIALIZED,
+};
+
/*
* scx_exit_info is passed to ops.exit() to describe why the BPF scheduler is
* being disabled.
@@ -73,6 +83,9 @@ struct scx_exit_info {
/* exit code if gracefully exiting */
s64 exit_code;
+ /* %SCX_EFLAG_* */
+ u64 flags;
+
/* textual representation of the above */
const char *reason;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 06/46] sched_ext: Make qmap dump operation non-destructive
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (4 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 05/46] sched_ext: Add SCX_EFLAG_INITIALIZED to indicate successful ops.init() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 07/46] tools/sched_ext: scx_qmap: Make debug output quieter by default Tejun Heo
` (39 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The qmap dump operation was destructively consuming queue entries while
displaying them. As dump can be triggered anytime, this can easily lead to
stalls. Add a temporary dump_store queue and modify the dump logic to pop
entries, display them, and then restore them back to the original queue.
This allows dump operations to be performed without affecting the
scheduler's queue state.
Note that if racing against new enqueues during dump, ordering can get
mixed up, but this is acceptable for debugging purposes.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
tools/sched_ext/scx_qmap.bpf.c | 18 +++++++++++++++++-
1 file changed, 17 insertions(+), 1 deletion(-)
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 69d877501cb7..cd50a94326e3 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -56,7 +56,8 @@ struct qmap {
queue1 SEC(".maps"),
queue2 SEC(".maps"),
queue3 SEC(".maps"),
- queue4 SEC(".maps");
+ queue4 SEC(".maps"),
+ dump_store SEC(".maps");
struct {
__uint(type, BPF_MAP_TYPE_ARRAY_OF_MAPS);
@@ -578,11 +579,26 @@ void BPF_STRUCT_OPS(qmap_dump, struct scx_dump_ctx *dctx)
return;
scx_bpf_dump("QMAP FIFO[%d]:", i);
+
+ /*
+ * Dump can be invoked anytime and there is no way to iterate in
+ * a non-destructive way. Pop and store in dump_store and then
+ * restore afterwards. If racing against new enqueues, ordering
+ * can get mixed up.
+ */
bpf_repeat(4096) {
if (bpf_map_pop_elem(fifo, &pid))
break;
+ bpf_map_push_elem(&dump_store, &pid, 0);
scx_bpf_dump(" %d", pid);
}
+
+ bpf_repeat(4096) {
+ if (bpf_map_pop_elem(&dump_store, &pid))
+ break;
+ bpf_map_push_elem(fifo, &pid, 0);
+ }
+
scx_bpf_dump("\n");
}
}
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 07/46] tools/sched_ext: scx_qmap: Make debug output quieter by default
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (5 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 06/46] sched_ext: Make qmap dump operation non-destructive Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 08/46] sched_ext: Separate out scx_kick_cpu() and add @sch to it Tejun Heo
` (38 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
scx_qmap currently outputs verbose debug messages including cgroup operations
and CPU online/offline events by default, which can be noisy during normal
operation. While the existing -P option controls DSQ dumps and event
statistics, there's no way to suppress the other debug messages.
Split the debug output controls to make scx_qmap quieter by default. The -P
option continues to control DSQ dumps and event statistics
(print_dsqs_and_events), while a new -M option controls debug messages like
cgroup operations and CPU events (print_msgs). This allows users to run
scx_qmap with minimal output and selectively enable debug information as
needed.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
tools/sched_ext/scx_qmap.bpf.c | 80 +++++++++++++++++++---------------
tools/sched_ext/scx_qmap.c | 12 +++--
2 files changed, 53 insertions(+), 39 deletions(-)
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index cd50a94326e3..3072b593f898 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -39,7 +39,8 @@ const volatile u32 stall_kernel_nth;
const volatile u32 dsp_inf_loop_after;
const volatile u32 dsp_batch;
const volatile bool highpri_boosting;
-const volatile bool print_shared_dsq;
+const volatile bool print_dsqs_and_events;
+const volatile bool print_msgs;
const volatile s32 disallow_tgid;
const volatile bool suppress_dump;
@@ -633,22 +634,25 @@ void BPF_STRUCT_OPS(qmap_dump_task, struct scx_dump_ctx *dctx, struct task_struc
s32 BPF_STRUCT_OPS(qmap_cgroup_init, struct cgroup *cgrp, struct scx_cgroup_init_args *args)
{
- bpf_printk("CGRP INIT %llu weight=%u period=%lu quota=%ld burst=%lu",
- cgrp->kn->id, args->weight, args->bw_period_us,
- args->bw_quota_us, args->bw_burst_us);
+ if (print_msgs)
+ bpf_printk("CGRP INIT %llu weight=%u period=%lu quota=%ld burst=%lu",
+ cgrp->kn->id, args->weight, args->bw_period_us,
+ args->bw_quota_us, args->bw_burst_us);
return 0;
}
void BPF_STRUCT_OPS(qmap_cgroup_set_weight, struct cgroup *cgrp, u32 weight)
{
- bpf_printk("CGRP SET %llu weight=%u", cgrp->kn->id, weight);
+ if (print_msgs)
+ bpf_printk("CGRP SET %llu weight=%u", cgrp->kn->id, weight);
}
void BPF_STRUCT_OPS(qmap_cgroup_set_bandwidth, struct cgroup *cgrp,
u64 period_us, u64 quota_us, u64 burst_us)
{
- bpf_printk("CGRP SET %llu period=%lu quota=%ld burst=%lu", cgrp->kn->id,
- period_us, quota_us, burst_us);
+ if (print_msgs)
+ bpf_printk("CGRP SET %llu period=%lu quota=%ld burst=%lu",
+ cgrp->kn->id, period_us, quota_us, burst_us);
}
/*
@@ -692,16 +696,20 @@ static void print_cpus(void)
void BPF_STRUCT_OPS(qmap_cpu_online, s32 cpu)
{
- bpf_printk("CPU %d coming online", cpu);
- /* @cpu is already online at this point */
- print_cpus();
+ if (print_msgs) {
+ bpf_printk("CPU %d coming online", cpu);
+ /* @cpu is already online at this point */
+ print_cpus();
+ }
}
void BPF_STRUCT_OPS(qmap_cpu_offline, s32 cpu)
{
- bpf_printk("CPU %d going offline", cpu);
- /* @cpu is still online at this point */
- print_cpus();
+ if (print_msgs) {
+ bpf_printk("CPU %d going offline", cpu);
+ /* @cpu is still online at this point */
+ print_cpus();
+ }
}
struct monitor_timer {
@@ -799,35 +807,36 @@ static void dump_shared_dsq(void)
static int monitor_timerfn(void *map, int *key, struct bpf_timer *timer)
{
- struct scx_event_stats events;
-
bpf_rcu_read_lock();
dispatch_highpri(true);
bpf_rcu_read_unlock();
monitor_cpuperf();
- if (print_shared_dsq)
+ if (print_dsqs_and_events) {
+ struct scx_event_stats events;
+
dump_shared_dsq();
- __COMPAT_scx_bpf_events(&events, sizeof(events));
-
- bpf_printk("%35s: %lld", "SCX_EV_SELECT_CPU_FALLBACK",
- scx_read_event(&events, SCX_EV_SELECT_CPU_FALLBACK));
- bpf_printk("%35s: %lld", "SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE",
- scx_read_event(&events, SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE));
- bpf_printk("%35s: %lld", "SCX_EV_DISPATCH_KEEP_LAST",
- scx_read_event(&events, SCX_EV_DISPATCH_KEEP_LAST));
- bpf_printk("%35s: %lld", "SCX_EV_ENQ_SKIP_EXITING",
- scx_read_event(&events, SCX_EV_ENQ_SKIP_EXITING));
- bpf_printk("%35s: %lld", "SCX_EV_REFILL_SLICE_DFL",
- scx_read_event(&events, SCX_EV_REFILL_SLICE_DFL));
- bpf_printk("%35s: %lld", "SCX_EV_BYPASS_DURATION",
- scx_read_event(&events, SCX_EV_BYPASS_DURATION));
- bpf_printk("%35s: %lld", "SCX_EV_BYPASS_DISPATCH",
- scx_read_event(&events, SCX_EV_BYPASS_DISPATCH));
- bpf_printk("%35s: %lld", "SCX_EV_BYPASS_ACTIVATE",
- scx_read_event(&events, SCX_EV_BYPASS_ACTIVATE));
+ __COMPAT_scx_bpf_events(&events, sizeof(events));
+
+ bpf_printk("%35s: %lld", "SCX_EV_SELECT_CPU_FALLBACK",
+ scx_read_event(&events, SCX_EV_SELECT_CPU_FALLBACK));
+ bpf_printk("%35s: %lld", "SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE",
+ scx_read_event(&events, SCX_EV_DISPATCH_LOCAL_DSQ_OFFLINE));
+ bpf_printk("%35s: %lld", "SCX_EV_DISPATCH_KEEP_LAST",
+ scx_read_event(&events, SCX_EV_DISPATCH_KEEP_LAST));
+ bpf_printk("%35s: %lld", "SCX_EV_ENQ_SKIP_EXITING",
+ scx_read_event(&events, SCX_EV_ENQ_SKIP_EXITING));
+ bpf_printk("%35s: %lld", "SCX_EV_REFILL_SLICE_DFL",
+ scx_read_event(&events, SCX_EV_REFILL_SLICE_DFL));
+ bpf_printk("%35s: %lld", "SCX_EV_BYPASS_DURATION",
+ scx_read_event(&events, SCX_EV_BYPASS_DURATION));
+ bpf_printk("%35s: %lld", "SCX_EV_BYPASS_DISPATCH",
+ scx_read_event(&events, SCX_EV_BYPASS_DISPATCH));
+ bpf_printk("%35s: %lld", "SCX_EV_BYPASS_ACTIVATE",
+ scx_read_event(&events, SCX_EV_BYPASS_ACTIVATE));
+ }
bpf_timer_start(timer, ONE_SEC_IN_NS, 0);
return 0;
@@ -839,7 +848,8 @@ s32 BPF_STRUCT_OPS_SLEEPABLE(qmap_init)
struct bpf_timer *timer;
s32 ret;
- print_cpus();
+ if (print_msgs)
+ print_cpus();
ret = scx_bpf_create_dsq(SHARED_DSQ, -1);
if (ret)
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index c4912ab2e76f..ef701d45ba43 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -20,7 +20,7 @@ const char help_fmt[] =
"See the top-level comment in .bpf.c for more details.\n"
"\n"
"Usage: %s [-s SLICE_US] [-e COUNT] [-t COUNT] [-T COUNT] [-l COUNT] [-b COUNT]\n"
-" [-P] [-d PID] [-D LEN] [-p] [-v]\n"
+" [-P] [-M] [-d PID] [-D LEN] [-p] [-v]\n"
"\n"
" -s SLICE_US Override slice duration\n"
" -e COUNT Trigger scx_bpf_error() after COUNT enqueues\n"
@@ -28,7 +28,8 @@ const char help_fmt[] =
" -T COUNT Stall every COUNT'th kernel thread\n"
" -l COUNT Trigger dispatch infinite looping after COUNT dispatches\n"
" -b COUNT Dispatch upto COUNT tasks together\n"
-" -P Print out DSQ content to trace_pipe every second, use with -b\n"
+" -P Print out DSQ content and event counters to trace_pipe every second\n"
+" -M Print out debug messages to trace_pipe\n"
" -H Boost nice -20 tasks in SHARED_DSQ, use with -b\n"
" -d PID Disallow a process from switching into SCHED_EXT (-1 for self)\n"
" -D LEN Set scx_exit_info.dump buffer length\n"
@@ -66,7 +67,7 @@ int main(int argc, char **argv)
skel->rodata->slice_ns = __COMPAT_ENUM_OR_ZERO("scx_public_consts", "SCX_SLICE_DFL");
- while ((opt = getopt(argc, argv, "s:e:t:T:l:b:PHd:D:Spvh")) != -1) {
+ while ((opt = getopt(argc, argv, "s:e:t:T:l:b:PMHd:D:Spvh")) != -1) {
switch (opt) {
case 's':
skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
@@ -87,7 +88,10 @@ int main(int argc, char **argv)
skel->rodata->dsp_batch = strtoul(optarg, NULL, 0);
break;
case 'P':
- skel->rodata->print_shared_dsq = true;
+ skel->rodata->print_dsqs_and_events = true;
+ break;
+ case 'M':
+ skel->rodata->print_msgs = true;
break;
case 'H':
skel->rodata->highpri_boosting = true;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 08/46] sched_ext: Separate out scx_kick_cpu() and add @sch to it
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (6 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 07/46] tools/sched_ext: scx_qmap: Make debug output quieter by default Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 09/46] sched_ext: Add the @sch parameter to __bstr_format() Tejun Heo
` (37 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation for multiple scheduler support, separate out scx_kick_cpu()
from scx_bpf_kick_cpu() and add the @sch parameter to it. scx_bpf_kick_cpu()
now acquires an RCU read lock, reads $scx_root, and calls scx_kick_cpu()
with it if non-NULL. The passed in @sch parameter is not used yet.
Internal uses of scx_bpf_kick_cpu() are converted to scx_kick_cpu(). Where
$sch is available, it's used. In the pick_task_scx() path where no
associated scheduler can be identified, $scx_root is used directly. Note
that $scx_root cannot be NULL in this case.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 43 +++++++++++++++++++++++++++----------------
1 file changed, 27 insertions(+), 16 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index d131e98156ac..560ac5a575bd 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -136,7 +136,7 @@ static struct kset *scx_kset;
#include <trace/events/sched_ext.h>
static void process_ddsp_deferred_locals(struct rq *rq);
-static void scx_bpf_kick_cpu(s32 cpu, u64 flags);
+static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags);
static void scx_vexit(struct scx_sched *sch, enum scx_exit_kind kind,
s64 exit_code, const char *fmt, va_list args);
@@ -2125,10 +2125,10 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
* balance(), we want to complete this scheduling cycle and then
* start a new one. IOW, we want to call resched_curr() on the
* next, most likely idle, task, not the current one. Use
- * scx_bpf_kick_cpu() for deferred kicking.
+ * scx_kick_cpu() for deferred kicking.
*/
if (unlikely(!--nr_loops)) {
- scx_bpf_kick_cpu(cpu_of(rq), 0);
+ scx_kick_cpu(sch, cpu_of(rq), 0);
break;
}
} while (dspc->nr_tasks);
@@ -2417,7 +2417,8 @@ static struct task_struct *pick_task_scx(struct rq *rq)
p = first_local_task(rq);
if (!p) {
if (kick_idle)
- scx_bpf_kick_cpu(cpu_of(rq), SCX_KICK_IDLE);
+ scx_kick_cpu(rcu_dereference_sched(scx_root),
+ cpu_of(rq), SCX_KICK_IDLE);
return NULL;
}
@@ -3721,7 +3722,7 @@ static void scx_clear_softlockup(void)
*
* - pick_next_task() suppresses zero slice warning.
*
- * - scx_bpf_kick_cpu() is disabled to avoid irq_work malfunction during PM
+ * - scx_kick_cpu() is disabled to avoid irq_work malfunction during PM
* operations.
*
* - scx_prio_less() reverts to the default core_sched_at order.
@@ -5809,17 +5810,7 @@ static const struct btf_kfunc_id_set scx_kfunc_set_unlocked = {
__bpf_kfunc_start_defs();
-/**
- * scx_bpf_kick_cpu - Trigger reschedule on a CPU
- * @cpu: cpu to kick
- * @flags: %SCX_KICK_* flags
- *
- * Kick @cpu into rescheduling. This can be used to wake up an idle CPU or
- * trigger rescheduling on a busy CPU. This can be called from any online
- * scx_ops operation and the actual kicking is performed asynchronously through
- * an irq work.
- */
-__bpf_kfunc void scx_bpf_kick_cpu(s32 cpu, u64 flags)
+static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags)
{
struct rq *this_rq;
unsigned long irq_flags;
@@ -5872,6 +5863,26 @@ __bpf_kfunc void scx_bpf_kick_cpu(s32 cpu, u64 flags)
local_irq_restore(irq_flags);
}
+/**
+ * scx_bpf_kick_cpu - Trigger reschedule on a CPU
+ * @cpu: cpu to kick
+ * @flags: %SCX_KICK_* flags
+ *
+ * Kick @cpu into rescheduling. This can be used to wake up an idle CPU or
+ * trigger rescheduling on a busy CPU. This can be called from any online
+ * scx_ops operation and the actual kicking is performed asynchronously through
+ * an irq work.
+ */
+__bpf_kfunc void scx_bpf_kick_cpu(s32 cpu, u64 flags)
+{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+ sch = rcu_dereference(scx_root);
+ if (likely(sch))
+ scx_kick_cpu(sch, cpu, flags);
+}
+
/**
* scx_bpf_dsq_nr_queued - Return the number of queued tasks
* @dsq_id: id of the DSQ
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 09/46] sched_ext: Add the @sch parameter to __bstr_format()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (7 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 08/46] sched_ext: Separate out scx_kick_cpu() and add @sch to it Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 10/46] sched_ext: Add the @sch parameter to ext_idle helpers Tejun Heo
` (36 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation for multiple scheduler support, add the @sch parameter to
__bstr_format() and update the callers to read $scx_root, verify that it's
not NULL and pass it in. The passed in @sch parameter is not used yet.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 28 +++++++++++++++++++++-------
1 file changed, 21 insertions(+), 7 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 560ac5a575bd..373146154829 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -6063,8 +6063,9 @@ __bpf_kfunc void bpf_iter_scx_dsq_destroy(struct bpf_iter_scx_dsq *it)
__bpf_kfunc_end_defs();
-static s32 __bstr_format(u64 *data_buf, char *line_buf, size_t line_size,
- char *fmt, unsigned long long *data, u32 data__sz)
+static s32 __bstr_format(struct scx_sched *sch, u64 *data_buf, char *line_buf,
+ size_t line_size, char *fmt, unsigned long long *data,
+ u32 data__sz)
{
struct bpf_bprintf_data bprintf_data = { .get_bin_args = true };
s32 ret;
@@ -6099,10 +6100,10 @@ static s32 __bstr_format(u64 *data_buf, char *line_buf, size_t line_size,
return ret;
}
-static s32 bstr_format(struct scx_bstr_buf *buf,
+static s32 bstr_format(struct scx_sched *sch, struct scx_bstr_buf *buf,
char *fmt, unsigned long long *data, u32 data__sz)
{
- return __bstr_format(buf->data, buf->line, sizeof(buf->line),
+ return __bstr_format(sch, buf->data, buf->line, sizeof(buf->line),
fmt, data, data__sz);
}
@@ -6121,10 +6122,13 @@ __bpf_kfunc_start_defs();
__bpf_kfunc void scx_bpf_exit_bstr(s64 exit_code, char *fmt,
unsigned long long *data, u32 data__sz)
{
+ struct scx_sched *sch;
unsigned long flags;
raw_spin_lock_irqsave(&scx_exit_bstr_buf_lock, flags);
- if (bstr_format(&scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
+ sch = rcu_dereference_bh(scx_root);
+ if (likely(sch) &&
+ bstr_format(sch, &scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
scx_kf_exit(SCX_EXIT_UNREG_BPF, exit_code, "%s", scx_exit_bstr_buf.line);
raw_spin_unlock_irqrestore(&scx_exit_bstr_buf_lock, flags);
}
@@ -6141,10 +6145,13 @@ __bpf_kfunc void scx_bpf_exit_bstr(s64 exit_code, char *fmt,
__bpf_kfunc void scx_bpf_error_bstr(char *fmt, unsigned long long *data,
u32 data__sz)
{
+ struct scx_sched *sch;
unsigned long flags;
raw_spin_lock_irqsave(&scx_exit_bstr_buf_lock, flags);
- if (bstr_format(&scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
+ sch = rcu_dereference_bh(scx_root);
+ if (likely(sch) &&
+ bstr_format(sch, &scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
scx_kf_exit(SCX_EXIT_ERROR_BPF, 0, "%s", scx_exit_bstr_buf.line);
raw_spin_unlock_irqrestore(&scx_exit_bstr_buf_lock, flags);
}
@@ -6164,17 +6171,24 @@ __bpf_kfunc void scx_bpf_error_bstr(char *fmt, unsigned long long *data,
__bpf_kfunc void scx_bpf_dump_bstr(char *fmt, unsigned long long *data,
u32 data__sz)
{
+ struct scx_sched *sch;
struct scx_dump_data *dd = &scx_dump_data;
struct scx_bstr_buf *buf = &dd->buf;
s32 ret;
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return;
+
if (raw_smp_processor_id() != dd->cpu) {
scx_kf_error("scx_bpf_dump() must only be called from ops.dump() and friends");
return;
}
/* append the formatted string to the line buf */
- ret = __bstr_format(buf->data, buf->line + dd->cursor,
+ ret = __bstr_format(sch, buf->data, buf->line + dd->cursor,
sizeof(buf->line) - dd->cursor, fmt, data, data__sz);
if (ret < 0) {
dump_line(dd->s, "%s[!] (\"%s\", %p, %u) failed to format (%d)",
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 10/46] sched_ext: Add the @sch parameter to ext_idle helpers
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (8 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 09/46] sched_ext: Add the @sch parameter to __bstr_format() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 11/46] sched_ext: Drop kf_cpu_valid() Tejun Heo
` (35 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation for multiple scheduler support, add the @sch parameter to
validate_node(), check_builtin_idle_enabled() and select_cpu_from_kfunc(),
and update their callers to read $scx_root, verify that it's not NULL and
pass it in. The passed in @sch parameter is not used yet.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext_idle.c | 109 ++++++++++++++++++++++++++++++++++------
1 file changed, 94 insertions(+), 15 deletions(-)
diff --git a/kernel/sched/ext_idle.c b/kernel/sched/ext_idle.c
index 7174e1c1a392..6e2504ae7357 100644
--- a/kernel/sched/ext_idle.c
+++ b/kernel/sched/ext_idle.c
@@ -819,7 +819,7 @@ void scx_idle_disable(void)
* Helpers that can be called from the BPF scheduler.
*/
-static int validate_node(int node)
+static int validate_node(struct scx_sched *sch, int node)
{
if (!static_branch_likely(&scx_builtin_idle_per_node)) {
scx_kf_error("per-node idle tracking is disabled");
@@ -847,7 +847,7 @@ static int validate_node(int node)
__bpf_kfunc_start_defs();
-static bool check_builtin_idle_enabled(void)
+static bool check_builtin_idle_enabled(struct scx_sched *sch)
{
if (static_branch_likely(&scx_builtin_idle_enabled))
return true;
@@ -856,7 +856,8 @@ static bool check_builtin_idle_enabled(void)
return false;
}
-static s32 select_cpu_from_kfunc(struct task_struct *p, s32 prev_cpu, u64 wake_flags,
+static s32 select_cpu_from_kfunc(struct scx_sched *sch, struct task_struct *p,
+ s32 prev_cpu, u64 wake_flags,
const struct cpumask *allowed, u64 flags)
{
struct rq *rq;
@@ -866,7 +867,7 @@ static s32 select_cpu_from_kfunc(struct task_struct *p, s32 prev_cpu, u64 wake_f
if (!kf_cpu_valid(prev_cpu, NULL))
return -EINVAL;
- if (!check_builtin_idle_enabled())
+ if (!check_builtin_idle_enabled(sch))
return -EBUSY;
/*
@@ -946,15 +947,21 @@ __bpf_kfunc int scx_bpf_cpu_node(s32 cpu)
__bpf_kfunc s32 scx_bpf_select_cpu_dfl(struct task_struct *p, s32 prev_cpu,
u64 wake_flags, bool *is_idle)
{
+ struct scx_sched *sch;
s32 cpu;
- cpu = select_cpu_from_kfunc(p, prev_cpu, wake_flags, NULL, 0);
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return -ENODEV;
+
+ cpu = select_cpu_from_kfunc(sch, p, prev_cpu, wake_flags, NULL, 0);
if (cpu >= 0) {
*is_idle = true;
return cpu;
}
*is_idle = false;
-
return prev_cpu;
}
@@ -981,7 +988,16 @@ __bpf_kfunc s32 scx_bpf_select_cpu_dfl(struct task_struct *p, s32 prev_cpu,
__bpf_kfunc s32 scx_bpf_select_cpu_and(struct task_struct *p, s32 prev_cpu, u64 wake_flags,
const struct cpumask *cpus_allowed, u64 flags)
{
- return select_cpu_from_kfunc(p, prev_cpu, wake_flags, cpus_allowed, flags);
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return -ENODEV;
+
+ return select_cpu_from_kfunc(sch, p, prev_cpu, wake_flags,
+ cpus_allowed, flags);
}
/**
@@ -995,7 +1011,15 @@ __bpf_kfunc s32 scx_bpf_select_cpu_and(struct task_struct *p, s32 prev_cpu, u64
*/
__bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask_node(int node)
{
- node = validate_node(node);
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return cpu_none_mask;
+
+ node = validate_node(sch, node);
if (node < 0)
return cpu_none_mask;
@@ -1011,12 +1035,20 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask_node(int node)
*/
__bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask(void)
{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return cpu_none_mask;
+
if (static_branch_unlikely(&scx_builtin_idle_per_node)) {
scx_kf_error("SCX_OPS_BUILTIN_IDLE_PER_NODE enabled");
return cpu_none_mask;
}
- if (!check_builtin_idle_enabled())
+ if (!check_builtin_idle_enabled(sch))
return cpu_none_mask;
return idle_cpumask(NUMA_NO_NODE)->cpu;
@@ -1034,7 +1066,15 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask(void)
*/
__bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask_node(int node)
{
- node = validate_node(node);
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return cpu_none_mask;
+
+ node = validate_node(sch, node);
if (node < 0)
return cpu_none_mask;
@@ -1054,12 +1094,20 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask_node(int node)
*/
__bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask(void)
{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return cpu_none_mask;
+
if (static_branch_unlikely(&scx_builtin_idle_per_node)) {
scx_kf_error("SCX_OPS_BUILTIN_IDLE_PER_NODE enabled");
return cpu_none_mask;
}
- if (!check_builtin_idle_enabled())
+ if (!check_builtin_idle_enabled(sch))
return cpu_none_mask;
if (sched_smt_active())
@@ -1095,7 +1143,15 @@ __bpf_kfunc void scx_bpf_put_idle_cpumask(const struct cpumask *idle_mask)
*/
__bpf_kfunc bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
{
- if (!check_builtin_idle_enabled())
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return false;
+
+ if (!check_builtin_idle_enabled(sch))
return false;
if (!kf_cpu_valid(cpu, NULL))
@@ -1126,7 +1182,15 @@ __bpf_kfunc bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
__bpf_kfunc s32 scx_bpf_pick_idle_cpu_node(const struct cpumask *cpus_allowed,
int node, u64 flags)
{
- node = validate_node(node);
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return -ENODEV;
+
+ node = validate_node(sch, node);
if (node < 0)
return node;
@@ -1158,12 +1222,20 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu_node(const struct cpumask *cpus_allowed,
__bpf_kfunc s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed,
u64 flags)
{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return -ENODEV;
+
if (static_branch_maybe(CONFIG_NUMA, &scx_builtin_idle_per_node)) {
scx_kf_error("per-node idle tracking is enabled");
return -EBUSY;
}
- if (!check_builtin_idle_enabled())
+ if (!check_builtin_idle_enabled(sch))
return -EBUSY;
return scx_pick_idle_cpu(cpus_allowed, NUMA_NO_NODE, flags);
@@ -1193,9 +1265,16 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed,
__bpf_kfunc s32 scx_bpf_pick_any_cpu_node(const struct cpumask *cpus_allowed,
int node, u64 flags)
{
+ struct scx_sched *sch;
s32 cpu;
- node = validate_node(node);
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return -ENODEV;
+
+ node = validate_node(sch, node);
if (node < 0)
return node;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 11/46] sched_ext: Drop kf_cpu_valid()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (9 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 10/46] sched_ext: Add the @sch parameter to ext_idle helpers Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 12/46] sched_ext: Add the @sch parameter to scx_dsq_insert_preamble/commit() Tejun Heo
` (34 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The intention behind kf_cpu_valid() was that when called from kfuncs,
kf_cpu_valid() would be able to implicitly determine the scx_sched instance
being operated on and thus wouldn't need @sch passed in explicitly. This
turned out to be unnecessarily complicated to implement and not have
justifiable practical benefits. Replace kf_cpu_valid() usages with
ops_cpu_valid() which takes explicit @sch.
Callers which don't have $sch available in the context are updated to read
$scx_root under RCU read lock, verify that it's not NULL and pass it in.
scx_bpf_cpu_rq() is restructured to use guard(rcu)() instead of explicit
rcu_read_[un]lock().
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 67 ++++++++++++++++++++++++-----------------
kernel/sched/ext_idle.c | 12 +++++---
2 files changed, 48 insertions(+), 31 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 373146154829..56ca09f46d1e 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -736,23 +736,6 @@ static bool ops_cpu_valid(struct scx_sched *sch, s32 cpu, const char *where)
}
}
-/**
- * kf_cpu_valid - Verify a CPU number, to be used on kfunc input args
- * @cpu: cpu number which came from a BPF ops
- * @where: extra information reported on error
- *
- * The same as ops_cpu_valid() but @sch is implicit.
- */
-static bool kf_cpu_valid(u32 cpu, const char *where)
-{
- if (__cpu_valid(cpu)) {
- return true;
- } else {
- scx_kf_error("invalid CPU %d%s%s", cpu, where ? " " : "", where ?: "");
- return false;
- }
-}
-
/**
* ops_sanitize_err - Sanitize a -errno value
* @sch: scx_sched to error out on error
@@ -5815,7 +5798,7 @@ static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags)
struct rq *this_rq;
unsigned long irq_flags;
- if (!kf_cpu_valid(cpu, NULL))
+ if (!ops_cpu_valid(sch, cpu, NULL))
return;
local_irq_save(irq_flags);
@@ -6224,7 +6207,12 @@ __bpf_kfunc void scx_bpf_dump_bstr(char *fmt, unsigned long long *data,
*/
__bpf_kfunc u32 scx_bpf_cpuperf_cap(s32 cpu)
{
- if (kf_cpu_valid(cpu, NULL))
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (likely(sch) && ops_cpu_valid(sch, cpu, NULL))
return arch_scale_cpu_capacity(cpu);
else
return SCX_CPUPERF_ONE;
@@ -6246,7 +6234,12 @@ __bpf_kfunc u32 scx_bpf_cpuperf_cap(s32 cpu)
*/
__bpf_kfunc u32 scx_bpf_cpuperf_cur(s32 cpu)
{
- if (kf_cpu_valid(cpu, NULL))
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (likely(sch) && ops_cpu_valid(sch, cpu, NULL))
return arch_scale_freq_capacity(cpu);
else
return SCX_CPUPERF_ONE;
@@ -6268,12 +6261,20 @@ __bpf_kfunc u32 scx_bpf_cpuperf_cur(s32 cpu)
*/
__bpf_kfunc void scx_bpf_cpuperf_set(s32 cpu, u32 perf)
{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(sch);
+ if (unlikely(!sch))
+ return;
+
if (unlikely(perf > SCX_CPUPERF_ONE)) {
scx_kf_error("Invalid cpuperf target %u for CPU %d", perf, cpu);
return;
}
- if (kf_cpu_valid(cpu, NULL)) {
+ if (ops_cpu_valid(sch, cpu, NULL)) {
struct rq *rq = cpu_rq(cpu), *locked_rq = scx_locked_rq();
struct rq_flags rf;
@@ -6379,18 +6380,21 @@ __bpf_kfunc struct rq *scx_bpf_cpu_rq(s32 cpu)
{
struct scx_sched *sch;
- if (!kf_cpu_valid(cpu, NULL))
- return NULL;
+ guard(rcu)();
- rcu_read_lock();
sch = rcu_dereference(scx_root);
- if (likely(sch) && !sch->warned_deprecated_rq) {
+ if (unlikely(!sch))
+ return NULL;
+
+ if (!ops_cpu_valid(sch, cpu, NULL))
+ return NULL;
+
+ if (!sch->warned_deprecated_rq) {
printk_deferred(KERN_WARNING "sched_ext: %s() is deprecated; "
"use scx_bpf_locked_rq() when holding rq lock "
"or scx_bpf_cpu_curr() to read remote curr safely.\n", __func__);
sch->warned_deprecated_rq = true;
}
- rcu_read_unlock();
return cpu_rq(cpu);
}
@@ -6425,8 +6429,17 @@ __bpf_kfunc struct rq *scx_bpf_locked_rq(void)
*/
__bpf_kfunc struct task_struct *scx_bpf_cpu_curr(s32 cpu)
{
- if (!kf_cpu_valid(cpu, NULL))
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
return NULL;
+
+ if (!ops_cpu_valid(sch, cpu, NULL))
+ return NULL;
+
return rcu_dereference(cpu_rq(cpu)->curr);
}
diff --git a/kernel/sched/ext_idle.c b/kernel/sched/ext_idle.c
index 6e2504ae7357..a576ec10522e 100644
--- a/kernel/sched/ext_idle.c
+++ b/kernel/sched/ext_idle.c
@@ -864,7 +864,7 @@ static s32 select_cpu_from_kfunc(struct scx_sched *sch, struct task_struct *p,
struct rq_flags rf;
s32 cpu;
- if (!kf_cpu_valid(prev_cpu, NULL))
+ if (!ops_cpu_valid(sch, prev_cpu, NULL))
return -EINVAL;
if (!check_builtin_idle_enabled(sch))
@@ -923,9 +923,13 @@ static s32 select_cpu_from_kfunc(struct scx_sched *sch, struct task_struct *p,
*/
__bpf_kfunc int scx_bpf_cpu_node(s32 cpu)
{
- if (!kf_cpu_valid(cpu, NULL))
- return NUMA_NO_NODE;
+ struct scx_sched *sch;
+
+ guard(rcu)();
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch) || !ops_cpu_valid(sch, cpu, NULL))
+ return NUMA_NO_NODE;
return cpu_to_node(cpu);
}
@@ -1154,7 +1158,7 @@ __bpf_kfunc bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
if (!check_builtin_idle_enabled(sch))
return false;
- if (!kf_cpu_valid(cpu, NULL))
+ if (!ops_cpu_valid(sch, cpu, NULL))
return false;
return scx_idle_test_and_clear_cpu(cpu);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 12/46] sched_ext: Add the @sch parameter to scx_dsq_insert_preamble/commit()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (10 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 11/46] sched_ext: Drop kf_cpu_valid() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 13/46] sched_ext: Drop scx_kf_exit() and scx_kf_error() Tejun Heo
` (33 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation for multiple scheduler support, add the @sch parameter to
scx_dsq_insert_preamble/commit() and update the callers to read $scx_root
and pass it in. The passed in @sch parameter is not used yet.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 29 ++++++++++++++++++++++-------
1 file changed, 22 insertions(+), 7 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 56ca09f46d1e..1455f8c56d5c 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -5240,7 +5240,8 @@ void __init init_sched_ext_class(void)
/********************************************************************************
* Helpers that can be called from the BPF scheduler.
*/
-static bool scx_dsq_insert_preamble(struct task_struct *p, u64 enq_flags)
+static bool scx_dsq_insert_preamble(struct scx_sched *sch, struct task_struct *p,
+ u64 enq_flags)
{
if (!scx_kf_allowed(SCX_KF_ENQUEUE | SCX_KF_DISPATCH))
return false;
@@ -5260,8 +5261,8 @@ static bool scx_dsq_insert_preamble(struct task_struct *p, u64 enq_flags)
return true;
}
-static void scx_dsq_insert_commit(struct task_struct *p, u64 dsq_id,
- u64 enq_flags)
+static void scx_dsq_insert_commit(struct scx_sched *sch, struct task_struct *p,
+ u64 dsq_id, u64 enq_flags)
{
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct task_struct *ddsp_task;
@@ -5325,7 +5326,14 @@ __bpf_kfunc_start_defs();
__bpf_kfunc void scx_bpf_dsq_insert(struct task_struct *p, u64 dsq_id, u64 slice,
u64 enq_flags)
{
- if (!scx_dsq_insert_preamble(p, enq_flags))
+ struct scx_sched *sch;
+
+ guard(rcu)();
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return;
+
+ if (!scx_dsq_insert_preamble(sch, p, enq_flags))
return;
if (slice)
@@ -5333,7 +5341,7 @@ __bpf_kfunc void scx_bpf_dsq_insert(struct task_struct *p, u64 dsq_id, u64 slice
else
p->scx.slice = p->scx.slice ?: 1;
- scx_dsq_insert_commit(p, dsq_id, enq_flags);
+ scx_dsq_insert_commit(sch, p, dsq_id, enq_flags);
}
/**
@@ -5360,7 +5368,14 @@ __bpf_kfunc void scx_bpf_dsq_insert(struct task_struct *p, u64 dsq_id, u64 slice
__bpf_kfunc void scx_bpf_dsq_insert_vtime(struct task_struct *p, u64 dsq_id,
u64 slice, u64 vtime, u64 enq_flags)
{
- if (!scx_dsq_insert_preamble(p, enq_flags))
+ struct scx_sched *sch;
+
+ guard(rcu)();
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return;
+
+ if (!scx_dsq_insert_preamble(sch, p, enq_flags))
return;
if (slice)
@@ -5370,7 +5385,7 @@ __bpf_kfunc void scx_bpf_dsq_insert_vtime(struct task_struct *p, u64 dsq_id,
p->scx.dsq_vtime = vtime;
- scx_dsq_insert_commit(p, dsq_id, enq_flags | SCX_ENQ_DSQ_PRIQ);
+ scx_dsq_insert_commit(sch, p, dsq_id, enq_flags | SCX_ENQ_DSQ_PRIQ);
}
__bpf_kfunc_end_defs();
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 13/46] sched_ext: Drop scx_kf_exit() and scx_kf_error()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (11 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 12/46] sched_ext: Add the @sch parameter to scx_dsq_insert_preamble/commit() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 14/46] sched_ext: Misc updates around scx_sched instance pointer Tejun Heo
` (32 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The intention behind scx_kf_exit/error() was that when called from kfuncs,
scx_kf_exit/error() would be able to implicitly determine the scx_sched
instance being operated on and thus wouldn't need the @sch parameter passed
in explicitly. This turned out to be unnecessarily complicated to implement
and not have enough practical benefits. Replace scx_kf_exit/error() usages
with scx_exit/error() which take an explicit @sch parameter.
- Add the @sch parameter to scx_kf_allowed(), scx_kf_allowed_on_arg_tasks,
mark_direct_dispatch() and other intermediate functions transitively.
- In callers that don't already have @sch available, grab RCU, read
$scx_root, verify it's not NULL and use it.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 126 +++++++++++++++++++++++-----------------
kernel/sched/ext_idle.c | 25 +++++---
2 files changed, 88 insertions(+), 63 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 1455f8c56d5c..0c99a55f199b 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -151,24 +151,7 @@ static __printf(4, 5) void scx_exit(struct scx_sched *sch,
va_end(args);
}
-static __printf(3, 4) void scx_kf_exit(enum scx_exit_kind kind, s64 exit_code,
- const char *fmt, ...)
-{
- struct scx_sched *sch;
- va_list args;
-
- rcu_read_lock();
- sch = rcu_dereference(scx_root);
- if (sch) {
- va_start(args, fmt);
- scx_vexit(sch, kind, exit_code, fmt, args);
- va_end(args);
- }
- rcu_read_unlock();
-}
-
#define scx_error(sch, fmt, args...) scx_exit((sch), SCX_EXIT_ERROR, 0, fmt, ##args)
-#define scx_kf_error(fmt, args...) scx_kf_exit(SCX_EXIT_ERROR, 0, fmt, ##args)
#define SCX_HAS_OP(sch, op) test_bit(SCX_OP_IDX(op), (sch)->has_op)
@@ -329,11 +312,11 @@ do { \
})
/* @mask is constant, always inline to cull unnecessary branches */
-static __always_inline bool scx_kf_allowed(u32 mask)
+static __always_inline bool scx_kf_allowed(struct scx_sched *sch, u32 mask)
{
if (unlikely(!(current->scx.kf_mask & mask))) {
- scx_kf_error("kfunc with mask 0x%x called from an operation only allowing 0x%x",
- mask, current->scx.kf_mask);
+ scx_error(sch, "kfunc with mask 0x%x called from an operation only allowing 0x%x",
+ mask, current->scx.kf_mask);
return false;
}
@@ -346,13 +329,13 @@ static __always_inline bool scx_kf_allowed(u32 mask)
*/
if (unlikely(highest_bit(mask) == SCX_KF_CPU_RELEASE &&
(current->scx.kf_mask & higher_bits(SCX_KF_CPU_RELEASE)))) {
- scx_kf_error("cpu_release kfunc called from a nested operation");
+ scx_error(sch, "cpu_release kfunc called from a nested operation");
return false;
}
if (unlikely(highest_bit(mask) == SCX_KF_DISPATCH &&
(current->scx.kf_mask & higher_bits(SCX_KF_DISPATCH)))) {
- scx_kf_error("dispatch kfunc called from a nested operation");
+ scx_error(sch, "dispatch kfunc called from a nested operation");
return false;
}
@@ -360,15 +343,16 @@ static __always_inline bool scx_kf_allowed(u32 mask)
}
/* see SCX_CALL_OP_TASK() */
-static __always_inline bool scx_kf_allowed_on_arg_tasks(u32 mask,
+static __always_inline bool scx_kf_allowed_on_arg_tasks(struct scx_sched *sch,
+ u32 mask,
struct task_struct *p)
{
- if (!scx_kf_allowed(mask))
+ if (!scx_kf_allowed(sch, mask))
return false;
if (unlikely((p != current->scx.kf_tasks[0] &&
p != current->scx.kf_tasks[1]))) {
- scx_kf_error("called on a task not being operated on");
+ scx_error(sch, "called on a task not being operated on");
return false;
}
@@ -1115,7 +1099,8 @@ static struct scx_dispatch_q *find_dsq_for_dispatch(struct scx_sched *sch,
return dsq;
}
-static void mark_direct_dispatch(struct task_struct *ddsp_task,
+static void mark_direct_dispatch(struct scx_sched *sch,
+ struct task_struct *ddsp_task,
struct task_struct *p, u64 dsq_id,
u64 enq_flags)
{
@@ -1129,10 +1114,10 @@ static void mark_direct_dispatch(struct task_struct *ddsp_task,
/* @p must match the task on the enqueue path */
if (unlikely(p != ddsp_task)) {
if (IS_ERR(ddsp_task))
- scx_kf_error("%s[%d] already direct-dispatched",
+ scx_error(sch, "%s[%d] already direct-dispatched",
p->comm, p->pid);
else
- scx_kf_error("scheduling for %s[%d] but trying to direct-dispatch %s[%d]",
+ scx_error(sch, "scheduling for %s[%d] but trying to direct-dispatch %s[%d]",
ddsp_task->comm, ddsp_task->pid,
p->comm, p->pid);
return;
@@ -5243,18 +5228,18 @@ void __init init_sched_ext_class(void)
static bool scx_dsq_insert_preamble(struct scx_sched *sch, struct task_struct *p,
u64 enq_flags)
{
- if (!scx_kf_allowed(SCX_KF_ENQUEUE | SCX_KF_DISPATCH))
+ if (!scx_kf_allowed(sch, SCX_KF_ENQUEUE | SCX_KF_DISPATCH))
return false;
lockdep_assert_irqs_disabled();
if (unlikely(!p)) {
- scx_kf_error("called with NULL task");
+ scx_error(sch, "called with NULL task");
return false;
}
if (unlikely(enq_flags & __SCX_ENQ_INTERNAL_MASK)) {
- scx_kf_error("invalid enq_flags 0x%llx", enq_flags);
+ scx_error(sch, "invalid enq_flags 0x%llx", enq_flags);
return false;
}
@@ -5269,12 +5254,12 @@ static void scx_dsq_insert_commit(struct scx_sched *sch, struct task_struct *p,
ddsp_task = __this_cpu_read(direct_dispatch_task);
if (ddsp_task) {
- mark_direct_dispatch(ddsp_task, p, dsq_id, enq_flags);
+ mark_direct_dispatch(sch, ddsp_task, p, dsq_id, enq_flags);
return;
}
if (unlikely(dspc->cursor >= scx_dsp_max_batch)) {
- scx_kf_error("dispatch buffer overflow");
+ scx_error(sch, "dispatch buffer overflow");
return;
}
@@ -5410,7 +5395,8 @@ static bool scx_dsq_move(struct bpf_iter_scx_dsq_kern *kit,
bool in_balance;
unsigned long flags;
- if (!scx_kf_allowed_if_unlocked() && !scx_kf_allowed(SCX_KF_DISPATCH))
+ if (!scx_kf_allowed_if_unlocked() &&
+ !scx_kf_allowed(sch, SCX_KF_DISPATCH))
return false;
/*
@@ -5495,7 +5481,15 @@ __bpf_kfunc_start_defs();
*/
__bpf_kfunc u32 scx_bpf_dispatch_nr_slots(void)
{
- if (!scx_kf_allowed(SCX_KF_DISPATCH))
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return 0;
+
+ if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return 0;
return scx_dsp_max_batch - __this_cpu_read(scx_dsp_ctx->cursor);
@@ -5510,14 +5504,21 @@ __bpf_kfunc u32 scx_bpf_dispatch_nr_slots(void)
__bpf_kfunc void scx_bpf_dispatch_cancel(void)
{
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
+ struct scx_sched *sch;
- if (!scx_kf_allowed(SCX_KF_DISPATCH))
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return;
+
+ if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return;
if (dspc->cursor > 0)
dspc->cursor--;
else
- scx_kf_error("dispatch buffer underflow");
+ scx_error(sch, "dispatch buffer underflow");
}
/**
@@ -5540,7 +5541,7 @@ __bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id)
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct scx_dispatch_q *dsq;
- if (!scx_kf_allowed(SCX_KF_DISPATCH))
+ if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return false;
flush_dispatch_buf(sch, dspc->rq);
@@ -5687,12 +5688,18 @@ __bpf_kfunc_start_defs();
*/
__bpf_kfunc u32 scx_bpf_reenqueue_local(void)
{
+ struct scx_sched *sch;
LIST_HEAD(tasks);
u32 nr_enqueued = 0;
struct rq *rq;
struct task_struct *p, *n;
- if (!scx_kf_allowed(SCX_KF_CPU_RELEASE))
+ guard(rcu)();
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return 0;
+
+ if (!scx_kf_allowed(sch, SCX_KF_CPU_RELEASE))
return 0;
rq = cpu_rq(smp_processor_id());
@@ -5837,7 +5844,7 @@ static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags)
struct rq *target_rq = cpu_rq(cpu);
if (unlikely(flags & (SCX_KICK_PREEMPT | SCX_KICK_WAIT)))
- scx_kf_error("PREEMPT/WAIT cannot be used with SCX_KICK_IDLE");
+ scx_error(sch, "PREEMPT/WAIT cannot be used with SCX_KICK_IDLE");
if (raw_spin_rq_trylock(target_rq)) {
if (can_skip_idle_kick(target_rq)) {
@@ -6070,20 +6077,20 @@ static s32 __bstr_format(struct scx_sched *sch, u64 *data_buf, char *line_buf,
if (data__sz % 8 || data__sz > MAX_BPRINTF_VARARGS * 8 ||
(data__sz && !data)) {
- scx_kf_error("invalid data=%p and data__sz=%u", (void *)data, data__sz);
+ scx_error(sch, "invalid data=%p and data__sz=%u", (void *)data, data__sz);
return -EINVAL;
}
ret = copy_from_kernel_nofault(data_buf, data, data__sz);
if (ret < 0) {
- scx_kf_error("failed to read data fields (%d)", ret);
+ scx_error(sch, "failed to read data fields (%d)", ret);
return ret;
}
ret = bpf_bprintf_prepare(fmt, UINT_MAX, data_buf, data__sz / 8,
&bprintf_data);
if (ret < 0) {
- scx_kf_error("format preparation failed (%d)", ret);
+ scx_error(sch, "format preparation failed (%d)", ret);
return ret;
}
@@ -6091,7 +6098,7 @@ static s32 __bstr_format(struct scx_sched *sch, u64 *data_buf, char *line_buf,
bprintf_data.bin_args);
bpf_bprintf_cleanup(&bprintf_data);
if (ret < 0) {
- scx_kf_error("(\"%s\", %p, %u) failed to format", fmt, data, data__sz);
+ scx_error(sch, "(\"%s\", %p, %u) failed to format", fmt, data, data__sz);
return ret;
}
@@ -6127,7 +6134,7 @@ __bpf_kfunc void scx_bpf_exit_bstr(s64 exit_code, char *fmt,
sch = rcu_dereference_bh(scx_root);
if (likely(sch) &&
bstr_format(sch, &scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
- scx_kf_exit(SCX_EXIT_UNREG_BPF, exit_code, "%s", scx_exit_bstr_buf.line);
+ scx_exit(sch, SCX_EXIT_UNREG_BPF, exit_code, "%s", scx_exit_bstr_buf.line);
raw_spin_unlock_irqrestore(&scx_exit_bstr_buf_lock, flags);
}
@@ -6150,7 +6157,7 @@ __bpf_kfunc void scx_bpf_error_bstr(char *fmt, unsigned long long *data,
sch = rcu_dereference_bh(scx_root);
if (likely(sch) &&
bstr_format(sch, &scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
- scx_kf_exit(SCX_EXIT_ERROR_BPF, 0, "%s", scx_exit_bstr_buf.line);
+ scx_exit(sch, SCX_EXIT_ERROR_BPF, 0, "%s", scx_exit_bstr_buf.line);
raw_spin_unlock_irqrestore(&scx_exit_bstr_buf_lock, flags);
}
@@ -6181,7 +6188,7 @@ __bpf_kfunc void scx_bpf_dump_bstr(char *fmt, unsigned long long *data,
return;
if (raw_smp_processor_id() != dd->cpu) {
- scx_kf_error("scx_bpf_dump() must only be called from ops.dump() and friends");
+ scx_error(sch, "scx_bpf_dump() must only be called from ops.dump() and friends");
return;
}
@@ -6285,7 +6292,7 @@ __bpf_kfunc void scx_bpf_cpuperf_set(s32 cpu, u32 perf)
return;
if (unlikely(perf > SCX_CPUPERF_ONE)) {
- scx_kf_error("Invalid cpuperf target %u for CPU %d", perf, cpu);
+ scx_error(sch, "Invalid cpuperf target %u for CPU %d", perf, cpu);
return;
}
@@ -6298,7 +6305,7 @@ __bpf_kfunc void scx_bpf_cpuperf_set(s32 cpu, u32 perf)
* to the corresponding CPU to prevent ABBA deadlocks.
*/
if (locked_rq && rq != locked_rq) {
- scx_kf_error("Invalid target CPU %d", cpu);
+ scx_error(sch, "Invalid target CPU %d", cpu);
return;
}
@@ -6422,16 +6429,20 @@ __bpf_kfunc struct rq *scx_bpf_cpu_rq(s32 cpu)
*/
__bpf_kfunc struct rq *scx_bpf_locked_rq(void)
{
+ struct scx_sched *sch;
struct rq *rq;
- preempt_disable();
+ guard(preempt)();
+
+ sch = rcu_dereference_sched(scx_root);
+ if (unlikely(!sch))
+ return NULL;
+
rq = scx_locked_rq();
if (!rq) {
- preempt_enable();
- scx_kf_error("accessing rq without holding rq lock");
+ scx_error(sch, "accessing rq without holding rq lock");
return NULL;
}
- preempt_enable();
return rq;
}
@@ -6474,8 +6485,15 @@ __bpf_kfunc struct cgroup *scx_bpf_task_cgroup(struct task_struct *p)
{
struct task_group *tg = p->sched_task_group;
struct cgroup *cgrp = &cgrp_dfl_root.cgrp;
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ goto out;
- if (!scx_kf_allowed_on_arg_tasks(__SCX_KF_RQ_LOCKED, p))
+ if (!scx_kf_allowed_on_arg_tasks(sch, __SCX_KF_RQ_LOCKED, p))
goto out;
cgrp = tg_cgrp(tg);
diff --git a/kernel/sched/ext_idle.c b/kernel/sched/ext_idle.c
index a576ec10522e..c57779f0ad57 100644
--- a/kernel/sched/ext_idle.c
+++ b/kernel/sched/ext_idle.c
@@ -822,7 +822,7 @@ void scx_idle_disable(void)
static int validate_node(struct scx_sched *sch, int node)
{
if (!static_branch_likely(&scx_builtin_idle_per_node)) {
- scx_kf_error("per-node idle tracking is disabled");
+ scx_error(sch, "per-node idle tracking is disabled");
return -EOPNOTSUPP;
}
@@ -832,13 +832,13 @@ static int validate_node(struct scx_sched *sch, int node)
/* Make sure node is in a valid range */
if (node < 0 || node >= nr_node_ids) {
- scx_kf_error("invalid node %d", node);
+ scx_error(sch, "invalid node %d", node);
return -EINVAL;
}
/* Make sure the node is part of the set of possible nodes */
if (!node_possible(node)) {
- scx_kf_error("unavailable node %d", node);
+ scx_error(sch, "unavailable node %d", node);
return -EINVAL;
}
@@ -852,7 +852,7 @@ static bool check_builtin_idle_enabled(struct scx_sched *sch)
if (static_branch_likely(&scx_builtin_idle_enabled))
return true;
- scx_kf_error("built-in idle tracking is disabled");
+ scx_error(sch, "built-in idle tracking is disabled");
return false;
}
@@ -880,7 +880,7 @@ static s32 select_cpu_from_kfunc(struct scx_sched *sch, struct task_struct *p,
if (scx_kf_allowed_if_unlocked()) {
rq = task_rq_lock(p, &rf);
} else {
- if (!scx_kf_allowed(SCX_KF_SELECT_CPU | SCX_KF_ENQUEUE))
+ if (!scx_kf_allowed(sch, SCX_KF_SELECT_CPU | SCX_KF_ENQUEUE))
return -EPERM;
rq = scx_locked_rq();
}
@@ -1048,7 +1048,7 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask(void)
return cpu_none_mask;
if (static_branch_unlikely(&scx_builtin_idle_per_node)) {
- scx_kf_error("SCX_OPS_BUILTIN_IDLE_PER_NODE enabled");
+ scx_error(sch, "SCX_OPS_BUILTIN_IDLE_PER_NODE enabled");
return cpu_none_mask;
}
@@ -1107,7 +1107,7 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask(void)
return cpu_none_mask;
if (static_branch_unlikely(&scx_builtin_idle_per_node)) {
- scx_kf_error("SCX_OPS_BUILTIN_IDLE_PER_NODE enabled");
+ scx_error(sch, "SCX_OPS_BUILTIN_IDLE_PER_NODE enabled");
return cpu_none_mask;
}
@@ -1235,7 +1235,7 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed,
return -ENODEV;
if (static_branch_maybe(CONFIG_NUMA, &scx_builtin_idle_per_node)) {
- scx_kf_error("per-node idle tracking is enabled");
+ scx_error(sch, "per-node idle tracking is enabled");
return -EBUSY;
}
@@ -1316,10 +1316,17 @@ __bpf_kfunc s32 scx_bpf_pick_any_cpu_node(const struct cpumask *cpus_allowed,
__bpf_kfunc s32 scx_bpf_pick_any_cpu(const struct cpumask *cpus_allowed,
u64 flags)
{
+ struct scx_sched *sch;
s32 cpu;
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return -ENODEV;
+
if (static_branch_maybe(CONFIG_NUMA, &scx_builtin_idle_per_node)) {
- scx_kf_error("per-node idle tracking is enabled");
+ scx_error(sch, "per-node idle tracking is enabled");
return -EBUSY;
}
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 14/46] sched_ext: Misc updates around scx_sched instance pointer
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (12 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 13/46] sched_ext: Drop scx_kf_exit() and scx_kf_error() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 15/46] sched_ext: Keep dying tasks on a separate list Tejun Heo
` (31 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation for multiple scheduler support:
- Add the @sch parameter to find_global_dsq() and refill_task_slice_dfl().
- Restructure scx_allow_ttwu_queue() and make it read scx_root into $sch.
- Make RCU protection in scx_dsq_move() and scx_bpf_dsq_move_to_local()
explicit.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 62 ++++++++++++++++++++++++++++++----------------
1 file changed, 40 insertions(+), 22 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 0c99a55f199b..32306203fba5 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -181,10 +181,9 @@ static bool u32_before(u32 a, u32 b)
return (s32)(a - b) < 0;
}
-static struct scx_dispatch_q *find_global_dsq(struct task_struct *p)
+static struct scx_dispatch_q *find_global_dsq(struct scx_sched *sch,
+ struct task_struct *p)
{
- struct scx_sched *sch = scx_root;
-
return sch->global_dsqs[cpu_to_node(task_cpu(p))];
}
@@ -880,10 +879,10 @@ static void dsq_mod_nr(struct scx_dispatch_q *dsq, s32 delta)
WRITE_ONCE(dsq->nr, dsq->nr + delta);
}
-static void refill_task_slice_dfl(struct task_struct *p)
+static void refill_task_slice_dfl(struct scx_sched *sch, struct task_struct *p)
{
p->scx.slice = SCX_SLICE_DFL;
- __scx_add_event(scx_root, SCX_EV_REFILL_SLICE_DFL, 1);
+ __scx_add_event(sch, SCX_EV_REFILL_SLICE_DFL, 1);
}
static void dispatch_enqueue(struct scx_sched *sch, struct scx_dispatch_q *dsq,
@@ -901,7 +900,7 @@ static void dispatch_enqueue(struct scx_sched *sch, struct scx_dispatch_q *dsq,
scx_error(sch, "attempting to dispatch to a destroyed dsq");
/* fall back to the global dsq */
raw_spin_unlock(&dsq->lock);
- dsq = find_global_dsq(p);
+ dsq = find_global_dsq(sch, p);
raw_spin_lock(&dsq->lock);
}
}
@@ -1080,20 +1079,20 @@ static struct scx_dispatch_q *find_dsq_for_dispatch(struct scx_sched *sch,
s32 cpu = dsq_id & SCX_DSQ_LOCAL_CPU_MASK;
if (!ops_cpu_valid(sch, cpu, "in SCX_DSQ_LOCAL_ON dispatch verdict"))
- return find_global_dsq(p);
+ return find_global_dsq(sch, p);
return &cpu_rq(cpu)->scx.local_dsq;
}
if (dsq_id == SCX_DSQ_GLOBAL)
- dsq = find_global_dsq(p);
+ dsq = find_global_dsq(sch, p);
else
dsq = find_user_dsq(sch, dsq_id);
if (unlikely(!dsq)) {
scx_error(sch, "non-existent DSQ 0x%llx for %s[%d]",
dsq_id, p->comm, p->pid);
- return find_global_dsq(p);
+ return find_global_dsq(sch, p);
}
return dsq;
@@ -1272,15 +1271,15 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
* higher priority it becomes from scx_prio_less()'s POV.
*/
touch_core_sched(rq, p);
- refill_task_slice_dfl(p);
+ refill_task_slice_dfl(sch, p);
local_norefill:
dispatch_enqueue(sch, &rq->scx.local_dsq, p, enq_flags);
return;
global:
touch_core_sched(rq, p); /* see the comment in local: */
- refill_task_slice_dfl(p);
- dispatch_enqueue(sch, find_global_dsq(p), p, enq_flags);
+ refill_task_slice_dfl(sch, p);
+ dispatch_enqueue(sch, find_global_dsq(sch, p), p, enq_flags);
}
static bool task_runnable(const struct task_struct *p)
@@ -1692,7 +1691,7 @@ static struct rq *move_task_between_dsqs(struct scx_sched *sch,
dst_rq = container_of(dst_dsq, struct rq, scx.local_dsq);
if (src_rq != dst_rq &&
unlikely(!task_can_run_on_remote_rq(sch, p, dst_rq, true))) {
- dst_dsq = find_global_dsq(p);
+ dst_dsq = find_global_dsq(sch, p);
dst_rq = src_rq;
}
} else {
@@ -1848,7 +1847,7 @@ static void dispatch_to_local_dsq(struct scx_sched *sch, struct rq *rq,
if (src_rq != dst_rq &&
unlikely(!task_can_run_on_remote_rq(sch, p, dst_rq, true))) {
- dispatch_enqueue(sch, find_global_dsq(p), p,
+ dispatch_enqueue(sch, find_global_dsq(sch, p), p,
enq_flags | SCX_ENQ_CLEAR_OPSS);
return;
}
@@ -2380,7 +2379,7 @@ static struct task_struct *pick_task_scx(struct rq *rq)
if (keep_prev) {
p = prev;
if (!p->scx.slice)
- refill_task_slice_dfl(p);
+ refill_task_slice_dfl(rcu_dereference_sched(scx_root), p);
} else {
p = first_local_task(rq);
if (!p) {
@@ -2391,14 +2390,14 @@ static struct task_struct *pick_task_scx(struct rq *rq)
}
if (unlikely(!p->scx.slice)) {
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = rcu_dereference_sched(scx_root);
if (!scx_rq_bypassing(rq) && !sch->warned_zero_slice) {
printk_deferred(KERN_WARNING "sched_ext: %s[%d] has zero slice in %s()\n",
p->comm, p->pid, __func__);
sch->warned_zero_slice = true;
}
- refill_task_slice_dfl(p);
+ refill_task_slice_dfl(sch, p);
}
}
@@ -2487,7 +2486,7 @@ static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flag
cpu = scx_select_cpu_dfl(p, prev_cpu, wake_flags, NULL, 0);
if (cpu >= 0) {
- refill_task_slice_dfl(p);
+ refill_task_slice_dfl(sch, p);
p->scx.ddsp_dsq_id = SCX_DSQ_LOCAL;
} else {
cpu = prev_cpu;
@@ -3572,9 +3571,22 @@ bool task_should_scx(int policy)
bool scx_allow_ttwu_queue(const struct task_struct *p)
{
- return !scx_enabled() ||
- (scx_root->ops.flags & SCX_OPS_ALLOW_QUEUED_WAKEUP) ||
- p->sched_class != &ext_sched_class;
+ struct scx_sched *sch;
+
+ if (!scx_enabled())
+ return true;
+
+ sch = rcu_dereference_sched(scx_root);
+ if (unlikely(!sch))
+ return true;
+
+ if (scx_root->ops.flags & SCX_OPS_ALLOW_QUEUED_WAKEUP)
+ return true;
+
+ if (unlikely(p->sched_class != &ext_sched_class))
+ return true;
+
+ return false;
}
/**
@@ -5537,9 +5549,15 @@ __bpf_kfunc void scx_bpf_dispatch_cancel(void)
*/
__bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id)
{
- struct scx_sched *sch = scx_root;
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct scx_dispatch_q *dsq;
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (unlikely(!sch))
+ return false;
if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return false;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 15/46] sched_ext: Keep dying tasks on a separate list
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (13 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 14/46] sched_ext: Misc updates around scx_sched instance pointer Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 16/46] sched_ext: Implement cgroup subtree iteration for scx_task_iter Tejun Heo
` (30 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
sched_ext needs to be able to iterate all tasks until the task finishes its
final scheduling event which can be after it is dropped out of pid hash. The
scx_tasks list keeps track of all tasks between fork and free for this
purpose.
The planned cgroup sub-scheduler support requires selectively iterating
tasks belonging to cgroup subtrees. While the live tasks can be iterated
using css_task_iter, there is a similar gap after removal from css. As there
won't be too many tasks between release and free at any moment, this can be
reasonably addressed by scanning dying tasks looking for tasks which are
inside the cgroup subtree.
In preparation, split out scx_tasks lists into two - scx_live_tasks and
scx_dying_tasks. scx_task_iter is updated to first iterate scx_live_tasks
and then scx_dying_tasks. As tasks can only move from the former to the
latter, tasks can't escape iteration; however, it's possible for newly dying
tasks to appear twice during iteration. The users of scx_task_iter are
updated to handle duplicate iterations.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/sched/ext.h | 2 ++
kernel/exit.c | 1 +
kernel/sched/ext.c | 70 +++++++++++++++++++++++++++++----------
3 files changed, 55 insertions(+), 18 deletions(-)
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index d82b7a9b0658..7290c4354ad6 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -207,6 +207,7 @@ struct sched_ext_entity {
struct list_head tasks_node;
};
+void sched_ext_exit(struct task_struct *p);
void sched_ext_free(struct task_struct *p);
void print_scx_info(const char *log_lvl, struct task_struct *p);
void scx_softlockup(u32 dur_s);
@@ -214,6 +215,7 @@ bool scx_rcu_cpu_stall(void);
#else /* !CONFIG_SCHED_CLASS_EXT */
+static inline void sched_ext_exit(struct task_struct *p) {}
static inline void sched_ext_free(struct task_struct *p) {}
static inline void print_scx_info(const char *log_lvl, struct task_struct *p) {}
static inline void scx_softlockup(u32 dur_s) {}
diff --git a/kernel/exit.c b/kernel/exit.c
index 343eb97543d5..0bdbaa85ef43 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -962,6 +962,7 @@ void __noreturn do_exit(long code)
exit_thread(tsk);
sched_autogroup_exit_task(tsk);
+ sched_ext_exit(tsk);
cgroup_exit(tsk);
/*
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 32306203fba5..6ae9ee5b9a50 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -20,13 +20,21 @@
static struct scx_sched __rcu *scx_root;
/*
- * During exit, a task may schedule after losing its PIDs. When disabling the
- * BPF scheduler, we need to be able to iterate tasks in every state to
- * guarantee system safety. Maintain a dedicated task list which contains every
- * task between its fork and eventual free.
+ * - We want to visit and perform sleepable operations on every task.
+ *
+ * - During exit, a task may schedule after losing its PIDs. When disabling the
+ * BPF scheduler, we need to be able to iterate tasks in every state to
+ * guarantee system safety.
+ *
+ * Maintain a dedicated task list which contains every task between its fork and
+ * eventual free. Live and exiting tasks are kept on separate lists so that the
+ * dying task iteration can be combined with cgroup task iteration. This leads
+ * to occasional double visiting of existing tasks but those aren't difficult
+ * to handle from users.
*/
static DEFINE_SPINLOCK(scx_tasks_lock);
-static LIST_HEAD(scx_tasks);
+static LIST_HEAD(scx_live_tasks);
+static LIST_HEAD(scx_dying_tasks);
/* ops enable/disable */
static DEFINE_MUTEX(scx_enable_mutex);
@@ -436,6 +444,7 @@ struct bpf_iter_scx_dsq {
* SCX task iterator.
*/
struct scx_task_iter {
+ struct list_head *head;
struct sched_ext_entity cursor;
struct task_struct *locked_task;
struct rq *rq;
@@ -458,7 +467,8 @@ struct scx_task_iter {
* RCU read lock or obtaining a reference count.
*
* All tasks which existed when the iteration started are guaranteed to be
- * visited as long as they still exist.
+ * visited as long as they still exist. Tasks which exit while iteration is in
+ * progress may be visited twice. The caller must be able to handle such cases.
*/
static void scx_task_iter_start(struct scx_task_iter *iter)
{
@@ -467,8 +477,9 @@ static void scx_task_iter_start(struct scx_task_iter *iter)
spin_lock_irq(&scx_tasks_lock);
+ iter->head = &scx_live_tasks;
iter->cursor = (struct sched_ext_entity){ .flags = SCX_TASK_CURSOR };
- list_add(&iter->cursor.tasks_node, &scx_tasks);
+ list_add(&iter->cursor.tasks_node, &scx_live_tasks);
iter->locked_task = NULL;
iter->cnt = 0;
iter->list_locked = true;
@@ -527,9 +538,11 @@ static void scx_task_iter_stop(struct scx_task_iter *iter)
* scx_task_iter_next - Next task
* @iter: iterator to walk
*
- * Visit the next task. See scx_task_iter_start() for details. Locks are dropped
- * and re-acquired every %SCX_TASK_ITER_BATCH iterations to avoid causing stalls
- * by holding scx_tasks_lock for too long.
+ * Visit the next task. Existing tasks may be visited twice. See
+ * scx_task_iter_start() for details.
+ *
+ * Locks are dropped and re-acquired every %SCX_TASK_ITER_BATCH iterations to
+ * avoid causing stalls by holding scx_tasks_lock for too long.
*/
static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
{
@@ -543,18 +556,23 @@ static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
cond_resched();
__scx_task_iter_maybe_relock(iter);
}
-
+retry:
list_for_each_entry(pos, cursor, tasks_node) {
- if (&pos->tasks_node == &scx_tasks)
- return NULL;
+ if (&pos->tasks_node == iter->head)
+ break;
if (!(pos->flags & SCX_TASK_CURSOR)) {
list_move(cursor, &pos->tasks_node);
return container_of(pos, struct task_struct, scx);
}
}
- /* can't happen, should always terminate at scx_tasks above */
- BUG();
+ if (iter->head == &scx_live_tasks) {
+ iter->head = &scx_dying_tasks;
+ list_move(cursor, iter->head);
+ goto retry;
+ }
+
+ return NULL;
}
/**
@@ -562,8 +580,8 @@ static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
* @iter: iterator to walk
*
* Visit the non-idle task with its rq lock held. Allows callers to specify
- * whether they would like to filter out dead tasks. See scx_task_iter_start()
- * for details.
+ * whether they would like to filter out dead tasks. Exiting tasks may be
+ * visited twice. See scx_task_iter_start() for details.
*/
static struct task_struct *scx_task_iter_next_locked(struct scx_task_iter *iter)
{
@@ -2905,7 +2923,7 @@ void scx_post_fork(struct task_struct *p)
}
spin_lock_irq(&scx_tasks_lock);
- list_add_tail(&p->scx.tasks_node, &scx_tasks);
+ list_add_tail(&p->scx.tasks_node, &scx_live_tasks);
spin_unlock_irq(&scx_tasks_lock);
percpu_up_read(&scx_fork_rwsem);
@@ -2926,6 +2944,13 @@ void scx_cancel_fork(struct task_struct *p)
percpu_up_read(&scx_fork_rwsem);
}
+void sched_ext_exit(struct task_struct *p)
+{
+ spin_lock_irq(&scx_tasks_lock);
+ list_move_tail(&p->scx.tasks_node, &scx_dying_tasks);
+ spin_unlock_irq(&scx_tasks_lock);
+}
+
void sched_ext_free(struct task_struct *p)
{
unsigned long flags;
@@ -3913,6 +3938,7 @@ static void scx_disable_workfn(struct kthread_work *work)
scx_task_iter_start(&sti);
while ((p = scx_task_iter_next_locked(&sti))) {
+ /* @p may be being visited twice, doesn't matter */
const struct sched_class *old_class = p->sched_class;
const struct sched_class *new_class =
__setscheduler_class(p->policy, p->prio);
@@ -4615,6 +4641,13 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
scx_task_iter_start(&sti);
while ((p = scx_task_iter_next_locked(&sti))) {
+ /*
+ * Task iteration may visit the same task twice when racing
+ * against exiting. Skip if @p is already initialized.
+ */
+ if (scx_get_task_state(p) == SCX_TASK_READY)
+ continue;
+
/*
* @p may already be dead, have lost all its usages counts and
* be waiting for RCU grace period before being freed. @p can't
@@ -4657,6 +4690,7 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
percpu_down_write(&scx_fork_rwsem);
scx_task_iter_start(&sti);
while ((p = scx_task_iter_next_locked(&sti))) {
+ /* @p may be being visited twice, doesn't matter */
const struct sched_class *old_class = p->sched_class;
const struct sched_class *new_class =
__setscheduler_class(p->policy, p->prio);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 16/46] sched_ext: Implement cgroup subtree iteration for scx_task_iter
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (14 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 15/46] sched_ext: Keep dying tasks on a separate list Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 17/46] sched_ext: Add @kargs to scx_fork() Tejun Heo
` (29 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
For the planned cgroup sub-scheduler support, enable/disable operations are
going to be subtree specific and iterating all tasks in the system for those
operations can be unnecessarily expensive and disruptive.
cgroup already has mechanisms to perform subtree task iterations. Implement
cgroup subtree iteration for scx_task_iter:
- Add optional @cgrp to scx_task_iter_start() which enables cgroup subtree
iteration.
- Make scx_task_iter use combination of css_next_descendant_pre() and
css_task_iter to iterate all live tasks for cgroup iterations.
- After live task iteration is finished, scan scx_dying_tasks and only visit
tasks that are in the cgroup subtree. As scx_dying_tasks is most likely
really short, this should be pretty cheap.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 76 ++++++++++++++++++++++++++++++++++++++++------
1 file changed, 67 insertions(+), 9 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 6ae9ee5b9a50..ca8221378924 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -451,11 +451,17 @@ struct scx_task_iter {
struct rq_flags rf;
u32 cnt;
bool list_locked;
+#ifdef CONFIG_CGROUPS
+ struct cgroup *cgrp;
+ struct cgroup_subsys_state *css_pos;
+ struct css_task_iter css_iter;
+#endif
};
/**
* scx_task_iter_start - Lock scx_tasks_lock and start a task iteration
* @iter: iterator to init
+ * @cgrp: Optional root of cgroup subhierarchy to iterate
*
* Initialize @iter and return with scx_tasks_lock held. Once initialized, @iter
* must eventually be stopped with scx_task_iter_stop().
@@ -469,8 +475,14 @@ struct scx_task_iter {
* All tasks which existed when the iteration started are guaranteed to be
* visited as long as they still exist. Tasks which exit while iteration is in
* progress may be visited twice. The caller must be able to handle such cases.
+ *
+ * @if @cgrp is NULL, scx_live_tasks are walked followed by scx_dying_tasks. If
+ * @cgrp is not NULL, @cgrp's tasks are walked using css_task_iter followed by
+ * scx_dying_tasks. To guarantee that all tasks are visited at least once, the
+ * caller must be holding scx_fork_rwsem. In the cgroup case, the caller must
+ * also be holding scx_cgroup_rwsem to prevent cgroup task migrations.
*/
-static void scx_task_iter_start(struct scx_task_iter *iter)
+static void scx_task_iter_start(struct scx_task_iter *iter, struct cgroup *cgrp)
{
BUILD_BUG_ON(__SCX_DSQ_ITER_ALL_FLAGS &
((1U << __SCX_DSQ_LNODE_PRIV_SHIFT) - 1));
@@ -478,8 +490,20 @@ static void scx_task_iter_start(struct scx_task_iter *iter)
spin_lock_irq(&scx_tasks_lock);
iter->head = &scx_live_tasks;
+#ifdef CONFIG_CGROUPS
+ if (cgrp) {
+ iter->cgrp = cgrp;
+ iter->css_pos = css_next_descendant_pre(NULL, &iter->cgrp->self);
+ css_task_iter_start(iter->css_pos, 0, &iter->css_iter);
+ /* walking cgroup tasks instead, skip scx_live_tasks */
+ iter->head = &scx_dying_tasks;
+ } else {
+ iter->cgrp = NULL;
+ iter->css_pos = NULL;
+ }
+#endif
iter->cursor = (struct sched_ext_entity){ .flags = SCX_TASK_CURSOR };
- list_add(&iter->cursor.tasks_node, &scx_live_tasks);
+ list_add(&iter->cursor.tasks_node, iter->head);
iter->locked_task = NULL;
iter->cnt = 0;
iter->list_locked = true;
@@ -530,6 +554,8 @@ static void __scx_task_iter_maybe_relock(struct scx_task_iter *iter)
static void scx_task_iter_stop(struct scx_task_iter *iter)
{
__scx_task_iter_maybe_relock(iter);
+ if (iter->css_pos)
+ css_task_iter_end(&iter->css_iter);
list_del_init(&iter->cursor.tasks_node);
scx_task_iter_unlock(iter);
}
@@ -557,13 +583,45 @@ static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
__scx_task_iter_maybe_relock(iter);
}
retry:
+
+#ifdef CONFIG_CGROUPS
+ /*
+ * For cgroup iterations, use css_task_iter for live tasks. iter->head
+ * is already set to scx_dying_tasks.
+ */
+ while (iter->css_pos) {
+ struct task_struct *p;
+
+ p = css_task_iter_next(&iter->css_iter);
+ if (p)
+ return p;
+
+ css_task_iter_end(&iter->css_iter);
+ iter->css_pos = css_next_descendant_pre(iter->css_pos,
+ &iter->cgrp->self);
+ if (iter->css_pos)
+ css_task_iter_start(iter->css_pos, 0, &iter->css_iter);
+ }
+#endif
+
list_for_each_entry(pos, cursor, tasks_node) {
+ struct task_struct *p = container_of(pos, struct task_struct, scx);
+
if (&pos->tasks_node == iter->head)
break;
- if (!(pos->flags & SCX_TASK_CURSOR)) {
- list_move(cursor, &pos->tasks_node);
- return container_of(pos, struct task_struct, scx);
- }
+ if (pos->flags & SCX_TASK_CURSOR)
+ continue;
+#ifdef CONFIG_CGROUPS
+ /*
+ * For cgroup iterations, this loop is only used for iterating
+ * dying tasks. Filter out tasks which aren't in the target
+ * subtree.
+ */
+ if (iter->cgrp && !task_under_cgroup_hierarchy(p, iter->cgrp))
+ continue;
+#endif
+ list_move(cursor, &pos->tasks_node);
+ return p;
}
if (iter->head == &scx_live_tasks) {
@@ -3936,7 +3994,7 @@ static void scx_disable_workfn(struct kthread_work *work)
scx_init_task_enabled = false;
- scx_task_iter_start(&sti);
+ scx_task_iter_start(&sti, NULL);
while ((p = scx_task_iter_next_locked(&sti))) {
/* @p may be being visited twice, doesn't matter */
const struct sched_class *old_class = p->sched_class;
@@ -4639,7 +4697,7 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
if (ret)
goto err_disable_unlock_all;
- scx_task_iter_start(&sti);
+ scx_task_iter_start(&sti, NULL);
while ((p = scx_task_iter_next_locked(&sti))) {
/*
* Task iteration may visit the same task twice when racing
@@ -4688,7 +4746,7 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
* scx_tasks_lock.
*/
percpu_down_write(&scx_fork_rwsem);
- scx_task_iter_start(&sti);
+ scx_task_iter_start(&sti, NULL);
while ((p = scx_task_iter_next_locked(&sti))) {
/* @p may be being visited twice, doesn't matter */
const struct sched_class *old_class = p->sched_class;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 17/46] sched_ext: Add @kargs to scx_fork()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (15 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 16/46] sched_ext: Implement cgroup subtree iteration for scx_task_iter Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 18/46] sched/core: Swap the order between sched_post_fork() and cgroup_post_fork() Tejun Heo
` (28 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69
Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo, Peter Zijlstra
Make sched_cgroup_fork() pass @kargs to scx_fork(). This will be used to
determine @p's cgroup for cgroup sub-sched support.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
---
kernel/sched/core.c | 2 +-
kernel/sched/ext.c | 2 +-
kernel/sched/ext.h | 4 ++--
3 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 27dda808ed83..f2c6aaf4cc77 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4805,7 +4805,7 @@ int sched_cgroup_fork(struct task_struct *p, struct kernel_clone_args *kargs)
p->sched_class->task_fork(p);
raw_spin_unlock_irqrestore(&p->pi_lock, flags);
- return scx_fork(p);
+ return scx_fork(p, kargs);
}
void sched_cancel_fork(struct task_struct *p)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index ca8221378924..ddceae539e11 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2950,7 +2950,7 @@ void scx_pre_fork(struct task_struct *p)
percpu_down_read(&scx_fork_rwsem);
}
-int scx_fork(struct task_struct *p)
+int scx_fork(struct task_struct *p, struct kernel_clone_args *kargs)
{
percpu_rwsem_assert_held(&scx_fork_rwsem);
diff --git a/kernel/sched/ext.h b/kernel/sched/ext.h
index 43429b33e52c..0b7fc46aee08 100644
--- a/kernel/sched/ext.h
+++ b/kernel/sched/ext.h
@@ -11,7 +11,7 @@
void scx_tick(struct rq *rq);
void init_scx_entity(struct sched_ext_entity *scx);
void scx_pre_fork(struct task_struct *p);
-int scx_fork(struct task_struct *p);
+int scx_fork(struct task_struct *p, struct kernel_clone_args *kargs);
void scx_post_fork(struct task_struct *p);
void scx_cancel_fork(struct task_struct *p);
bool scx_can_stop_tick(struct rq *rq);
@@ -44,7 +44,7 @@ bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
static inline void scx_tick(struct rq *rq) {}
static inline void scx_pre_fork(struct task_struct *p) {}
-static inline int scx_fork(struct task_struct *p) { return 0; }
+static inline int scx_fork(struct task_struct *p, struct kernel_clone_args *kargs) { return 0; }
static inline void scx_post_fork(struct task_struct *p) {}
static inline void scx_cancel_fork(struct task_struct *p) {}
static inline u32 scx_cpuperf_target(s32 cpu) { return 0; }
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 18/46] sched/core: Swap the order between sched_post_fork() and cgroup_post_fork()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (16 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 17/46] sched_ext: Add @kargs to scx_fork() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 19/46] cgroup: Expose some cgroup helpers Tejun Heo
` (27 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The planned sched_ext cgroup sub-scheduler support needs the newly forked
task to be associated with its cgroup in its post_fork() hook. There is no
existing ordering requirement between the two now. Swap them and note the
new ordering requirement.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/fork.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/kernel/fork.c b/kernel/fork.c
index af673856499d..6e5a1397d2e6 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -2415,8 +2415,12 @@ __latent_entropy struct task_struct *copy_process(
fd_install(pidfd, pidfile);
proc_fork_connector(p);
- sched_post_fork(p);
+ /*
+ * sched_ext needs @p to be associated with its cgroup in its post_fork
+ * hook. cgroup_post_fork() should come before sched_post_fork().
+ */
cgroup_post_fork(p, args);
+ sched_post_fork(p);
perf_event_fork(p);
trace_task_newtask(p, clone_flags);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 19/46] cgroup: Expose some cgroup helpers
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (17 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 18/46] sched/core: Swap the order between sched_post_fork() and cgroup_post_fork() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 20/46] sched_ext: Update p->scx.disallow warning in scx_init_task() Tejun Heo
` (26 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Expose the following through cgroup.h:
- cgroup_on_dfl()
- cgroup_is_dead()
- cgroup_for_each_live_child()
- cgroup_for_each_live_descendant_pre()
- cgroup_for_each_live_descendant_post()
Until now, these didn't need to be exposed because controllers only cared
about the css hierarchy. The planned sched_ext hierarchical scheduler
support will be based on the default cgroup hierarchy, which is in line
with the existing BPF cgroup support, and thus needs these exposed.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/cgroup.h | 65 ++++++++++++++++++++++++++++++++-
kernel/cgroup/cgroup-internal.h | 6 ---
kernel/cgroup/cgroup.c | 55 ----------------------------
3 files changed, 63 insertions(+), 63 deletions(-)
diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index b18fb5fcb38e..3abdff370dd3 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -41,6 +41,14 @@ struct kernel_clone_args;
#ifdef CONFIG_CGROUPS
+/*
+ * To avoid confusing the compiler (and generating warnings) with code
+ * that attempts to access what would be a 0-element array (i.e. sized
+ * to a potentially empty array when CGROUP_SUBSYS_COUNT == 0), this
+ * constant expression can be added.
+ */
+#define CGROUP_HAS_SUBSYS_CONFIG (CGROUP_SUBSYS_COUNT > 0)
+
enum css_task_iter_flags {
CSS_TASK_ITER_PROCS = (1U << 0), /* walk only threadgroup leaders */
CSS_TASK_ITER_THREADED = (1U << 1), /* walk all threaded css_sets in the domain */
@@ -75,6 +83,7 @@ enum cgroup_lifetime_events {
extern struct file_system_type cgroup_fs_type;
extern struct cgroup_root cgrp_dfl_root;
extern struct css_set init_css_set;
+extern struct mutex cgroup_mutex;
extern spinlock_t css_set_lock;
extern struct blocking_notifier_head cgroup_lifetime_notifier;
@@ -102,6 +111,8 @@ extern struct blocking_notifier_head cgroup_lifetime_notifier;
#define cgroup_subsys_on_dfl(ss) \
static_branch_likely(&ss ## _on_dfl_key)
+bool cgroup_on_dfl(const struct cgroup *cgrp);
+
bool css_has_online_children(struct cgroup_subsys_state *css);
struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss);
struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgroup,
@@ -272,6 +283,32 @@ void css_task_iter_end(struct css_task_iter *it);
for ((pos) = css_next_descendant_post(NULL, (css)); (pos); \
(pos) = css_next_descendant_post((pos), (css)))
+/* iterate over child cgrps, lock should be held throughout iteration */
+#define cgroup_for_each_live_child(child, cgrp) \
+ list_for_each_entry((child), &(cgrp)->self.children, self.sibling) \
+ if (({ lockdep_assert_held(&cgroup_mutex); \
+ cgroup_is_dead(child); })) \
+ ; \
+ else
+
+/* walk live descendants in pre order */
+#define cgroup_for_each_live_descendant_pre(dsct, d_css, cgrp) \
+ css_for_each_descendant_pre((d_css), cgroup_css((cgrp), NULL)) \
+ if (({ lockdep_assert_held(&cgroup_mutex); \
+ (dsct) = (d_css)->cgroup; \
+ cgroup_is_dead(dsct); })) \
+ ; \
+ else
+
+/* walk live descendants in postorder */
+#define cgroup_for_each_live_descendant_post(dsct, d_css, cgrp) \
+ css_for_each_descendant_post((d_css), cgroup_css((cgrp), NULL)) \
+ if (({ lockdep_assert_held(&cgroup_mutex); \
+ (dsct) = (d_css)->cgroup; \
+ cgroup_is_dead(dsct); })) \
+ ; \
+ else
+
/**
* cgroup_taskset_for_each - iterate cgroup_taskset
* @task: the loop cursor
@@ -334,6 +371,27 @@ static inline u64 cgroup_id(const struct cgroup *cgrp)
return cgrp->kn->id;
}
+/**
+ * cgroup_css - obtain a cgroup's css for the specified subsystem
+ * @cgrp: the cgroup of interest
+ * @ss: the subsystem of interest (%NULL returns @cgrp->self)
+ *
+ * Return @cgrp's css (cgroup_subsys_state) associated with @ss. This
+ * function must be called either under cgroup_mutex or rcu_read_lock() and
+ * the caller is responsible for pinning the returned css if it wants to
+ * keep accessing it outside the said locks. This function may return
+ * %NULL if @cgrp doesn't have @subsys_id enabled.
+ */
+static inline struct cgroup_subsys_state *cgroup_css(struct cgroup *cgrp,
+ struct cgroup_subsys *ss)
+{
+ if (CGROUP_HAS_SUBSYS_CONFIG && ss)
+ return rcu_dereference_check(cgrp->subsys[ss->id],
+ lockdep_is_held(&cgroup_mutex));
+ else
+ return &cgrp->self;
+}
+
/**
* css_is_dying - test whether the specified css is dying
* @css: target css
@@ -365,6 +423,11 @@ static inline bool css_is_self(struct cgroup_subsys_state *css)
return false;
}
+static inline bool cgroup_is_dead(const struct cgroup *cgrp)
+{
+ return !(cgrp->self.flags & CSS_ONLINE);
+}
+
static inline void cgroup_get(struct cgroup *cgrp)
{
css_get(&cgrp->self);
@@ -380,8 +443,6 @@ static inline void cgroup_put(struct cgroup *cgrp)
css_put(&cgrp->self);
}
-extern struct mutex cgroup_mutex;
-
static inline void cgroup_lock(void)
{
mutex_lock(&cgroup_mutex);
diff --git a/kernel/cgroup/cgroup-internal.h b/kernel/cgroup/cgroup-internal.h
index b14e61c64a34..31b6f4cb05ba 100644
--- a/kernel/cgroup/cgroup-internal.h
+++ b/kernel/cgroup/cgroup-internal.h
@@ -184,11 +184,6 @@ extern bool cgrp_dfl_visible;
for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT && \
(((ss) = cgroup_subsys[ssid]) || true); (ssid)++)
-static inline bool cgroup_is_dead(const struct cgroup *cgrp)
-{
- return !(cgrp->self.flags & CSS_ONLINE);
-}
-
static inline bool notify_on_release(const struct cgroup *cgrp)
{
return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
@@ -222,7 +217,6 @@ static inline void get_css_set(struct css_set *cset)
}
bool cgroup_ssid_enabled(int ssid);
-bool cgroup_on_dfl(const struct cgroup *cgrp);
struct cgroup_root *cgroup_root_from_kf(struct kernfs_root *kf_root);
struct cgroup *task_cgroup_from_root(struct task_struct *task,
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 312c6a8b55bb..4107af27965f 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -69,14 +69,6 @@
/* let's not notify more than 100 times per second */
#define CGROUP_FILE_NOTIFY_MIN_INTV DIV_ROUND_UP(HZ, 100)
-/*
- * To avoid confusing the compiler (and generating warnings) with code
- * that attempts to access what would be a 0-element array (i.e. sized
- * to a potentially empty array when CGROUP_SUBSYS_COUNT == 0), this
- * constant expression can be added.
- */
-#define CGROUP_HAS_SUBSYS_CONFIG (CGROUP_SUBSYS_COUNT > 0)
-
/*
* cgroup_mutex is the master lock. Any modification to cgroup or its
* hierarchy must be performed while holding it.
@@ -480,27 +472,6 @@ static u16 cgroup_ss_mask(struct cgroup *cgrp)
return cgrp->root->subsys_mask;
}
-/**
- * cgroup_css - obtain a cgroup's css for the specified subsystem
- * @cgrp: the cgroup of interest
- * @ss: the subsystem of interest (%NULL returns @cgrp->self)
- *
- * Return @cgrp's css (cgroup_subsys_state) associated with @ss. This
- * function must be called either under cgroup_mutex or rcu_read_lock() and
- * the caller is responsible for pinning the returned css if it wants to
- * keep accessing it outside the said locks. This function may return
- * %NULL if @cgrp doesn't have @subsys_id enabled.
- */
-static struct cgroup_subsys_state *cgroup_css(struct cgroup *cgrp,
- struct cgroup_subsys *ss)
-{
- if (CGROUP_HAS_SUBSYS_CONFIG && ss)
- return rcu_dereference_check(cgrp->subsys[ss->id],
- lockdep_is_held(&cgroup_mutex));
- else
- return &cgrp->self;
-}
-
/**
* cgroup_e_css_by_mask - obtain a cgroup's effective css for the specified ss
* @cgrp: the cgroup of interest
@@ -712,32 +683,6 @@ EXPORT_SYMBOL_GPL(of_css);
} \
} while (false)
-/* iterate over child cgrps, lock should be held throughout iteration */
-#define cgroup_for_each_live_child(child, cgrp) \
- list_for_each_entry((child), &(cgrp)->self.children, self.sibling) \
- if (({ lockdep_assert_held(&cgroup_mutex); \
- cgroup_is_dead(child); })) \
- ; \
- else
-
-/* walk live descendants in pre order */
-#define cgroup_for_each_live_descendant_pre(dsct, d_css, cgrp) \
- css_for_each_descendant_pre((d_css), cgroup_css((cgrp), NULL)) \
- if (({ lockdep_assert_held(&cgroup_mutex); \
- (dsct) = (d_css)->cgroup; \
- cgroup_is_dead(dsct); })) \
- ; \
- else
-
-/* walk live descendants in postorder */
-#define cgroup_for_each_live_descendant_post(dsct, d_css, cgrp) \
- css_for_each_descendant_post((d_css), cgroup_css((cgrp), NULL)) \
- if (({ lockdep_assert_held(&cgroup_mutex); \
- (dsct) = (d_css)->cgroup; \
- cgroup_is_dead(dsct); })) \
- ; \
- else
-
/*
* The default css_set - used by init and its children prior to any
* hierarchies being mounted. It contains a pointer to the root state
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 20/46] sched_ext: Update p->scx.disallow warning in scx_init_task()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (18 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 19/46] cgroup: Expose some cgroup helpers Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 21/46] sched_ext: Minor reorganization of enable/disable path Tejun Heo
` (25 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
- Always trigger the warning if p->scx.disallow is set for fork inits. There
is no reason to set it during forks.
- Flip the positions of if/else arms to ease adding error conditions.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index ddceae539e11..0ee716ff4dab 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2825,7 +2825,10 @@ static int scx_init_task(struct task_struct *p, struct task_group *tg, bool fork
scx_set_task_state(p, SCX_TASK_INIT);
if (p->scx.disallow) {
- if (!fork) {
+ if (unlikely(fork)) {
+ scx_error(sch, "ops.init_task() set task->scx.disallow for %s[%d] during fork",
+ p->comm, p->pid);
+ } else {
struct rq *rq;
struct rq_flags rf;
@@ -2844,9 +2847,6 @@ static int scx_init_task(struct task_struct *p, struct task_group *tg, bool fork
}
task_rq_unlock(rq, p, &rf);
- } else if (p->policy == SCHED_EXT) {
- scx_error(sch, "ops.init_task() set task->scx.disallow for %s[%d] during fork",
- p->comm, p->pid);
}
}
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 21/46] sched_ext: Minor reorganization of enable/disable path
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (19 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 20/46] sched_ext: Update p->scx.disallow warning in scx_init_task() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 22/46] sched_ext: Factor out scx_claim_exit() from scx_disable() Tejun Heo
` (24 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
- Separate out scx_root_disable() from scx_disable_workfn().
- Rename scx_enable() to scx_root_enable().
- Add @sch to scx_disable(). The callers are now responsible for providing
the scx_sched to disable.
No functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 70 ++++++++++++++++++++++++++--------------------
1 file changed, 40 insertions(+), 30 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 0ee716ff4dab..54f65a196d94 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -3018,8 +3018,8 @@ void sched_ext_free(struct task_struct *p)
spin_unlock_irqrestore(&scx_tasks_lock, flags);
/*
- * @p is off scx_tasks and wholly ours. scx_enable()'s READY -> ENABLED
- * transitions can't race us. Disable ops for @p.
+ * @p is off scx_tasks and wholly ours. scx_root_enable()'s READY ->
+ * ENABLED transitions can't race us. Disable ops for @p.
*/
if (scx_get_task_state(p) != SCX_TASK_NONE) {
struct rq_flags rf;
@@ -3933,24 +3933,12 @@ static const char *scx_exit_reason(enum scx_exit_kind kind)
}
}
-static void scx_disable_workfn(struct kthread_work *work)
+static void scx_root_disable(struct scx_sched *sch)
{
- struct scx_sched *sch = container_of(work, struct scx_sched, disable_work);
struct scx_exit_info *ei = sch->exit_info;
struct scx_task_iter sti;
struct task_struct *p;
- int kind, cpu;
-
- kind = atomic_read(&sch->exit_kind);
- while (true) {
- if (kind == SCX_EXIT_DONE) /* already disabled? */
- return;
- WARN_ON_ONCE(kind == SCX_EXIT_NONE);
- if (atomic_try_cmpxchg(&sch->exit_kind, &kind, SCX_EXIT_DONE))
- break;
- }
- ei->kind = kind;
- ei->reason = scx_exit_reason(ei->kind);
+ int cpu;
/* guarantee forward progress by bypassing scx_ops */
scx_bypass(true);
@@ -4078,21 +4066,35 @@ static void scx_disable_workfn(struct kthread_work *work)
scx_bypass(false);
}
-static void scx_disable(enum scx_exit_kind kind)
+static void scx_disable_workfn(struct kthread_work *work)
+{
+ struct scx_sched *sch = container_of(work, struct scx_sched, disable_work);
+ struct scx_exit_info *ei = sch->exit_info;
+ int kind;
+
+ kind = atomic_read(&sch->exit_kind);
+ while (true) {
+ if (kind == SCX_EXIT_DONE) /* already disabled? */
+ return;
+ WARN_ON_ONCE(kind == SCX_EXIT_NONE);
+ if (atomic_try_cmpxchg(&sch->exit_kind, &kind, SCX_EXIT_DONE))
+ break;
+ }
+ ei->kind = kind;
+ ei->reason = scx_exit_reason(ei->kind);
+
+ scx_root_disable(sch);
+}
+
+static void scx_disable(struct scx_sched *sch, enum scx_exit_kind kind)
{
int none = SCX_EXIT_NONE;
- struct scx_sched *sch;
if (WARN_ON_ONCE(kind == SCX_EXIT_NONE || kind == SCX_EXIT_DONE))
kind = SCX_EXIT_ERROR;
- rcu_read_lock();
- sch = rcu_dereference(scx_root);
- if (sch) {
- atomic_try_cmpxchg(&sch->exit_kind, &none, kind);
- kthread_queue_work(sch->helper, &sch->disable_work);
- }
- rcu_read_unlock();
+ atomic_try_cmpxchg(&sch->exit_kind, &none, kind);
+ kthread_queue_work(sch->helper, &sch->disable_work);
}
static void dump_newline(struct seq_buf *s)
@@ -4558,7 +4560,7 @@ static int validate_ops(struct scx_sched *sch, const struct sched_ext_ops *ops)
return 0;
}
-static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
+static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
{
struct scx_sched *sch;
struct scx_task_iter sti;
@@ -4808,7 +4810,7 @@ static int scx_enable(struct sched_ext_ops *ops, struct bpf_link *link)
* Flush scx_disable_work to ensure that error is reported before init
* completion. sch's base reference will be put by bpf_scx_unreg().
*/
- scx_error(sch, "scx_enable() failed (%d)", ret);
+ scx_error(sch, "scx_root_enable() failed (%d)", ret);
kthread_flush_work(&sch->disable_work);
return 0;
}
@@ -4940,7 +4942,7 @@ static int bpf_scx_check_member(const struct btf_type *t,
static int bpf_scx_reg(void *kdata, struct bpf_link *link)
{
- return scx_enable(kdata, link);
+ return scx_root_enable(kdata, link);
}
static void bpf_scx_unreg(void *kdata, struct bpf_link *link)
@@ -4948,7 +4950,7 @@ static void bpf_scx_unreg(void *kdata, struct bpf_link *link)
struct sched_ext_ops *ops = kdata;
struct scx_sched *sch = ops->priv;
- scx_disable(SCX_EXIT_UNREG);
+ scx_disable(sch, SCX_EXIT_UNREG);
kthread_flush_work(&sch->disable_work);
kobject_put(&sch->kobj);
}
@@ -5074,7 +5076,15 @@ static struct bpf_struct_ops bpf_sched_ext_ops = {
static void sysrq_handle_sched_ext_reset(u8 key)
{
- scx_disable(SCX_EXIT_SYSRQ);
+ struct scx_sched *sch;
+
+ rcu_read_lock();
+ sch = rcu_dereference(scx_root);
+ if (likely(sch))
+ scx_disable(sch, SCX_EXIT_SYSRQ);
+ else
+ pr_info("sched_ext: BPF schedulers not loaded\n");
+ rcu_read_unlock();
}
static const struct sysrq_key_op sysrq_sched_ext_reset_op = {
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 22/46] sched_ext: Factor out scx_claim_exit() from scx_disable()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (20 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 21/46] sched_ext: Minor reorganization of enable/disable path Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 23/46] sched_ext: Introduce cgroup sub-sched support Tejun Heo
` (23 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Both scx_disable() and scx_vexit() use cmpxchg on sch->exit_kind to claim
exit ownership but the latter lacks some sanity checks. Factor out
scx_claim_exit() from scx_disable() and use it in both places. This will
also be used to implement sub-sched disabling.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 54f65a196d94..b608c2c04730 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4086,15 +4086,20 @@ static void scx_disable_workfn(struct kthread_work *work)
scx_root_disable(sch);
}
-static void scx_disable(struct scx_sched *sch, enum scx_exit_kind kind)
+static bool scx_claim_exit(struct scx_sched *sch, enum scx_exit_kind kind)
{
int none = SCX_EXIT_NONE;
if (WARN_ON_ONCE(kind == SCX_EXIT_NONE || kind == SCX_EXIT_DONE))
kind = SCX_EXIT_ERROR;
- atomic_try_cmpxchg(&sch->exit_kind, &none, kind);
- kthread_queue_work(sch->helper, &sch->disable_work);
+ return atomic_try_cmpxchg(&sch->exit_kind, &none, kind);
+}
+
+static void scx_disable(struct scx_sched *sch, enum scx_exit_kind kind)
+{
+ if (scx_claim_exit(sch, kind))
+ kthread_queue_work(sch->helper, &sch->disable_work);
}
static void dump_newline(struct seq_buf *s)
@@ -4412,9 +4417,8 @@ static void scx_vexit(struct scx_sched *sch,
const char *fmt, va_list args)
{
struct scx_exit_info *ei = sch->exit_info;
- int none = SCX_EXIT_NONE;
- if (!atomic_try_cmpxchg(&sch->exit_kind, &none, kind))
+ if (!scx_claim_exit(sch, kind))
return;
ei->exit_code = exit_code;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 23/46] sched_ext: Introduce cgroup sub-sched support
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (21 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 22/46] sched_ext: Factor out scx_claim_exit() from scx_disable() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 24/46] HACK_NOT_FOR_UPSTREAM: BPF: Implement prog grouping hack Tejun Heo
` (22 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
A system often runs multiple workloads especially in multi-tenant server
environments where a system is split into partitions servicing separate
more-or-less independent workloads each requiring an application-specific
scheduler. To support such and other use cases, sched_ext is in the process
of growing multiple scheduler support.
When partitioning a system in terms of CPUs for such use cases, an
oft-taken approach is hard partitioning the system using cpuset. While it
would be possible to tie sched_ext multiple scheduler support to cpuset
partitions, such an approach would have fundamental limitations stemming
from the lack of dynamism and flexibility.
Users often don't care which specific CPUs are assigned to which workload
and want to take advantage of optimizations which are enabled by running
workloads on a larger machine - e.g. opportunistic over-commit, improving
latency critical workload characteristics while maintaining bandwidth
fairness, employing control mechanisms based on different criteria than
on-CPU time for e.g. flexible memory bandwidth isolation, packing similar
parts from different workloads on same L3s to improve cache efficiency,
and so on.
As this sort of dynamic behaviors are impossible or difficult to implement
with hard partitioning, sched_ext is implementing cgroup sub-sched support
where schedulers can be attached to the cgroup hierarchy and a parent
scheduler is responsible for controlling the CPUs that each child can use
at any given moment. This makes CPU distribution dynamically controlled by
BPF allowing high flexibility.
This patch adds the skeletal sched_ext cgroup sub-sched support:
- sched_ext_ops.sub_cgroup_id and .sub_attach/detach() are added. Non-zero
sub_cgroup_id indicates that the scheduler is to be attached to the
identified cgroup. A sub-sched is attached to the cgroup iff the nearest
ancestor scheduler implements .sub_attach() and grants the attachment. Max
nesting depth is limited by SCX_SUB_MAX_DEPTH.
- When a scheduler exits, all its descendant schedulers are exited
together. Also, cgroup.scx_sched added which points to the effective
scheduler instance for the cgroup. This is updated on scheduler
init/exit and inherited on cgroup online. When a cgroup is offlined, the
attached scheduler is automatically exited.
- Sub-sched support is gated on CONFIG_EXT_SUB_SCHED which is
automatically enabled if both SCX and cgroups are enabled. Sub-sched
support is not tied to the CPU controller but rather the cgroup
hierarchy itself. This is intentional as the support for cpu.weight and
cpu.max based resource control is orthogonal to sub-sched support. Note
that CONFIG_CGROUPS around cgroup subtree iteration support for
scx_task_iter is replaced with CONFIG_EXT_SUB_SCHED for consistency.
- This allows loading sub-scheds and most framework operations such as
propagating disable down the hierarchy work. However, sub-scheds are not
operational yet and all tasks stay with the root sched. This will serve
as the basis for building up full sub-sched support.
- DSQs point to the scx_sched they belong to.
- scx_qmap is updated to allow attachment of sub-scheds and also serving
as sub-scheds.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/cgroup-defs.h | 4 +
include/linux/sched/ext.h | 3 +
init/Kconfig | 4 +
kernel/sched/ext.c | 503 +++++++++++++++++++++++++++++++--
kernel/sched/ext_internal.h | 65 ++++-
tools/sched_ext/scx_qmap.bpf.c | 9 +-
tools/sched_ext/scx_qmap.c | 13 +-
7 files changed, 567 insertions(+), 34 deletions(-)
diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 6b93a64115fe..56ea297981f5 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -17,6 +17,7 @@
#include <linux/refcount.h>
#include <linux/percpu-refcount.h>
#include <linux/percpu-rwsem.h>
+#include <linux/sched.h>
#include <linux/u64_stats_sync.h>
#include <linux/workqueue.h>
#include <linux/bpf-cgroup-defs.h>
@@ -590,6 +591,9 @@ struct cgroup {
#ifdef CONFIG_BPF_SYSCALL
struct bpf_local_storage __rcu *bpf_cgrp_storage;
#endif
+#ifdef CONFIG_EXT_SUB_SCHED
+ struct scx_sched __rcu *scx_sched;
+#endif
/* All ancestors including self */
struct cgroup *ancestors[];
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 7290c4354ad6..992c8b43db75 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -65,6 +65,7 @@ struct scx_dispatch_q {
u64 id;
struct rhash_head hash_node;
struct llist_node free_node;
+ struct scx_sched *sched;
struct rcu_head rcu;
};
@@ -136,6 +137,8 @@ struct scx_dsq_list_node {
u32 priv; /* can be used by iter cursor */
};
+struct scx_sched;
+
/*
* The following is embedded in task_struct and contains all fields necessary
* for a task to be scheduled by SCX.
diff --git a/init/Kconfig b/init/Kconfig
index 836320251219..4ca8c048dfee 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1137,6 +1137,10 @@ config EXT_GROUP_SCHED
endif #CGROUP_SCHED
+config EXT_SUB_SCHED
+ def_bool y
+ depends on SCHED_CLASS_EXT
+
config SCHED_MM_CID
def_bool y
depends on SMP && RSEQ
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index b608c2c04730..5eb1d6919595 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -9,6 +9,8 @@
#include <linux/btf_ids.h>
#include "ext_idle.h"
+static DEFINE_RAW_SPINLOCK(scx_sched_lock);
+
/*
* NOTE: sched_ext is in the process of growing multiple scheduler support and
* scx_root usage is in a transitional state. Naked dereferences are safe if the
@@ -19,6 +21,12 @@
*/
static struct scx_sched __rcu *scx_root;
+/*
+ * All scheds, writers must hold both scx_enable_mutex and scx_sched_lock.
+ * Readers can hold either or rcu_read_lock().
+ */
+static LIST_HEAD(scx_sched_all);
+
/*
* - We want to visit and perform sleepable operations on every task.
*
@@ -144,6 +152,7 @@ static struct kset *scx_kset;
#include <trace/events/sched_ext.h>
static void process_ddsp_deferred_locals(struct rq *rq);
+static void scx_disable(struct scx_sched *sch, enum scx_exit_kind kind);
static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags);
static void scx_vexit(struct scx_sched *sch, enum scx_exit_kind kind,
s64 exit_code, const char *fmt, va_list args);
@@ -189,6 +198,82 @@ static bool u32_before(u32 a, u32 b)
return (s32)(a - b) < 0;
}
+/**
+ * scx_parent - Find the parent sched
+ * @sch: sched to find the parent of
+ *
+ * Returns the parent scheduler or %NULL if @sch is root.
+ */
+static struct scx_sched *scx_parent(struct scx_sched *sch)
+{
+ if (sch->level)
+ return sch->ancestors[sch->level - 1];
+ else
+ return NULL;
+}
+
+#ifdef CONFIG_EXT_SUB_SCHED
+static struct scx_sched *scx_next_descendant_pre(struct scx_sched *pos,
+ struct scx_sched *root)
+{
+ struct scx_sched *next;
+
+ lockdep_assert(lockdep_is_held(&scx_enable_mutex) ||
+ lockdep_is_held(&scx_sched_lock));
+
+ /* if first iteration, visit @root */
+ if (!pos)
+ return root;
+
+ /* visit the first child if exists */
+ next = list_first_entry_or_null(&pos->children, struct scx_sched, sibling);
+ if (next)
+ return next;
+
+ /* no child, visit my or the closest ancestor's next sibling */
+ while (pos != root) {
+ if (!list_is_last(&pos->sibling, &scx_parent(pos)->children))
+ return list_next_entry(pos, sibling);
+ pos = scx_parent(pos);
+ }
+
+ return NULL;
+}
+#else /* CONFIG_EXT_SUB_SCHED */
+static struct scx_sched *scx_next_descendant_pre(struct scx_sched *pos,
+ struct scx_sched *root)
+{
+ return pos ? NULL : root;
+}
+#endif /* CONFIG_EXT_SUB_SCHED */
+
+/**
+ * scx_is_descendant - Test whether sched is a descendant
+ * @sch: sched to test
+ * @ancestor: ancestor sched to test against
+ *
+ * Test whether @sch is a descendant of @ancestor.
+ */
+static bool scx_is_descendant(struct scx_sched *sch, struct scx_sched *ancestor)
+{
+ if (sch->level < ancestor->level)
+ return false;
+ return sch->ancestors[ancestor->level] == ancestor;
+}
+
+/**
+ * scx_for_each_descendant_pre - pre-order walk of a sched's descendants
+ * @pos: iteration cursor
+ * @root: sched to walk the descendants of
+ *
+ * Walk @root's descendants. @root is included in the iteration and the first
+ * node to be visited. Must be called with either scx_enable_mutex or
+ * scx_sched_lock held.
+ */
+#define scx_for_each_descendant_pre(pos, root) \
+ for ((pos) = scx_next_descendant_pre(NULL, (root)); (pos); \
+ (pos) = scx_next_descendant_pre((pos), (root)))
+
static struct scx_dispatch_q *find_global_dsq(struct scx_sched *sch,
struct task_struct *p)
{
@@ -451,7 +536,7 @@ struct scx_task_iter {
struct rq_flags rf;
u32 cnt;
bool list_locked;
-#ifdef CONFIG_CGROUPS
+#ifdef CONFIG_EXT_SUB_SCHED
struct cgroup *cgrp;
struct cgroup_subsys_state *css_pos;
struct css_task_iter css_iter;
@@ -490,7 +575,7 @@ static void scx_task_iter_start(struct scx_task_iter *iter, struct cgroup *cgrp)
spin_lock_irq(&scx_tasks_lock);
iter->head = &scx_live_tasks;
-#ifdef CONFIG_CGROUPS
+#ifdef CONFIG_EXT_SUB_SCHED
if (cgrp) {
iter->cgrp = cgrp;
iter->css_pos = css_next_descendant_pre(NULL, &iter->cgrp->self);
@@ -584,7 +669,7 @@ static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
}
retry:
-#ifdef CONFIG_CGROUPS
+#ifdef CONFIG_EXT_SUB_SCHED
/*
* For cgroup iterations, use css_task_iter for live tasks. iter->head
* is already set to scx_dying_tasks.
@@ -611,7 +696,7 @@ static struct task_struct *scx_task_iter_next(struct scx_task_iter *iter)
break;
if (pos->flags & SCX_TASK_CURSOR)
continue;
-#ifdef CONFIG_CGROUPS
+#ifdef CONFIG_EXT_SUB_SCHED
/*
* For cgroup iterations, this loop is only used for iterating
* dying tasks. Filter out tasks which aren't in the target
@@ -2825,7 +2910,10 @@ static int scx_init_task(struct task_struct *p, struct task_group *tg, bool fork
scx_set_task_state(p, SCX_TASK_INIT);
if (p->scx.disallow) {
- if (unlikely(fork)) {
+ if (unlikely(scx_parent(sch))) {
+ scx_error(sch, "non-root ops.init_task() set task->scx.disallow for %s[%d]",
+ p->comm, p->pid);
+ } else if (unlikely(fork)) {
scx_error(sch, "ops.init_task() set task->scx.disallow for %s[%d] during fork",
p->comm, p->pid);
} else {
@@ -3284,25 +3372,39 @@ void scx_group_set_bandwidth(struct task_group *tg,
percpu_up_read(&scx_cgroup_ops_rwsem);
}
+#endif /* CONFIG_EXT_GROUP_SCHED */
+#if defined(CONFIG_EXT_GROUP_SCHED) || defined(CONFIG_EXT_SUB_SCHED)
static void scx_cgroup_lock(void)
{
+#ifdef CONFIG_EXT_GROUP_SCHED
percpu_down_write(&scx_cgroup_ops_rwsem);
+#endif
cgroup_lock();
}
static void scx_cgroup_unlock(void)
{
cgroup_unlock();
+#ifdef CONFIG_EXT_GROUP_SCHED
percpu_up_write(&scx_cgroup_ops_rwsem);
+#endif
}
-#else /* CONFIG_EXT_GROUP_SCHED */
+/* for each descendant of @cgrp including self, set ->scx_sched to @sch */
+static void set_cgroup_sched(struct cgroup *cgrp, struct scx_sched *sch)
+{
+ struct cgroup *pos;
+ struct cgroup_subsys_state *css;
+ cgroup_for_each_live_descendant_pre(pos, css, cgrp)
+ rcu_assign_pointer(pos->scx_sched, sch);
+}
+#else /* CONFIG_EXT_GROUP_SCHED || CONFIG_EXT_SUB_SCHED */
static void scx_cgroup_lock(void) {}
static void scx_cgroup_unlock(void) {}
-
-#endif /* CONFIG_EXT_GROUP_SCHED */
+static void set_cgroup_sched(struct cgroup *cgrp, struct scx_sched *sch) {}
+#endif /* CONFIG_EXT_GROUP_SCHED || CONFIG_EXT_SUB_SCHED */
/*
* Omitted operations:
@@ -3352,13 +3454,15 @@ DEFINE_SCHED_CLASS(ext) = {
#endif
};
-static void init_dsq(struct scx_dispatch_q *dsq, u64 dsq_id)
+static void init_dsq(struct scx_dispatch_q *dsq, u64 dsq_id,
+ struct scx_sched *sch)
{
memset(dsq, 0, sizeof(*dsq));
raw_spin_lock_init(&dsq->lock);
INIT_LIST_HEAD(&dsq->list);
dsq->id = dsq_id;
+ dsq->sched = sch;
}
static void free_dsq_irq_workfn(struct irq_work *irq_work)
@@ -3554,6 +3658,11 @@ static void scx_sched_free_rcu_work(struct work_struct *work)
struct scx_dispatch_q *dsq;
int node;
+#ifdef CONFIG_EXT_SUB_SCHED
+ kfree(sch->cgrp_path);
+ if (sch->cgrp)
+ cgroup_put(sch->cgrp);
+#endif /* CONFIG_EXT_SUB_SCHED */
kthread_stop(sch->helper->task);
free_percpu(sch->pcpu);
@@ -3922,6 +4031,8 @@ static const char *scx_exit_reason(enum scx_exit_kind kind)
return "unregistered from the main kernel";
case SCX_EXIT_SYSRQ:
return "disabled by sysrq-S";
+ case SCX_EXIT_PARENT:
+ return "parent exiting";
case SCX_EXIT_ERROR:
return "runtime error";
case SCX_EXIT_ERROR_BPF:
@@ -3933,6 +4044,85 @@ static const char *scx_exit_reason(enum scx_exit_kind kind)
}
}
+#ifdef CONFIG_EXT_SUB_SCHED
+static DECLARE_WAIT_QUEUE_HEAD(scx_unlink_waitq);
+
+/* propagate disable to all proper descendants and wait for them to unlink */
+static void scx_propagate_disable_and_flush(struct scx_sched *sch)
+{
+ struct scx_sched *child;
+
+ /*
+ * Trigger them in parallel. Each has a dedicated helper kthread and can
+ * run in parallel. While most of disabling is serialized, running them
+ * in separate threads allows excluding ops.exit(), which can take
+ * arbitrarily long, from the critical path and thus avoids staying in
+ * the bypass mode for prolonged time.
+ */
+ raw_spin_lock_irq(&scx_sched_lock);
+ list_for_each_entry(child, &sch->children, sibling)
+ scx_disable(child, SCX_EXIT_PARENT);
+ raw_spin_unlock_irq(&scx_sched_lock);
+
+ /*
+ * Child scheds that finished the critical part of disabling will take
+ * themselves off @sch->children. Wait for it to drian. As propagation
+ * is recursive, empty @sch->children means that all proper descendant
+ * scheds reached unlinking stage.
+ */
+ wait_event(scx_unlink_waitq, list_empty(&sch->children));
+}
+
+static void scx_sub_disable(struct scx_sched *sch)
+{
+ struct scx_sched *parent = scx_parent(sch);
+
+ scx_propagate_disable_and_flush(sch);
+
+ mutex_lock(&scx_enable_mutex);
+ percpu_down_write(&scx_fork_rwsem);
+ scx_cgroup_lock();
+
+ set_cgroup_sched(sch->cgrp, parent);
+
+ /* TODO - perform actual disabling here */
+
+ scx_cgroup_unlock();
+ percpu_up_write(&scx_fork_rwsem);
+
+ raw_spin_lock_irq(&scx_sched_lock);
+ list_del_init(&sch->sibling);
+ list_del_rcu(&sch->all);
+ raw_spin_unlock_irq(&scx_sched_lock);
+
+ mutex_unlock(&scx_enable_mutex);
+
+ /*
+ * @sch is now unlinked from the parent's children list. Notify and call
+ * ops.sub_detach/exit(). Note that ops.sub_detach/exit() must be called
+ * after unlinking and releasing all locks. See
+ * scx_propagate_disable_and_flush().
+ */
+ wake_up_all(&scx_unlink_waitq);
+
+ if (sch->ops.sub_detach && sch->sub_attached) {
+ struct scx_sub_detach_args sub_detach_args = {
+ .ops = &sch->ops,
+ .cgroup_path = sch->cgrp_path,
+ };
+ SCX_CALL_OP(parent, SCX_KF_UNLOCKED, sub_detach, NULL,
+ &sub_detach_args);
+ }
+
+ if (sch->ops.exit)
+ SCX_CALL_OP(sch, SCX_KF_UNLOCKED, exit, NULL, sch->exit_info);
+ kobject_del(&sch->kobj);
+}
+#else /* CONFIG_EXT_SUB_SCHED */
+static void scx_propgate_disable(struct scx_sched *sch) { }
+static void scx_sub_disable(struct scx_sched *sch) { }
+#endif /* CONFIG_EXT_SUB_SCHED */
+
static void scx_root_disable(struct scx_sched *sch)
{
struct scx_exit_info *ei = sch->exit_info;
@@ -3940,8 +4130,9 @@ static void scx_root_disable(struct scx_sched *sch)
struct task_struct *p;
int cpu;
- /* guarantee forward progress by bypassing scx_ops */
+ /* guarantee forward progress and disable all descendants */
scx_bypass(true);
+ scx_propagate_disable_and_flush(sch);
switch (scx_set_enable_state(SCX_DISABLING)) {
case SCX_DISABLING:
@@ -4004,6 +4195,11 @@ static void scx_root_disable(struct scx_sched *sch)
scx_exit_task(p);
}
scx_task_iter_stop(&sti);
+
+ scx_cgroup_lock();
+ set_cgroup_sched(sch->cgrp, NULL);
+ scx_cgroup_unlock();
+
percpu_up_write(&scx_fork_rwsem);
/*
@@ -4040,6 +4236,10 @@ static void scx_root_disable(struct scx_sched *sch)
cancel_delayed_work_sync(&scx_watchdog_work);
+ raw_spin_lock_irq(&scx_sched_lock);
+ list_del_rcu(&sch->all);
+ raw_spin_unlock_irq(&scx_sched_lock);
+
/*
* scx_root clearing must be inside cpus_read_lock(). See
* handle_hotplug().
@@ -4083,7 +4283,10 @@ static void scx_disable_workfn(struct kthread_work *work)
ei->kind = kind;
ei->reason = scx_exit_reason(ei->kind);
- scx_root_disable(sch);
+ if (scx_parent(sch))
+ scx_sub_disable(sch);
+ else
+ scx_root_disable(sch);
}
static bool scx_claim_exit(struct scx_sched *sch, enum scx_exit_kind kind)
@@ -4438,12 +4641,15 @@ static void scx_vexit(struct scx_sched *sch,
irq_work_queue(&sch->error_irq_work);
}
-static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops)
+static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
+ struct cgroup *cgrp,
+ struct scx_sched *parent)
{
struct scx_sched *sch;
- int node, ret;
+ s32 level = parent ? parent->level + 1 : 0;
+ s32 node, ret;
- sch = kzalloc(sizeof(*sch), GFP_KERNEL);
+ sch = kzalloc(struct_size(sch, ancestors, level), GFP_KERNEL);
if (!sch)
return ERR_PTR(-ENOMEM);
@@ -4473,7 +4679,7 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops)
goto err_free_gdsqs;
}
- init_dsq(dsq, SCX_DSQ_GLOBAL);
+ init_dsq(dsq, SCX_DSQ_GLOBAL, sch);
sch->global_dsqs[node] = dsq;
}
@@ -4486,6 +4692,12 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops)
goto err_free_pcpu;
sched_set_fifo(sch->helper->task);
+ if (parent)
+ memcpy(sch->ancestors, parent->ancestors,
+ level * sizeof(parent->ancestors[0]));
+ sch->ancestors[level] = sch;
+ sch->level = level;
+
atomic_set(&sch->exit_kind, SCX_EXIT_NONE);
init_irq_work(&sch->error_irq_work, scx_error_irq_workfn);
kthread_init_work(&sch->disable_work, scx_disable_workfn);
@@ -4493,10 +4705,46 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops)
ops->priv = sch;
sch->kobj.kset = scx_kset;
+
+#ifdef CONFIG_EXT_SUB_SCHED
+ char *buf = kzalloc(PATH_MAX, GFP_KERNEL);
+ if (!buf)
+ goto err_stop_helper;
+ cgroup_path(cgrp, buf, PATH_MAX);
+ sch->cgrp_path = kstrdup(buf, GFP_KERNEL);
+ kfree(buf);
+ if (!sch->cgrp_path)
+ goto err_stop_helper;
+
+ sch->cgrp = cgrp;
+ INIT_LIST_HEAD(&sch->children);
+ INIT_LIST_HEAD(&sch->sibling);
+
+ if (parent)
+ ret = kobject_init_and_add(&sch->kobj, &scx_ktype,
+ &parent->sub_kset->kobj,
+ "sub-%llu", cgroup_id(cgrp));
+ else
+ ret = kobject_init_and_add(&sch->kobj, &scx_ktype, NULL, "root");
+
+ if (ret < 0) {
+ kfree(sch->cgrp_path);
+ goto err_stop_helper;
+ }
+
+ if (ops->sub_attach) {
+ sch->sub_kset = kset_create_and_add("sub", NULL, &sch->kobj);
+ if (!sch->sub_kset) {
+ kobject_put(&sch->kobj);
+ return ERR_PTR(-ENOMEM);
+ }
+ }
+
+#else /* CONFIG_EXT_SUB_SCHED */
ret = kobject_init_and_add(&sch->kobj, &scx_ktype, NULL, "root");
if (ret < 0)
goto err_stop_helper;
-
+#endif /* CONFIG_EXT_SUB_SCHED */
return sch;
err_stop_helper:
@@ -4585,7 +4833,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
goto err_unlock;
}
- sch = scx_alloc_and_add_sched(ops);
+ sch = scx_alloc_and_add_sched(ops, &cgrp_dfl_root.cgrp, NULL);
if (IS_ERR(sch)) {
ret = PTR_ERR(sch);
goto err_unlock;
@@ -4600,8 +4848,12 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
atomic_long_set(&scx_nr_rejected, 0);
- for_each_possible_cpu(cpu)
- cpu_rq(cpu)->scx.cpuperf_target = SCX_CPUPERF_ONE;
+ for_each_possible_cpu(cpu) {
+ struct rq *rq = cpu_rq(cpu);
+
+ rq->scx.local_dsq.sched = sch;
+ rq->scx.cpuperf_target = SCX_CPUPERF_ONE;
+ }
/*
* Keep CPUs stable during enable so that the BPF scheduler can track
@@ -4615,6 +4867,10 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
*/
rcu_assign_pointer(scx_root, sch);
+ raw_spin_lock_irq(&scx_sched_lock);
+ list_add_tail_rcu(&sch->all, &scx_sched_all);
+ raw_spin_unlock_irq(&scx_sched_lock);
+
scx_idle_enable(ops);
if (sch->ops.init) {
@@ -4699,6 +4955,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
* never sees uninitialized tasks.
*/
scx_cgroup_lock();
+ set_cgroup_sched(sch->cgrp, sch);
ret = scx_cgroup_init(sch);
if (ret)
goto err_disable_unlock_all;
@@ -4819,6 +5076,181 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
return 0;
}
+#ifdef CONFIG_EXT_SUB_SCHED
+/* verify that a scheduler can be attached to @cgrp and return the parent */
+static struct scx_sched *find_parent_sched(struct cgroup *cgrp)
+{
+ struct scx_sched *parent = cgrp->scx_sched;
+ struct scx_sched *pos;
+
+ lockdep_assert_held(&scx_sched_lock);
+
+ /* can't attach twice to the same cgroup */
+ if (parent->cgrp == cgrp)
+ return ERR_PTR(-EBUSY);
+
+ /* does $parent allow sub-scheds? */
+ if (!parent->ops.sub_attach)
+ return ERR_PTR(-EOPNOTSUPP);
+
+ /* can't insert between $parent and its exiting children */
+ list_for_each_entry(pos, &parent->children, sibling)
+ if (cgroup_is_descendant(pos->cgrp, cgrp))
+ return ERR_PTR(-EBUSY);
+
+ return parent;
+}
+
+static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
+{
+ struct cgroup *cgrp;
+ struct scx_sched *parent, *sch;
+ int ret;
+
+ mutex_lock(&scx_enable_mutex);
+
+ if (!scx_enabled()) {
+ ret = -ENODEV;
+ goto out_unlock;
+ }
+
+ cgrp = cgroup_get_from_id(ops->sub_cgroup_id);
+ if (IS_ERR(cgrp)) {
+ ret = PTR_ERR(cgrp);
+ goto out_unlock;
+ }
+
+ raw_spin_lock_irq(&scx_sched_lock);
+ parent = find_parent_sched(cgrp);
+ if (IS_ERR(parent)) {
+ raw_spin_unlock_irq(&scx_sched_lock);
+ ret = PTR_ERR(parent);
+ goto out_put_cgrp;
+ }
+ kobject_get(&parent->kobj);
+ raw_spin_unlock_irq(&scx_sched_lock);
+
+ sch = scx_alloc_and_add_sched(ops, cgrp, parent);
+ kobject_put(&parent->kobj);
+ if (IS_ERR(sch)) {
+ ret = PTR_ERR(sch);
+ goto out_put_cgrp;
+ }
+
+ raw_spin_lock_irq(&scx_sched_lock);
+ list_add_tail(&sch->sibling, &parent->children);
+ list_add_tail_rcu(&sch->all, &scx_sched_all);
+ raw_spin_unlock_irq(&scx_sched_lock);
+
+ if (sch->level >= SCX_SUB_MAX_DEPTH) {
+ scx_error(sch, "max nesting depth %d violated",
+ SCX_SUB_MAX_DEPTH);
+ goto err_disable;
+ }
+
+ if (sch->ops.init) {
+ ret = SCX_CALL_OP_RET(sch, SCX_KF_UNLOCKED, init, NULL);
+ if (ret) {
+ ret = ops_sanitize_err(sch, "init", ret);
+ scx_error(sch, "ops.init() failed (%d)", ret);
+ goto err_disable;
+ }
+ sch->exit_info->flags |= SCX_EFLAG_INITIALIZED;
+ }
+
+ if (validate_ops(sch, ops))
+ goto err_disable;
+
+ struct scx_sub_attach_args sub_attach_args = {
+ .ops = &sch->ops,
+ .cgroup_path = sch->cgrp_path,
+ };
+
+ ret = SCX_CALL_OP_RET(parent, SCX_KF_UNLOCKED, sub_attach, NULL,
+ &sub_attach_args);
+ if (ret) {
+ ret = ops_sanitize_err(sch, "sub_attach", ret);
+ scx_error(sch, "parent rejected (%d)", ret);
+ goto err_disable;
+ }
+ sch->sub_attached = true;
+
+ percpu_down_write(&scx_fork_rwsem);
+ scx_cgroup_lock();
+
+ /*
+ * Set cgroup->scx_sched's and check CSS_ONLINE. Either we see
+ * !CSS_ONLINE or scx_cgroup_lifetime_notify() sees and shoots us down.
+ */
+ set_cgroup_sched(sch->cgrp, sch);
+ if (!(cgrp->self.flags & CSS_ONLINE)) {
+ scx_error(sch, "cgroup is not online");
+ goto err_unlock_and_disable;
+ }
+
+ /* TODO - perform actual enabling here */
+
+ scx_cgroup_unlock();
+ percpu_up_write(&scx_fork_rwsem);
+
+ pr_info("sched_ext: BPF sub-scheduler \"%s\" enabled\n", sch->ops.name);
+ kobject_uevent(&sch->kobj, KOBJ_ADD);
+ ret = 0;
+ goto out_unlock;
+
+out_put_cgrp:
+ cgroup_put(cgrp);
+out_unlock:
+ mutex_unlock(&scx_enable_mutex);
+ return ret;
+
+err_unlock_and_disable:
+ scx_cgroup_unlock();
+ percpu_up_write(&scx_fork_rwsem);
+err_disable:
+ mutex_unlock(&scx_enable_mutex);
+ kthread_flush_work(&sch->disable_work);
+ return 0;
+}
+
+static int scx_cgroup_lifetime_notify(struct notifier_block *nb,
+ unsigned long action, void *data)
+{
+ struct cgroup *cgrp = data;
+ struct cgroup *parent = cgroup_parent(cgrp);
+
+ if (!cgroup_on_dfl(cgrp))
+ return NOTIFY_OK;
+
+ switch (action) {
+ case CGROUP_LIFETIME_ONLINE:
+ /* inherit ->scx_sched from $parent */
+ if (parent)
+ rcu_assign_pointer(cgrp->scx_sched, parent->scx_sched);
+ break;
+ case CGROUP_LIFETIME_OFFLINE:
+ /* if there is a sched attached, shoot it down */
+ if (cgrp->scx_sched && cgrp->scx_sched->cgrp == cgrp)
+ scx_exit(cgrp->scx_sched, SCX_EXIT_UNREG_KERN,
+ SCX_ECODE_RSN_CGROUP_OFFLINE,
+ "cgroup %llu going offline", cgroup_id(cgrp));
+ break;
+ }
+
+ return NOTIFY_OK;
+}
+
+static struct notifier_block scx_cgroup_lifetime_nb = {
+ .notifier_call = scx_cgroup_lifetime_notify,
+};
+
+static int __init scx_cgroup_lifetime_notifier_init(void)
+{
+ return blocking_notifier_chain_register(&cgroup_lifetime_notifier,
+ &scx_cgroup_lifetime_nb);
+}
+core_initcall(scx_cgroup_lifetime_notifier_init);
+#endif /* CONFIG_EXT_SUB_SCHED */
/********************************************************************************
* bpf_struct_ops plumbing.
@@ -4913,6 +5345,11 @@ static int bpf_scx_init_member(const struct btf_type *t,
case offsetof(struct sched_ext_ops, hotplug_seq):
ops->hotplug_seq = *(u64 *)(udata + moff);
return 1;
+#ifdef CONFIG_EXT_SUB_SCHED
+ case offsetof(struct sched_ext_ops, sub_cgroup_id):
+ ops->sub_cgroup_id = *(u64 *)(udata + moff);
+ return 1;
+#endif /* CONFIG_EXT_SUB_SCHED */
}
return 0;
@@ -4935,6 +5372,8 @@ static int bpf_scx_check_member(const struct btf_type *t,
case offsetof(struct sched_ext_ops, cpu_offline):
case offsetof(struct sched_ext_ops, init):
case offsetof(struct sched_ext_ops, exit):
+ case offsetof(struct sched_ext_ops, sub_attach):
+ case offsetof(struct sched_ext_ops, sub_detach):
break;
default:
if (prog->sleepable)
@@ -4946,7 +5385,13 @@ static int bpf_scx_check_member(const struct btf_type *t,
static int bpf_scx_reg(void *kdata, struct bpf_link *link)
{
- return scx_root_enable(kdata, link);
+ struct sched_ext_ops *ops = kdata;
+
+#ifdef CONFIG_EXT_SUB_SCHED
+ if (ops->sub_cgroup_id > 1)
+ return scx_sub_enable(ops, link);
+#endif /* CONFIG_EXT_SUB_SCHED */
+ return scx_root_enable(ops, link);
}
static void bpf_scx_unreg(void *kdata, struct bpf_link *link)
@@ -5011,7 +5456,9 @@ static void sched_ext_ops__cgroup_move(struct task_struct *p, struct cgroup *fro
static void sched_ext_ops__cgroup_cancel_move(struct task_struct *p, struct cgroup *from, struct cgroup *to) {}
static void sched_ext_ops__cgroup_set_weight(struct cgroup *cgrp, u32 weight) {}
static void sched_ext_ops__cgroup_set_bandwidth(struct cgroup *cgrp, u64 period_us, u64 quota_us, u64 burst_us) {}
-#endif
+#endif /* CONFIG_EXT_GROUP_SCHED */
+static s32 sched_ext_ops__sub_attach(struct scx_sub_attach_args *args) { return -EINVAL; }
+static void sched_ext_ops__sub_detach(struct scx_sub_detach_args *args) {}
static void sched_ext_ops__cpu_online(s32 cpu) {}
static void sched_ext_ops__cpu_offline(s32 cpu) {}
static s32 sched_ext_ops__init(void) { return -EINVAL; }
@@ -5050,6 +5497,8 @@ static struct sched_ext_ops __bpf_ops_sched_ext_ops = {
.cgroup_set_weight = sched_ext_ops__cgroup_set_weight,
.cgroup_set_bandwidth = sched_ext_ops__cgroup_set_bandwidth,
#endif
+ .sub_attach = sched_ext_ops__sub_attach,
+ .sub_detach = sched_ext_ops__sub_detach,
.cpu_online = sched_ext_ops__cpu_online,
.cpu_offline = sched_ext_ops__cpu_offline,
.init = sched_ext_ops__init,
@@ -5319,7 +5768,9 @@ void __init init_sched_ext_class(void)
struct rq *rq = cpu_rq(cpu);
int n = cpu_to_node(cpu);
- init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL);
+ /* local_dsq's local_dsqs will be set during scx_root_enable() */
+ init_dsq(&rq->scx.local_dsq, SCX_DSQ_LOCAL, NULL);
+
INIT_LIST_HEAD(&rq->scx.runnable_list);
INIT_LIST_HEAD(&rq->scx.ddsp_deferred_locals);
@@ -5905,16 +6356,16 @@ __bpf_kfunc s32 scx_bpf_create_dsq(u64 dsq_id, s32 node)
if (!dsq)
return -ENOMEM;
- init_dsq(dsq, dsq_id);
-
rcu_read_lock();
sch = rcu_dereference(scx_root);
- if (sch)
+ if (sch) {
+ init_dsq(dsq, dsq_id, sch);
ret = rhashtable_lookup_insert_fast(&sch->dsq_hash, &dsq->hash_node,
dsq_hash_params);
- else
+ } else {
ret = -ENODEV;
+ }
rcu_read_unlock();
if (ret)
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index b3617abed510..2e3aa3888ce0 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -23,6 +23,8 @@ enum scx_consts {
* scx_tasks_lock to avoid causing e.g. CSD and RCU stalls.
*/
SCX_TASK_ITER_BATCH = 32,
+
+ SCX_SUB_MAX_DEPTH = 4,
};
enum scx_exit_kind {
@@ -33,6 +35,7 @@ enum scx_exit_kind {
SCX_EXIT_UNREG_BPF, /* BPF-initiated unregistration */
SCX_EXIT_UNREG_KERN, /* kernel-initiated unregistration */
SCX_EXIT_SYSRQ, /* requested by 'S' sysrq */
+ SCX_EXIT_PARENT, /* parent exiting */
SCX_EXIT_ERROR = 1024, /* runtime error, error msg contains details */
SCX_EXIT_ERROR_BPF, /* ERROR but triggered through scx_bpf_error() */
@@ -57,6 +60,7 @@ enum scx_exit_kind {
enum scx_exit_code {
/* Reasons */
SCX_ECODE_RSN_HOTPLUG = 1LLU << 32,
+ SCX_ECODE_RSN_CGROUP_OFFLINE = 2LLU << 32,
/* Actions */
SCX_ECODE_ACT_RESTART = 1LLU << 48,
@@ -208,7 +212,7 @@ struct scx_exit_task_args {
bool cancelled;
};
-/* argument container for ops->cgroup_init() */
+/* argument container for ops.cgroup_init() */
struct scx_cgroup_init_args {
/* the weight of the cgroup [1..10000] */
u32 weight;
@@ -231,12 +235,12 @@ enum scx_cpu_preempt_reason {
};
/*
- * Argument container for ops->cpu_acquire(). Currently empty, but may be
+ * Argument container for ops.cpu_acquire(). Currently empty, but may be
* expanded in the future.
*/
struct scx_cpu_acquire_args {};
-/* argument container for ops->cpu_release() */
+/* argument container for ops.cpu_release() */
struct scx_cpu_release_args {
/* the reason the CPU was preempted */
enum scx_cpu_preempt_reason reason;
@@ -245,9 +249,7 @@ struct scx_cpu_release_args {
struct task_struct *task;
};
-/*
- * Informational context provided to dump operations.
- */
+/* informational context provided to dump operations */
struct scx_dump_ctx {
enum scx_exit_kind kind;
s64 exit_code;
@@ -256,6 +258,18 @@ struct scx_dump_ctx {
u64 at_jiffies;
};
+/* argument container for ops.sub_attach() */
+struct scx_sub_attach_args {
+ struct sched_ext_ops *ops;
+ char *cgroup_path;
+};
+
+/* argument container for ops.sub_detach() */
+struct scx_sub_detach_args {
+ struct sched_ext_ops *ops;
+ char *cgroup_path;
+};
+
/**
* struct sched_ext_ops - Operation table for BPF scheduler implementation
*
@@ -705,6 +719,20 @@ struct sched_ext_ops {
#endif /* CONFIG_EXT_GROUP_SCHED */
+ /**
+ * @sub_attach: Attach a sub-scheduler
+ * @args: argument container, see the struct definition
+ *
+ * Return 0 to accept the sub-scheduler. -errno to reject.
+ */
+ s32 (*sub_attach)(struct scx_sub_attach_args *args);
+
+ /**
+ * @sub_detach: Detach a sub-scheduler
+ * @args: argument container, see the struct definition
+ */
+ void (*sub_detach)(struct scx_sub_detach_args *args);
+
/*
* All online ops must come before ops.cpu_online().
*/
@@ -746,6 +774,10 @@ struct sched_ext_ops {
*/
void (*exit)(struct scx_exit_info *info);
+ /*
+ * Data fields must comes after all ops fields.
+ */
+
/**
* @dispatch_max_batch: Max nr of tasks that dispatch() can dispatch
*/
@@ -780,6 +812,12 @@ struct sched_ext_ops {
*/
u64 hotplug_seq;
+ /**
+ * @cgroup_id: When >1, attach the scheduler as a sub-scheduler on the
+ * specified cgroup.
+ */
+ u64 sub_cgroup_id;
+
/**
* @name: BPF scheduler's name
*
@@ -884,8 +922,20 @@ struct scx_sched {
struct scx_dispatch_q **global_dsqs;
struct scx_sched_pcpu __percpu *pcpu;
+ s32 level;
bool warned_zero_slice:1;
bool warned_deprecated_rq:1;
+ bool sub_attached:1;
+
+ struct list_head all;
+
+#ifdef CONFIG_EXT_SUB_SCHED
+ struct list_head children;
+ struct list_head sibling;
+ struct cgroup *cgrp;
+ char *cgrp_path;
+ struct kset *sub_kset;
+#endif /* CONFIG_EXT_SUB_SCHED */
atomic_t exit_kind;
struct scx_exit_info *exit_info;
@@ -896,6 +946,9 @@ struct scx_sched {
struct irq_work error_irq_work;
struct kthread_work disable_work;
struct rcu_work rcu_work;
+
+ /* all ancestors including self */
+ struct scx_sched *ancestors[];
};
enum scx_wake_flags {
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 3072b593f898..9631e4d04d88 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -41,6 +41,7 @@ const volatile u32 dsp_batch;
const volatile bool highpri_boosting;
const volatile bool print_dsqs_and_events;
const volatile bool print_msgs;
+const volatile u64 sub_cgroup_id;
const volatile s32 disallow_tgid;
const volatile bool suppress_dump;
@@ -848,7 +849,7 @@ s32 BPF_STRUCT_OPS_SLEEPABLE(qmap_init)
struct bpf_timer *timer;
s32 ret;
- if (print_msgs)
+ if (print_msgs && !sub_cgroup_id)
print_cpus();
ret = scx_bpf_create_dsq(SHARED_DSQ, -1);
@@ -874,6 +875,11 @@ void BPF_STRUCT_OPS(qmap_exit, struct scx_exit_info *ei)
UEI_RECORD(uei, ei);
}
+s32 BPF_STRUCT_OPS(qmap_sub_attach, struct scx_sub_attach_args *args)
+{
+ return 0;
+}
+
SCX_OPS_DEFINE(qmap_ops,
.select_cpu = (void *)qmap_select_cpu,
.enqueue = (void *)qmap_enqueue,
@@ -889,6 +895,7 @@ SCX_OPS_DEFINE(qmap_ops,
.cgroup_init = (void *)qmap_cgroup_init,
.cgroup_set_weight = (void *)qmap_cgroup_set_weight,
.cgroup_set_bandwidth = (void *)qmap_cgroup_set_bandwidth,
+ .sub_attach = (void *)qmap_sub_attach,
.cpu_online = (void *)qmap_cpu_online,
.cpu_offline = (void *)qmap_cpu_offline,
.init = (void *)qmap_init,
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index ef701d45ba43..5d762d10f4db 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -10,6 +10,7 @@
#include <inttypes.h>
#include <signal.h>
#include <libgen.h>
+#include <sys/stat.h>
#include <bpf/bpf.h>
#include <scx/common.h>
#include "scx_qmap.bpf.skel.h"
@@ -67,7 +68,7 @@ int main(int argc, char **argv)
skel->rodata->slice_ns = __COMPAT_ENUM_OR_ZERO("scx_public_consts", "SCX_SLICE_DFL");
- while ((opt = getopt(argc, argv, "s:e:t:T:l:b:PMHd:D:Spvh")) != -1) {
+ while ((opt = getopt(argc, argv, "s:e:t:T:l:b:PMHc:d:D:Spvh")) != -1) {
switch (opt) {
case 's':
skel->rodata->slice_ns = strtoull(optarg, NULL, 0) * 1000;
@@ -96,6 +97,16 @@ int main(int argc, char **argv)
case 'H':
skel->rodata->highpri_boosting = true;
break;
+ case 'c': {
+ struct stat st;
+ if (stat(optarg, &st) < 0) {
+ perror("stat");
+ return 1;
+ }
+ skel->struct_ops.qmap_ops->sub_cgroup_id = st.st_ino;
+ skel->rodata->sub_cgroup_id = st.st_ino;
+ break;
+ }
case 'd':
skel->rodata->disallow_tgid = strtol(optarg, NULL, 0);
if (skel->rodata->disallow_tgid < 0)
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 24/46] HACK_NOT_FOR_UPSTREAM: BPF: Implement prog grouping hack
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (22 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 23/46] sched_ext: Introduce cgroup sub-sched support Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 25/46] sched_ext: Introduce scx_task_sched[_rcu]() Tejun Heo
` (21 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Hopefully, we can have something better instead.
NOT_SIGNED_OFF
---
include/linux/bpf.h | 5 +++++
include/linux/sched.h | 2 ++
kernel/bpf/syscall.c | 23 +++++++++++++++++++++++
kernel/sched/ext.c | 36 ++++++++++++++++++++++++++++++++++++
tools/sched_ext/scx_qmap.c | 13 +++++++++++++
5 files changed, 79 insertions(+)
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index cc700925b802..5101ae3ba2b6 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -1581,6 +1581,11 @@ struct bpf_stream_stage {
struct bpf_prog_aux {
atomic64_t refcnt;
+
+ /* XXX - See kernel/sched/ext.c::scx_sub_enable() */
+ u64 priv_user;
+ void *priv;
+
u32 used_map_cnt;
u32 used_btf_cnt;
u32 max_ctx_offset;
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2b272382673d..576aed48beb2 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1596,6 +1596,8 @@ struct task_struct {
struct bpf_local_storage __rcu *bpf_storage;
/* Used for BPF run context */
struct bpf_run_ctx *bpf_ctx;
+ /* XXX - See kernel/sched/ext.c::scx_sub_enable() */
+ u64 bpf_prog_aux_priv;
#endif
/* Used by BPF for per-TASK xdp storage */
struct bpf_net_context *bpf_net_context;
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 0fbfa8532c39..e85dbe7fe5ce 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -2761,6 +2761,27 @@ static bool is_perfmon_prog_type(enum bpf_prog_type prog_type)
}
}
+static int prog_aux_priv_param_set(const char *input, const struct kernel_param *kp)
+{
+ return kstrtoull(input, 0, ¤t->bpf_prog_aux_priv);
+}
+
+static int prog_aux_priv_param_get(char *buf, const struct kernel_param *kp)
+{
+ return scnprintf(buf, PAGE_SIZE, "%llu\n", current->bpf_prog_aux_priv);
+}
+
+static const struct kernel_param_ops prog_aux_priv_param_ops = {
+ .set = prog_aux_priv_param_set,
+ .get = prog_aux_priv_param_get,
+};
+
+#undef MODULE_PARAM_PREFIX
+#define MODULE_PARAM_PREFIX "bpf."
+module_param_cb(prog_aux_priv, &prog_aux_priv_param_ops, NULL, 0664);
+MODULE_PARM_DESC("prog_aux_priv",
+ "Set prog->aux->priv to this value for all BPF programs loaded by %current");
+
/* last field in 'union bpf_attr' used by this command */
#define BPF_PROG_LOAD_LAST_FIELD fd_array_cnt
@@ -2898,6 +2919,8 @@ static int bpf_prog_load(union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size)
prog->expected_attach_type = attr->expected_attach_type;
prog->sleepable = !!(attr->prog_flags & BPF_F_SLEEPABLE);
+ /* XXX - See kernel/sched/ext.c::scx_sub_enable() */
+ prog->aux->priv_user = current->bpf_prog_aux_priv;
prog->aux->attach_btf = attach_btf;
prog->aux->attach_btf_id = attr->attach_btf_id;
prog->aux->dst_prog = dst_prog;
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 5eb1d6919595..a0251442b8ac 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4116,6 +4116,24 @@ static void scx_sub_disable(struct scx_sched *sch)
if (sch->ops.exit)
SCX_CALL_OP(sch, SCX_KF_UNLOCKED, exit, NULL, sch->exit_info);
+
+ /*
+ * XXX - NULL prog->aux->priv is interpreted as scx_root, so use an
+ * ERR_PTR value to mark the associated progs dead. Note that this is
+ * racy as e.g. a tracepoint program associated with a scheduler which
+ * hasn't finished scx_sub_enable() yet may end up affecting scx_root
+ * inadvertently. Plug the hole when this hack is replaced with a proper
+ * BPF construct.
+ */
+ u32 prog_id = 0;
+ struct bpf_prog *prog;
+ while ((prog = bpf_prog_get_curr_or_next(&prog_id))) {
+ if (prog->aux->priv == sch)
+ RCU_INIT_POINTER(prog->aux->priv, ERR_PTR(-ENODEV));
+ bpf_prog_put(prog);
+ prog_id++;
+ }
+
kobject_del(&sch->kobj);
}
#else /* CONFIG_EXT_SUB_SCHED */
@@ -5148,6 +5166,24 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
goto err_disable;
}
+ /*
+ * XXX - We want all BPF programs loaded together with this scheduler
+ * instance to point to this scheduler instance. BPF currently doesn't
+ * have such feature so work around with a hack. The loading userspace
+ * thread sets %current->bpf_prog_aux_priv to the associated cgroup ID
+ * which gets transferred to bpf->aux->priv_user in bpf_prog_load().
+ * Here, we can find all progs that have the matching cgroup ID and set
+ * their prog->aux->priv to $sch.
+ */
+ u32 prog_id = 0;
+ struct bpf_prog *prog;
+ while ((prog = bpf_prog_get_curr_or_next(&prog_id))) {
+ if (prog->aux->priv_user == cgroup_id(cgrp))
+ rcu_assign_pointer(prog->aux->priv, sch);
+ bpf_prog_put(prog);
+ prog_id++;
+ }
+
if (sch->ops.init) {
ret = SCX_CALL_OP_RET(sch, SCX_KF_UNLOCKED, init, NULL);
if (ret) {
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 5d762d10f4db..cefc439c9e4a 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -99,12 +99,25 @@ int main(int argc, char **argv)
break;
case 'c': {
struct stat st;
+ int fd, len;
+ char buf[19];
if (stat(optarg, &st) < 0) {
perror("stat");
return 1;
}
skel->struct_ops.qmap_ops->sub_cgroup_id = st.st_ino;
skel->rodata->sub_cgroup_id = st.st_ino;
+ fd = open("/sys/module/bpf/parameters/prog_aux_priv", O_RDWR);
+ if (fd < 0) {
+ perror("open(\"/sys/module/bpf/parameters/prog_aux_priv\")");
+ return 1;
+ }
+ len = snprintf(buf, sizeof(buf), "0x%lx", st.st_ino);
+ if (write(fd, buf, len) != len) {
+ perror("write(\"/sys/module/bpf/parameters/prog_aux_priv\")");
+ return 1;
+ }
+ close(fd);
break;
}
case 'd':
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 25/46] sched_ext: Introduce scx_task_sched[_rcu]()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (23 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 24/46] HACK_NOT_FOR_UPSTREAM: BPF: Implement prog grouping hack Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 26/46] sched_ext: Introduce scx_prog_sched() Tejun Heo
` (20 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation of multiple scheduler support, add p->scx.sched which points
to the scx_sched instance that the task is scheduled by, which is currently
always scx_root. Add scx_task_sched[_rcu]() accessors which return the
assocaited scx_sched of the specified task and replace the raw scx_root
dereferences with it where applicable.
As scx_root is still the only scheduler, this shouldn't introduce
user-visible behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/sched/ext.h | 7 ++++
kernel/sched/ext.c | 68 +++++++++++++++++++++++--------------
kernel/sched/ext_internal.h | 42 +++++++++++++++++++++++
3 files changed, 92 insertions(+), 25 deletions(-)
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 992c8b43db75..73f9df0759e2 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -144,6 +144,13 @@ struct scx_sched;
* for a task to be scheduled by SCX.
*/
struct sched_ext_entity {
+#ifdef CONFIG_CGROUPS
+ /*
+ * Associated scx_sched. Updated either during fork or while holding
+ * both p->pi_lock and rq lock.
+ */
+ struct scx_sched __rcu *sched;
+#endif
struct scx_dispatch_q *dsq;
struct scx_dsq_list_node dsq_list; /* dispatch order */
struct rb_node dsq_priq; /* p->scx.dsq_vtime order */
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index a0251442b8ac..e041c2f0cdc3 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -19,7 +19,7 @@ static DEFINE_RAW_SPINLOCK(scx_sched_lock);
* are used as temporary markers to indicate that the dereferences need to be
* updated to point to the associated scheduler instances rather than scx_root.
*/
-static struct scx_sched __rcu *scx_root;
+struct scx_sched __rcu *scx_root;
/*
* All scheds, writers must hold both scx_enable_mutex and scx_sched_lock.
@@ -213,6 +213,11 @@ static struct scx_sched *scx_parent(struct scx_sched *sch)
}
#ifdef CONFIG_EXT_SUB_SCHED
+static void scx_set_task_sched(struct task_struct *p, struct scx_sched *sch)
+{
+ rcu_assign_pointer(p->scx.sched, sch);
+}
+
static struct scx_sched *scx_next_descendant_pre(struct scx_sched *pos,
struct scx_sched *root)
{
@@ -240,6 +245,10 @@ static struct scx_sched *scx_next_descendant_pre(struct scx_sched *pos,
return NULL;
}
#else /* CONFIG_EXT_SUB_SCHED */
+static void scx_set_task_sched(struct task_struct *p, struct scx_sched *sch)
+{
+}
+
static struct scx_sched *scx_next_descendant_pre(struct scx_sched *pos,
struct scx_sched *root)
{
@@ -1355,7 +1364,7 @@ static bool scx_rq_online(struct rq *rq)
static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
int sticky_cpu)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
struct task_struct **ddsp_taskp;
unsigned long qseq;
@@ -1473,7 +1482,7 @@ static void clr_task_runnable(struct task_struct *p, bool reset_runnable_at)
static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
int sticky_cpu = p->scx.sticky_cpu;
if (enq_flags & ENQUEUE_WAKEUP)
@@ -1520,7 +1529,7 @@ static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int enq_flags
static void ops_dequeue(struct rq *rq, struct task_struct *p, u64 deq_flags)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
unsigned long opss;
/* dequeue is always temporary, don't reset runnable_at */
@@ -1569,7 +1578,7 @@ static void ops_dequeue(struct rq *rq, struct task_struct *p, u64 deq_flags)
static bool dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
if (!(p->scx.flags & SCX_TASK_QUEUED)) {
WARN_ON_ONCE(task_runnable(p));
@@ -1613,8 +1622,8 @@ static bool dequeue_task_scx(struct rq *rq, struct task_struct *p, int deq_flags
static void yield_task_scx(struct rq *rq)
{
- struct scx_sched *sch = scx_root;
struct task_struct *p = rq->curr;
+ struct scx_sched *sch = scx_task_sched(p);
if (SCX_HAS_OP(sch, yield))
SCX_CALL_OP_2TASKS_RET(sch, SCX_KF_REST, yield, rq, p, NULL);
@@ -1624,10 +1633,10 @@ static void yield_task_scx(struct rq *rq)
static bool yield_to_task_scx(struct rq *rq, struct task_struct *to)
{
- struct scx_sched *sch = scx_root;
struct task_struct *from = rq->curr;
+ struct scx_sched *sch = scx_task_sched(from);
- if (SCX_HAS_OP(sch, yield))
+ if (SCX_HAS_OP(sch, yield) && sch == scx_task_sched(to))
return SCX_CALL_OP_2TASKS_RET(sch, SCX_KF_REST, yield, rq,
from, to);
else
@@ -2329,7 +2338,7 @@ static void process_ddsp_deferred_locals(struct rq *rq)
*/
while ((p = list_first_entry_or_null(&rq->scx.ddsp_deferred_locals,
struct task_struct, scx.dsq_list.node))) {
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
struct scx_dispatch_q *dsq;
list_del_init(&p->scx.dsq_list.node);
@@ -2343,7 +2352,7 @@ static void process_ddsp_deferred_locals(struct rq *rq)
static void set_next_task_scx(struct rq *rq, struct task_struct *p, bool first)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
if (p->scx.flags & SCX_TASK_QUEUED) {
/*
@@ -2446,7 +2455,7 @@ static void switch_class(struct rq *rq, struct task_struct *next)
static void put_prev_task_scx(struct rq *rq, struct task_struct *p,
struct task_struct *next)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
update_curr_scx(rq);
/* see dequeue_task_scx() on why we skip when !QUEUED */
@@ -2540,7 +2549,7 @@ static struct task_struct *pick_task_scx(struct rq *rq)
if (keep_prev) {
p = prev;
if (!p->scx.slice)
- refill_task_slice_dfl(rcu_dereference_sched(scx_root), p);
+ refill_task_slice_dfl(scx_task_sched(p), p);
} else {
p = first_local_task(rq);
if (!p) {
@@ -2551,7 +2560,7 @@ static struct task_struct *pick_task_scx(struct rq *rq)
}
if (unlikely(!p->scx.slice)) {
- struct scx_sched *sch = rcu_dereference_sched(scx_root);
+ struct scx_sched *sch = scx_task_sched(p);
if (!scx_rq_bypassing(rq) && !sch->warned_zero_slice) {
printk_deferred(KERN_WARNING "sched_ext: %s[%d] has zero slice in %s()\n",
@@ -2607,7 +2616,7 @@ bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flags)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
bool rq_bypass;
/*
@@ -2668,7 +2677,7 @@ static void task_woken_scx(struct rq *rq, struct task_struct *p)
static void set_cpus_allowed_scx(struct task_struct *p,
struct affinity_context *ac)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
set_cpus_allowed_common(p, ac);
@@ -2809,7 +2818,7 @@ void scx_tick(struct rq *rq)
static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(curr);
update_curr_scx(rq);
@@ -2985,11 +2994,12 @@ static void scx_disable_task(struct task_struct *p)
static void scx_exit_task(struct task_struct *p)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
struct scx_exit_task_args args = {
.cancelled = false,
};
+ lockdep_assert_held(&p->pi_lock);
lockdep_assert_rq_held(task_rq(p));
switch (scx_get_task_state(p)) {
@@ -3011,6 +3021,7 @@ static void scx_exit_task(struct task_struct *p)
if (SCX_HAS_OP(sch, exit_task))
SCX_CALL_OP_TASK(sch, SCX_KF_REST, exit_task, task_rq(p),
p, &args);
+ scx_set_task_sched(p, NULL);
scx_set_task_state(p, SCX_TASK_NONE);
}
@@ -3040,12 +3051,18 @@ void scx_pre_fork(struct task_struct *p)
int scx_fork(struct task_struct *p, struct kernel_clone_args *kargs)
{
+ int ret;
+
percpu_rwsem_assert_held(&scx_fork_rwsem);
- if (scx_init_task_enabled)
- return scx_init_task(p, task_group(p), true);
- else
- return 0;
+ if (scx_init_task_enabled) {
+ ret = scx_init_task(p, task_group(p), true);
+ if (!ret)
+ scx_set_task_sched(p, scx_root);
+ return ret;
+ }
+
+ return 0;
}
void scx_post_fork(struct task_struct *p)
@@ -3122,7 +3139,7 @@ void sched_ext_free(struct task_struct *p)
static void reweight_task_scx(struct rq *rq, struct task_struct *p,
const struct load_weight *lw)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
lockdep_assert_rq_held(task_rq(p));
@@ -3138,7 +3155,7 @@ static void prio_changed_scx(struct rq *rq, struct task_struct *p, int oldprio)
static void switching_to_scx(struct rq *rq, struct task_struct *p)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_task_sched(p);
scx_enable_task(p);
@@ -3768,11 +3785,11 @@ bool scx_allow_ttwu_queue(const struct task_struct *p)
if (!scx_enabled())
return true;
- sch = rcu_dereference_sched(scx_root);
+ sch = scx_task_sched(p);
if (unlikely(!sch))
return true;
- if (scx_root->ops.flags & SCX_OPS_ALLOW_QUEUED_WAKEUP)
+ if (sch->ops.flags & SCX_OPS_ALLOW_QUEUED_WAKEUP)
return true;
if (unlikely(p->sched_class != &ext_sched_class))
@@ -5006,6 +5023,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
goto err_disable_unlock_all;
}
+ scx_set_task_sched(p, sch);
scx_set_task_state(p, SCX_TASK_READY);
put_task_struct(p);
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 2e3aa3888ce0..aaf606e5d0de 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -1109,6 +1109,7 @@ enum scx_ops_state {
#define SCX_OPSS_STATE_MASK ((1LU << SCX_OPSS_QSEQ_SHIFT) - 1)
#define SCX_OPSS_QSEQ_MASK (~SCX_OPSS_STATE_MASK)
+extern struct scx_sched __rcu *scx_root;
DECLARE_PER_CPU(struct rq *, scx_locked_rq_state);
/*
@@ -1129,3 +1130,44 @@ static inline bool scx_rq_bypassing(struct rq *rq)
{
return unlikely(rq->scx.flags & SCX_RQ_BYPASSING);
}
+
+#ifdef CONFIG_EXT_SUB_SCHED
+/**
+ * scx_task_sched - Find scx_sched scheduling a task
+ * @p: task of interest
+ *
+ * Return @p's scheduler instance. Must be called with @p's rq locked.
+ */
+static inline struct scx_sched *scx_task_sched(const struct task_struct *p)
+{
+ return rcu_dereference_protected(p->scx.sched,
+ lockdep_is_held(&p->pi_lock) ||
+ lockdep_is_held(__rq_lockp(task_rq(p))));
+}
+
+/**
+ * scx_task_sched_rcu - Find scx_sched scheduling a task
+ * @p: task of interest
+ *
+ * Return @p's scheduler instance. The returned scx_sched is RCU protected.
+ */
+static inline struct scx_sched *scx_task_sched_rcu(const struct task_struct *p)
+{
+ return rcu_dereference_check(p->scx.sched,
+ rcu_read_lock_bh_held() ||
+ rcu_read_lock_sched_held());
+}
+#else /* CONFIG_EXT_SUB_SCHED */
+static inline struct scx_sched *scx_task_sched(const struct task_struct *p)
+{
+ return rcu_dereference_protected(scx_root,
+ lockdep_is_held(__rq_lockp(task_rq(p))));
+}
+
+static inline struct scx_sched *scx_task_sched_rcu(const struct task_struct *p)
+{
+ return rcu_dereference_check(scx_root,
+ rcu_read_lock_bh_held() ||
+ rcu_read_lock_sched_held());
+}
+#endif /* CONFIG_EXT_SUB_SCHED */
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 26/46] sched_ext: Introduce scx_prog_sched()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (24 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 25/46] sched_ext: Introduce scx_task_sched[_rcu]() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 27/46] sched_ext: Ignore insertions of not-owned tasks into DSQs Tejun Heo
` (19 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation of multiple scheduler support, add add scx_prog_sched()
accessor which returns the assocaited scx_sched given the magic BPF
parameter @aux__prog on kfuncs.
As scx_root is still the only scheduler, this shouldn't introduce
user-visible behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 110 +++++++++++++++++++++++-------------
kernel/sched/ext_idle.c | 72 +++++++++++++++--------
kernel/sched/ext_internal.h | 33 +++++++++++
3 files changed, 154 insertions(+), 61 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index e041c2f0cdc3..b5f5106ddbf8 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -5902,6 +5902,7 @@ __bpf_kfunc_start_defs();
* @dsq_id: DSQ to insert into
* @slice: duration @p can run for in nsecs, 0 to keep the current value
* @enq_flags: SCX_ENQ_*
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Insert @p into the FIFO queue of the DSQ identified by @dsq_id. It is safe to
* call this function spuriously. Can be called from ops.enqueue(),
@@ -5932,12 +5933,13 @@ __bpf_kfunc_start_defs();
* scx_bpf_kick_cpu() to trigger scheduling.
*/
__bpf_kfunc void scx_bpf_dsq_insert(struct task_struct *p, u64 dsq_id, u64 slice,
- u64 enq_flags)
+ u64 enq_flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return;
@@ -5979,7 +5981,13 @@ __bpf_kfunc void scx_bpf_dsq_insert_vtime(struct task_struct *p, u64 dsq_id,
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ /*
+ * FIXME - This should be scx_prog_sched() so that we can verify that
+ * the calling scheduler has authority over @p. However, due to BPF
+ * funcation all parameter count limit, we can't add @aux__prog to this
+ * function. This can be fixed by packing the parameters into a struct.
+ */
+ sch = scx_task_sched(p);
if (unlikely(!sch))
return;
@@ -6099,16 +6107,17 @@ __bpf_kfunc_start_defs();
/**
* scx_bpf_dispatch_nr_slots - Return the number of remaining dispatch slots
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Can only be called from ops.dispatch().
*/
-__bpf_kfunc u32 scx_bpf_dispatch_nr_slots(void)
+__bpf_kfunc u32 scx_bpf_dispatch_nr_slots(const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return 0;
@@ -6120,18 +6129,19 @@ __bpf_kfunc u32 scx_bpf_dispatch_nr_slots(void)
/**
* scx_bpf_dispatch_cancel - Cancel the latest dispatch
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Cancel the latest dispatch. Can be called multiple times to cancel further
* dispatches. Can only be called from ops.dispatch().
*/
-__bpf_kfunc void scx_bpf_dispatch_cancel(void)
+__bpf_kfunc void scx_bpf_dispatch_cancel(const struct bpf_prog_aux *aux__prog)
{
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return;
@@ -6147,6 +6157,7 @@ __bpf_kfunc void scx_bpf_dispatch_cancel(void)
/**
* scx_bpf_dsq_move_to_local - move a task from a DSQ to the current CPU's local DSQ
* @dsq_id: DSQ to move task from
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Move a task from the non-local DSQ identified by @dsq_id to the current CPU's
* local DSQ for execution. Can only be called from ops.dispatch().
@@ -6158,7 +6169,8 @@ __bpf_kfunc void scx_bpf_dispatch_cancel(void)
* Returns %true if a task has been moved, %false if there isn't any task to
* move.
*/
-__bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id)
+__bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct scx_dispatch_q *dsq;
@@ -6166,7 +6178,7 @@ __bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id)
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return false;
@@ -6310,12 +6322,13 @@ __bpf_kfunc_start_defs();
/**
* scx_bpf_reenqueue_local - Re-enqueue tasks on a local DSQ
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Iterate over all of the tasks currently enqueued on the local DSQ of the
* caller's CPU, and re-enqueue them in the BPF scheduler. Returns the number of
* processed tasks. Can only be called from ops.cpu_release().
*/
-__bpf_kfunc u32 scx_bpf_reenqueue_local(void)
+__bpf_kfunc u32 scx_bpf_reenqueue_local(const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
LIST_HEAD(tasks);
@@ -6324,7 +6337,7 @@ __bpf_kfunc u32 scx_bpf_reenqueue_local(void)
struct task_struct *p, *n;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return 0;
@@ -6389,11 +6402,13 @@ __bpf_kfunc_start_defs();
* scx_bpf_create_dsq - Create a custom DSQ
* @dsq_id: DSQ to create
* @node: NUMA node to allocate from
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Create a custom DSQ identified by @dsq_id. Can be called from any sleepable
* scx callback, and any BPF_PROG_TYPE_SYSCALL prog.
*/
-__bpf_kfunc s32 scx_bpf_create_dsq(u64 dsq_id, s32 node)
+__bpf_kfunc s32 scx_bpf_create_dsq(u64 dsq_id, s32 node,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_dispatch_q *dsq;
struct scx_sched *sch;
@@ -6412,7 +6427,7 @@ __bpf_kfunc s32 scx_bpf_create_dsq(u64 dsq_id, s32 node)
rcu_read_lock();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (sch) {
init_dsq(dsq, dsq_id, sch);
ret = rhashtable_lookup_insert_fast(&sch->dsq_hash, &dsq->hash_node,
@@ -6501,18 +6516,20 @@ static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags)
* scx_bpf_kick_cpu - Trigger reschedule on a CPU
* @cpu: cpu to kick
* @flags: %SCX_KICK_* flags
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Kick @cpu into rescheduling. This can be used to wake up an idle CPU or
* trigger rescheduling on a busy CPU. This can be called from any online
* scx_ops operation and the actual kicking is performed asynchronously through
* an irq work.
*/
-__bpf_kfunc void scx_bpf_kick_cpu(s32 cpu, u64 flags)
+__bpf_kfunc void scx_bpf_kick_cpu(s32 cpu, u64 flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (likely(sch))
scx_kick_cpu(sch, cpu, flags);
}
@@ -6592,10 +6609,11 @@ __bpf_kfunc void scx_bpf_destroy_dsq(u64 dsq_id)
* tasks which are already queued when this function is invoked.
*/
__bpf_kfunc int bpf_iter_scx_dsq_new(struct bpf_iter_scx_dsq *it, u64 dsq_id,
- u64 flags)
+ u64 flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct bpf_iter_scx_dsq_kern *kit = (void *)it;
- struct scx_sched *sch;
+ struct scx_sched *sch = scx_prog_sched(aux__prog);
BUILD_BUG_ON(sizeof(struct bpf_iter_scx_dsq_kern) >
sizeof(struct bpf_iter_scx_dsq));
@@ -6608,7 +6626,6 @@ __bpf_kfunc int bpf_iter_scx_dsq_new(struct bpf_iter_scx_dsq *it, u64 dsq_id,
*/
kit->dsq = NULL;
- sch = rcu_dereference_check(scx_root, rcu_read_lock_bh_held());
if (unlikely(!sch))
return -ENODEV;
@@ -6749,18 +6766,20 @@ __bpf_kfunc_start_defs();
* @fmt: error message format string
* @data: format string parameters packaged using ___bpf_fill() macro
* @data__sz: @data len, must end in '__sz' for the verifier
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Indicate that the BPF scheduler wants to exit gracefully, and initiate ops
* disabling.
*/
__bpf_kfunc void scx_bpf_exit_bstr(s64 exit_code, char *fmt,
- unsigned long long *data, u32 data__sz)
+ unsigned long long *data, u32 data__sz,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
unsigned long flags;
raw_spin_lock_irqsave(&scx_exit_bstr_buf_lock, flags);
- sch = rcu_dereference_bh(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (likely(sch) &&
bstr_format(sch, &scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
scx_exit(sch, SCX_EXIT_UNREG_BPF, exit_code, "%s", scx_exit_bstr_buf.line);
@@ -6772,18 +6791,20 @@ __bpf_kfunc void scx_bpf_exit_bstr(s64 exit_code, char *fmt,
* @fmt: error message format string
* @data: format string parameters packaged using ___bpf_fill() macro
* @data__sz: @data len, must end in '__sz' for the verifier
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Indicate that the BPF scheduler encountered a fatal error and initiate ops
* disabling.
*/
__bpf_kfunc void scx_bpf_error_bstr(char *fmt, unsigned long long *data,
- u32 data__sz)
+ u32 data__sz,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
unsigned long flags;
raw_spin_lock_irqsave(&scx_exit_bstr_buf_lock, flags);
- sch = rcu_dereference_bh(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (likely(sch) &&
bstr_format(sch, &scx_exit_bstr_buf, fmt, data, data__sz) >= 0)
scx_exit(sch, SCX_EXIT_ERROR_BPF, 0, "%s", scx_exit_bstr_buf.line);
@@ -6795,6 +6816,7 @@ __bpf_kfunc void scx_bpf_error_bstr(char *fmt, unsigned long long *data,
* @fmt: format string
* @data: format string parameters packaged using ___bpf_fill() macro
* @data__sz: @data len, must end in '__sz' for the verifier
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* To be called through scx_bpf_dump() helper from ops.dump(), dump_cpu() and
* dump_task() to generate extra debug dump specific to the BPF scheduler.
@@ -6803,7 +6825,8 @@ __bpf_kfunc void scx_bpf_error_bstr(char *fmt, unsigned long long *data,
* multiple calls. The last line is automatically terminated.
*/
__bpf_kfunc void scx_bpf_dump_bstr(char *fmt, unsigned long long *data,
- u32 data__sz)
+ u32 data__sz,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
struct scx_dump_data *dd = &scx_dump_data;
@@ -6812,7 +6835,7 @@ __bpf_kfunc void scx_bpf_dump_bstr(char *fmt, unsigned long long *data,
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return;
@@ -6851,18 +6874,19 @@ __bpf_kfunc void scx_bpf_dump_bstr(char *fmt, unsigned long long *data,
/**
* scx_bpf_cpuperf_cap - Query the maximum relative capacity of a CPU
* @cpu: CPU of interest
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Return the maximum relative capacity of @cpu in relation to the most
* performant CPU in the system. The return value is in the range [1,
* %SCX_CPUPERF_ONE]. See scx_bpf_cpuperf_cur().
*/
-__bpf_kfunc u32 scx_bpf_cpuperf_cap(s32 cpu)
+__bpf_kfunc u32 scx_bpf_cpuperf_cap(s32 cpu, const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (likely(sch) && ops_cpu_valid(sch, cpu, NULL))
return arch_scale_cpu_capacity(cpu);
else
@@ -6872,6 +6896,7 @@ __bpf_kfunc u32 scx_bpf_cpuperf_cap(s32 cpu)
/**
* scx_bpf_cpuperf_cur - Query the current relative performance of a CPU
* @cpu: CPU of interest
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Return the current relative performance of @cpu in relation to its maximum.
* The return value is in the range [1, %SCX_CPUPERF_ONE].
@@ -6883,13 +6908,13 @@ __bpf_kfunc u32 scx_bpf_cpuperf_cap(s32 cpu)
*
* The result is in the range [1, %SCX_CPUPERF_ONE].
*/
-__bpf_kfunc u32 scx_bpf_cpuperf_cur(s32 cpu)
+__bpf_kfunc u32 scx_bpf_cpuperf_cur(s32 cpu, const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (likely(sch) && ops_cpu_valid(sch, cpu, NULL))
return arch_scale_freq_capacity(cpu);
else
@@ -6900,6 +6925,7 @@ __bpf_kfunc u32 scx_bpf_cpuperf_cur(s32 cpu)
* scx_bpf_cpuperf_set - Set the relative performance target of a CPU
* @cpu: CPU of interest
* @perf: target performance level [0, %SCX_CPUPERF_ONE]
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Set the target performance level of @cpu to @perf. @perf is in linear
* relative scale between 0 and %SCX_CPUPERF_ONE. This determines how the
@@ -6910,13 +6936,14 @@ __bpf_kfunc u32 scx_bpf_cpuperf_cur(s32 cpu)
* use. Consult hardware and cpufreq documentation for more information. The
* current performance level can be monitored using scx_bpf_cpuperf_cur().
*/
-__bpf_kfunc void scx_bpf_cpuperf_set(s32 cpu, u32 perf)
+__bpf_kfunc void scx_bpf_cpuperf_set(s32 cpu, u32 perf,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(sch);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return;
@@ -7026,14 +7053,16 @@ __bpf_kfunc s32 scx_bpf_task_cpu(const struct task_struct *p)
/**
* scx_bpf_cpu_rq - Fetch the rq of a CPU
* @cpu: CPU of the rq
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*/
-__bpf_kfunc struct rq *scx_bpf_cpu_rq(s32 cpu)
+__bpf_kfunc struct rq *scx_bpf_cpu_rq(s32 cpu,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return NULL;
@@ -7052,18 +7081,19 @@ __bpf_kfunc struct rq *scx_bpf_cpu_rq(s32 cpu)
/**
* scx_bpf_locked_rq - Return the rq currently locked by SCX
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Returns the rq if a rq lock is currently held by SCX.
* Otherwise emits an error and returns NULL.
*/
-__bpf_kfunc struct rq *scx_bpf_locked_rq(void)
+__bpf_kfunc struct rq *scx_bpf_locked_rq(const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
struct rq *rq;
guard(preempt)();
- sch = rcu_dereference_sched(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return NULL;
@@ -7079,16 +7109,18 @@ __bpf_kfunc struct rq *scx_bpf_locked_rq(void)
/**
* scx_bpf_cpu_curr - Return remote CPU's curr task
* @cpu: CPU of interest
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Callers must hold RCU read lock (KF_RCU).
*/
-__bpf_kfunc struct task_struct *scx_bpf_cpu_curr(s32 cpu)
+__bpf_kfunc struct task_struct *scx_bpf_cpu_curr(s32 cpu,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return NULL;
@@ -7101,6 +7133,7 @@ __bpf_kfunc struct task_struct *scx_bpf_cpu_curr(s32 cpu)
/**
* scx_bpf_task_cgroup - Return the sched cgroup of a task
* @p: task of interest
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* @p->sched_task_group->css.cgroup represents the cgroup @p is associated with
* from the scheduler's POV. SCX operations should use this function to
@@ -7110,7 +7143,8 @@ __bpf_kfunc struct task_struct *scx_bpf_cpu_curr(s32 cpu)
* operations. The restriction guarantees that @p's rq is locked by the caller.
*/
#ifdef CONFIG_CGROUP_SCHED
-__bpf_kfunc struct cgroup *scx_bpf_task_cgroup(struct task_struct *p)
+__bpf_kfunc struct cgroup *scx_bpf_task_cgroup(struct task_struct *p,
+ const struct bpf_prog_aux *aux__prog)
{
struct task_group *tg = p->sched_task_group;
struct cgroup *cgrp = &cgrp_dfl_root.cgrp;
@@ -7118,7 +7152,7 @@ __bpf_kfunc struct cgroup *scx_bpf_task_cgroup(struct task_struct *p)
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
goto out;
diff --git a/kernel/sched/ext_idle.c b/kernel/sched/ext_idle.c
index c57779f0ad57..ce9c2f345ec5 100644
--- a/kernel/sched/ext_idle.c
+++ b/kernel/sched/ext_idle.c
@@ -920,14 +920,15 @@ static s32 select_cpu_from_kfunc(struct scx_sched *sch, struct task_struct *p,
* scx_bpf_cpu_node - Return the NUMA node the given @cpu belongs to, or
* trigger an error if @cpu is invalid
* @cpu: target CPU
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*/
-__bpf_kfunc int scx_bpf_cpu_node(s32 cpu)
+__bpf_kfunc int scx_bpf_cpu_node(s32 cpu, const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch) || !ops_cpu_valid(sch, cpu, NULL))
return NUMA_NO_NODE;
return cpu_to_node(cpu);
@@ -939,6 +940,7 @@ __bpf_kfunc int scx_bpf_cpu_node(s32 cpu)
* @prev_cpu: CPU @p was on previously
* @wake_flags: %SCX_WAKE_* flags
* @is_idle: out parameter indicating whether the returned CPU is idle
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Can be called from ops.select_cpu(), ops.enqueue(), or from an unlocked
* context such as a BPF test_run() call, as long as built-in CPU selection
@@ -949,14 +951,15 @@ __bpf_kfunc int scx_bpf_cpu_node(s32 cpu)
* currently idle and thus a good candidate for direct dispatching.
*/
__bpf_kfunc s32 scx_bpf_select_cpu_dfl(struct task_struct *p, s32 prev_cpu,
- u64 wake_flags, bool *is_idle)
+ u64 wake_flags, bool *is_idle,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
s32 cpu;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return -ENODEV;
@@ -996,7 +999,12 @@ __bpf_kfunc s32 scx_bpf_select_cpu_and(struct task_struct *p, s32 prev_cpu, u64
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ /*
+ * FIXME - This should be scx_prog_sched(). However, due to BPF
+ * funcation all parameter count limit, we can't add @aux__prog to this
+ * function. This can be fixed by packing the parameters into a struct.
+ */
+ sch = scx_task_sched(p);
if (unlikely(!sch))
return -ENODEV;
@@ -1008,18 +1016,20 @@ __bpf_kfunc s32 scx_bpf_select_cpu_and(struct task_struct *p, s32 prev_cpu, u64
* scx_bpf_get_idle_cpumask_node - Get a referenced kptr to the
* idle-tracking per-CPU cpumask of a target NUMA node.
* @node: target NUMA node
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Returns an empty cpumask if idle tracking is not enabled, if @node is
* not valid, or running on a UP kernel. In this case the actual error will
* be reported to the BPF scheduler via scx_error().
*/
-__bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask_node(int node)
+__bpf_kfunc const struct cpumask *
+scx_bpf_get_idle_cpumask_node(int node, const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return cpu_none_mask;
@@ -1033,17 +1043,19 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask_node(int node)
/**
* scx_bpf_get_idle_cpumask - Get a referenced kptr to the idle-tracking
* per-CPU cpumask.
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Returns an empty mask if idle tracking is not enabled, or running on a
* UP kernel.
*/
-__bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask(void)
+__bpf_kfunc const struct cpumask *
+scx_bpf_get_idle_cpumask(const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return cpu_none_mask;
@@ -1063,18 +1075,20 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_cpumask(void)
* idle-tracking, per-physical-core cpumask of a target NUMA node. Can be
* used to determine if an entire physical core is free.
* @node: target NUMA node
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Returns an empty cpumask if idle tracking is not enabled, if @node is
* not valid, or running on a UP kernel. In this case the actual error will
* be reported to the BPF scheduler via scx_error().
*/
-__bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask_node(int node)
+__bpf_kfunc const struct cpumask *
+scx_bpf_get_idle_smtmask_node(int node, const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return cpu_none_mask;
@@ -1092,17 +1106,19 @@ __bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask_node(int node)
* scx_bpf_get_idle_smtmask - Get a referenced kptr to the idle-tracking,
* per-physical-core cpumask. Can be used to determine if an entire physical
* core is free.
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Returns an empty mask if idle tracking is not enabled, or running on a
* UP kernel.
*/
-__bpf_kfunc const struct cpumask *scx_bpf_get_idle_smtmask(void)
+__bpf_kfunc const struct cpumask *
+scx_bpf_get_idle_smtmask(const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return cpu_none_mask;
@@ -1138,6 +1154,7 @@ __bpf_kfunc void scx_bpf_put_idle_cpumask(const struct cpumask *idle_mask)
/**
* scx_bpf_test_and_clear_cpu_idle - Test and clear @cpu's idle state
* @cpu: cpu to test and clear idle for
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Returns %true if @cpu was idle and its idle state was successfully cleared.
* %false otherwise.
@@ -1145,13 +1162,14 @@ __bpf_kfunc void scx_bpf_put_idle_cpumask(const struct cpumask *idle_mask)
* Unavailable if ops.update_idle() is implemented and
* %SCX_OPS_KEEP_BUILTIN_IDLE is not set.
*/
-__bpf_kfunc bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
+__bpf_kfunc bool
+scx_bpf_test_and_clear_cpu_idle(s32 cpu, const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return false;
@@ -1169,6 +1187,7 @@ __bpf_kfunc bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
* @cpus_allowed: Allowed cpumask
* @node: target NUMA node
* @flags: %SCX_PICK_IDLE_* flags
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Pick and claim an idle cpu in @cpus_allowed from the NUMA node @node.
*
@@ -1184,13 +1203,14 @@ __bpf_kfunc bool scx_bpf_test_and_clear_cpu_idle(s32 cpu)
* %SCX_OPS_BUILTIN_IDLE_PER_NODE is not set.
*/
__bpf_kfunc s32 scx_bpf_pick_idle_cpu_node(const struct cpumask *cpus_allowed,
- int node, u64 flags)
+ int node, u64 flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return -ENODEV;
@@ -1205,6 +1225,7 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu_node(const struct cpumask *cpus_allowed,
* scx_bpf_pick_idle_cpu - Pick and claim an idle cpu
* @cpus_allowed: Allowed cpumask
* @flags: %SCX_PICK_IDLE_CPU_* flags
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Pick and claim an idle cpu in @cpus_allowed. Returns the picked idle cpu
* number on success. -%EBUSY if no matching cpu was found.
@@ -1224,13 +1245,14 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu_node(const struct cpumask *cpus_allowed,
* scx_bpf_pick_idle_cpu_node() instead.
*/
__bpf_kfunc s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed,
- u64 flags)
+ u64 flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return -ENODEV;
@@ -1251,6 +1273,7 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed,
* @cpus_allowed: Allowed cpumask
* @node: target NUMA node
* @flags: %SCX_PICK_IDLE_CPU_* flags
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Pick and claim an idle cpu in @cpus_allowed. If none is available, pick any
* CPU in @cpus_allowed. Guaranteed to succeed and returns the picked idle cpu
@@ -1267,14 +1290,15 @@ __bpf_kfunc s32 scx_bpf_pick_idle_cpu(const struct cpumask *cpus_allowed,
* CPU.
*/
__bpf_kfunc s32 scx_bpf_pick_any_cpu_node(const struct cpumask *cpus_allowed,
- int node, u64 flags)
+ int node, u64 flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
s32 cpu;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return -ENODEV;
@@ -1300,6 +1324,7 @@ __bpf_kfunc s32 scx_bpf_pick_any_cpu_node(const struct cpumask *cpus_allowed,
* scx_bpf_pick_any_cpu - Pick and claim an idle cpu if available or pick any CPU
* @cpus_allowed: Allowed cpumask
* @flags: %SCX_PICK_IDLE_CPU_* flags
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
*
* Pick and claim an idle cpu in @cpus_allowed. If none is available, pick any
* CPU in @cpus_allowed. Guaranteed to succeed and returns the picked idle cpu
@@ -1314,14 +1339,15 @@ __bpf_kfunc s32 scx_bpf_pick_any_cpu_node(const struct cpumask *cpus_allowed,
* scx_bpf_pick_any_cpu_node() instead.
*/
__bpf_kfunc s32 scx_bpf_pick_any_cpu(const struct cpumask *cpus_allowed,
- u64 flags)
+ u64 flags,
+ const struct bpf_prog_aux *aux__prog)
{
struct scx_sched *sch;
s32 cpu;
guard(rcu)();
- sch = rcu_dereference(scx_root);
+ sch = scx_prog_sched(aux__prog);
if (unlikely(!sch))
return -ENODEV;
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index aaf606e5d0de..64d0f0787c8e 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -1157,6 +1157,32 @@ static inline struct scx_sched *scx_task_sched_rcu(const struct task_struct *p)
rcu_read_lock_bh_held() ||
rcu_read_lock_sched_held());
}
+
+/**
+ * scx_prog_sched - Find scx_sched associated with a BPF prog
+ * @aux: aux__prog passed in from BPF to a kfunc
+ *
+ * To be called from kfuncs. Return the scheduler instance associated with the
+ * BPF program given the special kfunc argument aux__prog. The returned
+ * scx_sched is RCU protected.
+ */
+static struct scx_sched *scx_prog_sched(const struct bpf_prog_aux *aux)
+{
+ struct scx_sched *sch;
+
+ sch = rcu_dereference_check(aux->priv,
+ rcu_read_lock_bh_held() ||
+ rcu_read_lock_sched_held());
+ if (unlikely(IS_ERR(sch)))
+ return NULL;
+
+ if (sch)
+ return sch;
+
+ return rcu_dereference_check(scx_root,
+ rcu_read_lock_bh_held() ||
+ rcu_read_lock_sched_held());
+}
#else /* CONFIG_EXT_SUB_SCHED */
static inline struct scx_sched *scx_task_sched(const struct task_struct *p)
{
@@ -1170,4 +1196,11 @@ static inline struct scx_sched *scx_task_sched_rcu(const struct task_struct *p)
rcu_read_lock_bh_held() ||
rcu_read_lock_sched_held());
}
+
+static struct scx_sched *scx_prog_sched(const struct bpf_prog_aux *aux)
+{
+ return rcu_dereference_check(scx_root,
+ rcu_read_lock_bh_held() ||
+ rcu_read_lock_sched_held());
+}
#endif /* CONFIG_EXT_SUB_SCHED */
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 27/46] sched_ext: Ignore insertions of not-owned tasks into DSQs
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (25 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 26/46] sched_ext: Introduce scx_prog_sched() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 28/46] sched_ext: scx_dsq_move() should validate the task belongs to the right scheduler Tejun Heo
` (18 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
As BPF schedulers are allowed to ignore dequeues, after a sub-sched
enabling moved tasks out of the parent scheduler, the parent scheduler may
try to dispatch an already moved task. As this doesn't necessarily
indicate a malfunction, ignore and count such attempts.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 9 +++++++++
kernel/sched/ext_internal.h | 12 ++++++++++++
2 files changed, 21 insertions(+)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index b5f5106ddbf8..891b956a92b6 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2125,6 +2125,12 @@ static void finish_dispatch(struct scx_sched *sch, struct rq *rq,
if ((opss & SCX_OPSS_QSEQ_MASK) != qseq_at_dispatch)
return;
+ /* see SCX_EV_INSERT_NOT_OWNED definition */
+ if (unlikely(sch != rcu_access_pointer(p->scx.sched))) {
+ __scx_add_event(sch, SCX_EV_INSERT_NOT_OWNED, 1);
+ return;
+ }
+
/*
* While we know @p is accessible, we don't yet have a claim on
* it - the BPF scheduler is allowed to dispatch tasks
@@ -3739,6 +3745,7 @@ static ssize_t scx_attr_events_show(struct kobject *kobj,
at += scx_attr_event_show(buf, at, &events, SCX_EV_BYPASS_DURATION);
at += scx_attr_event_show(buf, at, &events, SCX_EV_BYPASS_DISPATCH);
at += scx_attr_event_show(buf, at, &events, SCX_EV_BYPASS_ACTIVATE);
+ at += scx_attr_event_show(buf, at, &events, SCX_EV_INSERT_NOT_OWNED);
return at;
}
SCX_ATTR(events);
@@ -4631,6 +4638,7 @@ static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
scx_dump_event(s, &events, SCX_EV_BYPASS_DURATION);
scx_dump_event(s, &events, SCX_EV_BYPASS_DISPATCH);
scx_dump_event(s, &events, SCX_EV_BYPASS_ACTIVATE);
+ scx_dump_event(s, &events, SCX_EV_INSERT_NOT_OWNED);
if (seq_buf_has_overflowed(&s) && dump_len >= sizeof(trunc_marker))
memcpy(ei->dump + dump_len - sizeof(trunc_marker),
@@ -7247,6 +7255,7 @@ static void scx_read_events(struct scx_sched *sch, struct scx_event_stats *event
scx_agg_event(events, e_cpu, SCX_EV_BYPASS_DURATION);
scx_agg_event(events, e_cpu, SCX_EV_BYPASS_DISPATCH);
scx_agg_event(events, e_cpu, SCX_EV_BYPASS_ACTIVATE);
+ scx_agg_event(events, e_cpu, SCX_EV_INSERT_NOT_OWNED);
}
}
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 64d0f0787c8e..154993921c38 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -895,6 +895,18 @@ struct scx_event_stats {
* The number of times the bypassing mode has been activated.
*/
s64 SCX_EV_BYPASS_ACTIVATE;
+
+ /*
+ * The number of times the scheduler attempted to insert a task that it
+ * doesn't own into a DSQ. Such attempts are ignored.
+ *
+ * As BPF schedulers are allowed to ignore dequeues, it's difficult to
+ * tell whether such an attempt is from a scheduler malfunction or an
+ * ignored dequeue around sub-sched enabling. If this count keeps going
+ * up regardless of sub-sched enabling, it likely indicates a bug in the
+ * scheduler.
+ */
+ s64 SCX_EV_INSERT_NOT_OWNED;
};
struct scx_sched_pcpu {
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 28/46] sched_ext: scx_dsq_move() should validate the task belongs to the right scheduler
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (26 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 27/46] sched_ext: Ignore insertions of not-owned tasks into DSQs Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 29/46] sched_ext: Refactor task init/exit helpers Tejun Heo
` (17 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
scx_bpf_dsq_move[_vtime]() calls scx_dsq_move() to move task from a DSQ to
another. However, @p doesn't necessarily have to come form the containing
iteration and can thus be a task which belongs to another scx_sched. Verify
that @p is on the same scx_sched as the DSQ being iterated.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 891b956a92b6..c60b58341d24 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -6027,8 +6027,8 @@ static const struct btf_kfunc_id_set scx_kfunc_set_enqueue_dispatch = {
static bool scx_dsq_move(struct bpf_iter_scx_dsq_kern *kit,
struct task_struct *p, u64 dsq_id, u64 enq_flags)
{
- struct scx_sched *sch = scx_root;
struct scx_dispatch_q *src_dsq = kit->dsq, *dst_dsq;
+ struct scx_sched *sch = src_dsq->sched;
struct rq *this_rq, *src_rq, *locked_rq;
bool dispatched = false;
bool in_balance;
@@ -6038,6 +6038,11 @@ static bool scx_dsq_move(struct bpf_iter_scx_dsq_kern *kit,
!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return false;
+ if (unlikely(sch != scx_task_sched(p))) {
+ scx_error(sch, "scx_bpf_dsq_move[_vtime]() on %s[%d] but the task belongs to a different scheduler",
+ p->comm, p->pid);
+ }
+
/*
* Can be called from either ops.dispatch() locking this_rq() or any
* context where no rq lock is held. If latter, lock @p's task_rq which
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 29/46] sched_ext: Refactor task init/exit helpers
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (27 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 28/46] sched_ext: scx_dsq_move() should validate the task belongs to the right scheduler Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 30/46] sched_ext: Make scx_prio_less() handle multiple schedulers Tejun Heo
` (16 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
- Add the @sch parameter to scx_init_task() and drop @tg as it can be
obtained from @p. Separate out __scx_init_task() which does everything
except for the task state transition.
- Add the @sch parameter to scx_enable_task(). Separate out
__scx_enable_task() which does everything except for the task state
transition.
- Add the @sch parameter to scx_disable_task().
- Rename scx_exit_task() to scx_disable_and_exit_task() and separate out
__scx_disable_and_exit_task() which does everything except for the task
state transition.
While some task state transitions are relocated, no meaningful behavior
changes are expected.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 60 ++++++++++++++++++++++++++++++----------------
1 file changed, 39 insertions(+), 21 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index c60b58341d24..5b02903ba3bb 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2901,9 +2901,9 @@ static void scx_set_task_state(struct task_struct *p, enum scx_task_state state)
p->scx.flags |= state << SCX_TASK_STATE_SHIFT;
}
-static int scx_init_task(struct task_struct *p, struct task_group *tg, bool fork)
+static int __scx_init_task(struct scx_sched *sch, struct task_struct *p, bool fork)
{
- struct scx_sched *sch = scx_root;
+ struct task_group *tg = task_group(p);
int ret;
p->scx.disallow = false;
@@ -2922,8 +2922,6 @@ static int scx_init_task(struct task_struct *p, struct task_group *tg, bool fork
}
}
- scx_set_task_state(p, SCX_TASK_INIT);
-
if (p->scx.disallow) {
if (unlikely(scx_parent(sch))) {
scx_error(sch, "non-root ops.init_task() set task->scx.disallow for %s[%d]",
@@ -2957,9 +2955,18 @@ static int scx_init_task(struct task_struct *p, struct task_group *tg, bool fork
return 0;
}
-static void scx_enable_task(struct task_struct *p)
+static int scx_init_task(struct scx_sched *sch, struct task_struct *p, bool fork)
+{
+ int ret;
+
+ ret = __scx_init_task(sch, p, fork);
+ if (!ret)
+ scx_set_task_state(p, SCX_TASK_INIT);
+ return ret;
+}
+
+static void __scx_enable_task(struct scx_sched *sch, struct task_struct *p)
{
- struct scx_sched *sch = scx_root;
struct rq *rq = task_rq(p);
u32 weight;
@@ -2978,16 +2985,20 @@ static void scx_enable_task(struct task_struct *p)
if (SCX_HAS_OP(sch, enable))
SCX_CALL_OP_TASK(sch, SCX_KF_REST, enable, rq, p);
- scx_set_task_state(p, SCX_TASK_ENABLED);
if (SCX_HAS_OP(sch, set_weight))
SCX_CALL_OP_TASK(sch, SCX_KF_REST, set_weight, rq,
p, p->scx.weight);
}
-static void scx_disable_task(struct task_struct *p)
+static void scx_enable_task(struct scx_sched *sch, struct task_struct *p)
+{
+ __scx_enable_task(sch, p);
+ scx_set_task_state(p, SCX_TASK_ENABLED);
+}
+
+static void scx_disable_task(struct scx_sched *sch, struct task_struct *p)
{
- struct scx_sched *sch = scx_root;
struct rq *rq = task_rq(p);
lockdep_assert_rq_held(rq);
@@ -2998,9 +3009,9 @@ static void scx_disable_task(struct task_struct *p)
scx_set_task_state(p, SCX_TASK_READY);
}
-static void scx_exit_task(struct task_struct *p)
+static void __scx_disable_and_exit_task(struct scx_sched *sch,
+ struct task_struct *p)
{
- struct scx_sched *sch = scx_task_sched(p);
struct scx_exit_task_args args = {
.cancelled = false,
};
@@ -3017,7 +3028,7 @@ static void scx_exit_task(struct task_struct *p)
case SCX_TASK_READY:
break;
case SCX_TASK_ENABLED:
- scx_disable_task(p);
+ scx_disable_task(sch, p);
break;
default:
WARN_ON_ONCE(true);
@@ -3027,6 +3038,13 @@ static void scx_exit_task(struct task_struct *p)
if (SCX_HAS_OP(sch, exit_task))
SCX_CALL_OP_TASK(sch, SCX_KF_REST, exit_task, task_rq(p),
p, &args);
+}
+
+static void scx_disable_and_exit_task(struct scx_sched *sch,
+ struct task_struct *p)
+{
+ __scx_disable_and_exit_task(sch, p);
+
scx_set_task_sched(p, NULL);
scx_set_task_state(p, SCX_TASK_NONE);
}
@@ -3062,7 +3080,7 @@ int scx_fork(struct task_struct *p, struct kernel_clone_args *kargs)
percpu_rwsem_assert_held(&scx_fork_rwsem);
if (scx_init_task_enabled) {
- ret = scx_init_task(p, task_group(p), true);
+ ret = scx_init_task(scx_root, p, true);
if (!ret)
scx_set_task_sched(p, scx_root);
return ret;
@@ -3086,7 +3104,7 @@ void scx_post_fork(struct task_struct *p)
struct rq *rq;
rq = task_rq_lock(p, &rf);
- scx_enable_task(p);
+ scx_enable_task(scx_task_sched(p), p);
task_rq_unlock(rq, p, &rf);
}
}
@@ -3106,7 +3124,7 @@ void scx_cancel_fork(struct task_struct *p)
rq = task_rq_lock(p, &rf);
WARN_ON_ONCE(scx_get_task_state(p) >= SCX_TASK_READY);
- scx_exit_task(p);
+ scx_disable_and_exit_task(scx_task_sched(p), p);
task_rq_unlock(rq, p, &rf);
}
@@ -3137,7 +3155,7 @@ void sched_ext_free(struct task_struct *p)
struct rq *rq;
rq = task_rq_lock(p, &rf);
- scx_exit_task(p);
+ scx_disable_and_exit_task(scx_task_sched(p), p);
task_rq_unlock(rq, p, &rf);
}
}
@@ -3163,7 +3181,7 @@ static void switching_to_scx(struct rq *rq, struct task_struct *p)
{
struct scx_sched *sch = scx_task_sched(p);
- scx_enable_task(p);
+ scx_enable_task(sch, p);
/*
* set_cpus_allowed_scx() is not called while @p is associated with a
@@ -3176,7 +3194,7 @@ static void switching_to_scx(struct rq *rq, struct task_struct *p)
static void switched_from_scx(struct rq *rq, struct task_struct *p)
{
- scx_disable_task(p);
+ scx_disable_task(scx_task_sched(p), p);
}
static void wakeup_preempt_scx(struct rq *rq, struct task_struct *p,int wake_flags) {}
@@ -4201,7 +4219,7 @@ static void scx_root_disable(struct scx_sched *sch)
/*
* Shut down cgroup support before tasks so that the cgroup attach path
- * doesn't race against scx_exit_task().
+ * doesn't race against scx_disable_and_exit_task().
*/
scx_cgroup_lock();
scx_cgroup_exit(sch);
@@ -4234,7 +4252,7 @@ static void scx_root_disable(struct scx_sched *sch)
sched_enq_and_set_task(&ctx);
check_class_changed(task_rq(p), p, old_class, p->prio);
- scx_exit_task(p);
+ scx_disable_and_exit_task(scx_task_sched(p), p);
}
scx_task_iter_stop(&sti);
@@ -5022,7 +5040,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
scx_task_iter_unlock(&sti);
- ret = scx_init_task(p, task_group(p), false);
+ ret = scx_init_task(sch, p, false);
if (ret) {
put_task_struct(p);
scx_task_iter_stop(&sti);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 30/46] sched_ext: Make scx_prio_less() handle multiple schedulers
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (28 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 29/46] sched_ext: Refactor task init/exit helpers Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 31/46] sched_ext: Move bypass_depth into scx_sched Tejun Heo
` (15 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Call ops.core_sched_before() iff both tasks belong to the same scx_sched.
Otherwise, use timestamp based ordering.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 5b02903ba3bb..2b8a1a950c74 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2602,16 +2602,17 @@ static struct task_struct *pick_task_scx(struct rq *rq)
bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
bool in_fi)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch_a = scx_task_sched(a);
+ struct scx_sched *sch_b = scx_task_sched(b);
/*
* The const qualifiers are dropped from task_struct pointers when
* calling ops.core_sched_before(). Accesses are controlled by the
* verifier.
*/
- if (SCX_HAS_OP(sch, core_sched_before) &&
+ if (sch_a == sch_b && SCX_HAS_OP(sch_a, core_sched_before) &&
!scx_rq_bypassing(task_rq(a)))
- return SCX_CALL_OP_2TASKS_RET(sch, SCX_KF_REST, core_sched_before,
+ return SCX_CALL_OP_2TASKS_RET(sch_a, SCX_KF_REST, core_sched_before,
NULL,
(struct task_struct *)a,
(struct task_struct *)b);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 31/46] sched_ext: Move bypass_depth into scx_sched
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (29 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 30/46] sched_ext: Make scx_prio_less() handle multiple schedulers Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 32/46] sched_ext: Make bypass mode sub-sched aware Tejun Heo
` (14 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation of multiple scheduler support, make bypass state
per-scx_sched:
- Replace global scx_bypass_depth with scx_sched_pcpu->bypass_depth.
- SCX_RQ_BYPASSING is replaced with SCX_SCHED_PCPU_BYPASSING which is stored
in newly added scx_sched_pcpu->flags.
- Add @sch to scx_bypass().
- Rename scx_rq_bypassing() to scx_bypassing() and replace @rq with @cpu as
the state is now in scx_sched_pcpu.
No behavior changes expected.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 100 ++++++++++++++++++++----------------
kernel/sched/ext_idle.c | 3 +-
kernel/sched/ext_internal.h | 12 ++++-
kernel/sched/sched.h | 1 -
4 files changed, 67 insertions(+), 49 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 2b8a1a950c74..5f22a79e19ec 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -51,7 +51,6 @@ DEFINE_STATIC_PERCPU_RWSEM(scx_fork_rwsem);
static atomic_t scx_enable_state_var = ATOMIC_INIT(SCX_DISABLED);
static unsigned long scx_in_softlockup;
static atomic_t scx_breather_depth = ATOMIC_INIT(0);
-static int scx_bypass_depth;
static bool scx_init_task_enabled;
static bool scx_switching_all;
DEFINE_STATIC_KEY_FALSE(__scx_switched_all);
@@ -1382,7 +1381,7 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
if (!scx_rq_online(rq))
goto local;
- if (scx_rq_bypassing(rq)) {
+ if (scx_bypassing(sch, cpu_of(rq))) {
__scx_add_event(sch, SCX_EV_BYPASS_DISPATCH, 1);
goto global;
}
@@ -2218,7 +2217,8 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
* See scx_disable_workfn() for the explanation on the bypassing
* test.
*/
- if (prev_on_rq && prev->scx.slice && !scx_rq_bypassing(rq)) {
+ if (prev_on_rq && prev->scx.slice &&
+ !scx_bypassing(sch, cpu_of(rq))) {
rq->scx.flags |= SCX_RQ_BAL_KEEP;
goto has_tasks;
}
@@ -2232,7 +2232,7 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
goto has_tasks;
if (unlikely(!SCX_HAS_OP(sch, dispatch)) ||
- scx_rq_bypassing(rq) || !scx_rq_online(rq))
+ scx_bypassing(sch, cpu_of(rq)) || !scx_rq_online(rq))
goto no_tasks;
dspc->rq = rq;
@@ -2282,7 +2282,8 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
* %SCX_OPS_ENQ_LAST is in effect.
*/
if (prev_on_rq &&
- (!(sch->ops.flags & SCX_OPS_ENQ_LAST) || scx_rq_bypassing(rq))) {
+ (!(sch->ops.flags & SCX_OPS_ENQ_LAST) ||
+ scx_bypassing(sch, cpu_of(rq)))) {
rq->scx.flags |= SCX_RQ_BAL_KEEP;
__scx_add_event(sch, SCX_EV_DISPATCH_KEEP_LAST, 1);
goto has_tasks;
@@ -2477,7 +2478,7 @@ static void put_prev_task_scx(struct rq *rq, struct task_struct *p,
* forcing a different task. Leave it at the head of the local
* DSQ.
*/
- if (p->scx.slice && !scx_rq_bypassing(rq)) {
+ if (p->scx.slice && !scx_bypassing(sch, cpu_of(rq))) {
dispatch_enqueue(sch, &rq->scx.local_dsq, p,
SCX_ENQ_HEAD);
goto switch_class;
@@ -2568,7 +2569,8 @@ static struct task_struct *pick_task_scx(struct rq *rq)
if (unlikely(!p->scx.slice)) {
struct scx_sched *sch = scx_task_sched(p);
- if (!scx_rq_bypassing(rq) && !sch->warned_zero_slice) {
+ if (!scx_bypassing(sch, cpu_of(rq)) &&
+ !sch->warned_zero_slice) {
printk_deferred(KERN_WARNING "sched_ext: %s[%d] has zero slice in %s()\n",
p->comm, p->pid, __func__);
sch->warned_zero_slice = true;
@@ -2611,7 +2613,7 @@ bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
* verifier.
*/
if (sch_a == sch_b && SCX_HAS_OP(sch_a, core_sched_before) &&
- !scx_rq_bypassing(task_rq(a)))
+ !scx_bypassing(sch_a, task_cpu(a)))
return SCX_CALL_OP_2TASKS_RET(sch_a, SCX_KF_REST, core_sched_before,
NULL,
(struct task_struct *)a,
@@ -2624,7 +2626,7 @@ bool scx_prio_less(const struct task_struct *a, const struct task_struct *b,
static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flags)
{
struct scx_sched *sch = scx_task_sched(p);
- bool rq_bypass;
+ bool bypassing;
/*
* sched_exec() calls with %WF_EXEC when @p is about to exec(2) as it
@@ -2639,8 +2641,8 @@ static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flag
if (unlikely(wake_flags & WF_EXEC))
return prev_cpu;
- rq_bypass = scx_rq_bypassing(task_rq(p));
- if (likely(SCX_HAS_OP(sch, select_cpu)) && !rq_bypass) {
+ bypassing = scx_bypassing(sch, task_cpu(p));
+ if (likely(SCX_HAS_OP(sch, select_cpu)) && !bypassing) {
s32 cpu;
struct task_struct **ddsp_taskp;
@@ -2670,7 +2672,7 @@ static int select_task_rq_scx(struct task_struct *p, int prev_cpu, int wake_flag
}
p->scx.selected_cpu = cpu;
- if (rq_bypass)
+ if (bypassing)
__scx_add_event(sch, SCX_EV_BYPASS_DISPATCH, 1);
return cpu;
}
@@ -2833,7 +2835,7 @@ static void task_tick_scx(struct rq *rq, struct task_struct *curr, int queued)
* While disabling, always resched and refresh core-sched timestamp as
* we can't trust the slice management or ops.core_sched_before().
*/
- if (scx_rq_bypassing(rq)) {
+ if (scx_bypassing(sch, cpu_of(rq))) {
curr->scx.slice = 0;
touch_core_sched(rq, curr);
} else if (SCX_HAS_OP(sch, tick)) {
@@ -3217,13 +3219,14 @@ int scx_check_setscheduler(struct task_struct *p, int policy)
bool scx_can_stop_tick(struct rq *rq)
{
struct task_struct *p = rq->curr;
-
- if (scx_rq_bypassing(rq))
- return false;
+ struct scx_sched *sch = scx_task_sched(p);
if (p->sched_class != &ext_sched_class)
return true;
+ if (scx_bypassing(sch, cpu_of(rq)))
+ return false;
+
/*
* @rq can dispatch from different DSQs, so we can't tell whether it
* needs the tick or not by looking at nr_running. Allow stopping ticks
@@ -3942,40 +3945,37 @@ static void scx_clear_softlockup(void)
*
* - scx_prio_less() reverts to the default core_sched_at order.
*/
-static void scx_bypass(bool bypass)
+static void scx_bypass(struct scx_sched *sch, bool bypass)
{
static DEFINE_RAW_SPINLOCK(bypass_lock);
static unsigned long bypass_timestamp;
- struct scx_sched *sch;
unsigned long flags;
int cpu;
raw_spin_lock_irqsave(&bypass_lock, flags);
- sch = rcu_dereference_bh(scx_root);
if (bypass) {
- scx_bypass_depth++;
- WARN_ON_ONCE(scx_bypass_depth <= 0);
- if (scx_bypass_depth != 1)
+ sch->bypass_depth++;
+ WARN_ON_ONCE(sch->bypass_depth <= 0);
+ if (sch->bypass_depth != 1)
goto unlock;
bypass_timestamp = ktime_get_ns();
- if (sch)
- scx_add_event(sch, SCX_EV_BYPASS_ACTIVATE, 1);
+ scx_add_event(sch, SCX_EV_BYPASS_ACTIVATE, 1);
} else {
- scx_bypass_depth--;
- WARN_ON_ONCE(scx_bypass_depth < 0);
- if (scx_bypass_depth != 0)
+ sch->bypass_depth--;
+ WARN_ON_ONCE(sch->bypass_depth < 0);
+ if (sch->bypass_depth != 0)
goto unlock;
- if (sch)
- scx_add_event(sch, SCX_EV_BYPASS_DURATION,
- ktime_get_ns() - bypass_timestamp);
+ scx_add_event(sch, SCX_EV_BYPASS_DURATION,
+ ktime_get_ns() - bypass_timestamp);
}
- atomic_inc(&scx_breather_depth);
+ if (!scx_parent(sch))
+ atomic_inc(&scx_breather_depth);
/*
* No task property is changing. We just need to make sure all currently
- * queued tasks are re-queued according to the new scx_rq_bypassing()
+ * queued tasks are re-queued according to the new scx_bypassing()
* state. As an optimization, walk each rq's runnable_list instead of
* the scx_tasks list.
*
@@ -3984,22 +3984,23 @@ static void scx_bypass(bool bypass)
*/
for_each_possible_cpu(cpu) {
struct rq *rq = cpu_rq(cpu);
+ struct scx_sched_pcpu *pcpu = per_cpu_ptr(sch->pcpu, cpu);
struct task_struct *p, *n;
raw_spin_rq_lock(rq);
if (bypass) {
- WARN_ON_ONCE(rq->scx.flags & SCX_RQ_BYPASSING);
- rq->scx.flags |= SCX_RQ_BYPASSING;
+ WARN_ON_ONCE(pcpu->flags & SCX_SCHED_PCPU_BYPASSING);
+ pcpu->flags |= SCX_SCHED_PCPU_BYPASSING;
} else {
- WARN_ON_ONCE(!(rq->scx.flags & SCX_RQ_BYPASSING));
- rq->scx.flags &= ~SCX_RQ_BYPASSING;
+ WARN_ON_ONCE(!(pcpu->flags & SCX_SCHED_PCPU_BYPASSING));
+ pcpu->flags &= ~SCX_SCHED_PCPU_BYPASSING;
}
/*
* We need to guarantee that no tasks are on the BPF scheduler
* while bypassing. Either we see enabled or the enable path
- * sees scx_rq_bypassing() before moving tasks to SCX.
+ * sees scx_bypassing() before moving tasks to SCX.
*/
if (!scx_enabled()) {
raw_spin_rq_unlock(rq);
@@ -4029,7 +4030,8 @@ static void scx_bypass(bool bypass)
raw_spin_rq_unlock(rq);
}
- atomic_dec(&scx_breather_depth);
+ if (!scx_parent(sch))
+ atomic_dec(&scx_breather_depth);
unlock:
raw_spin_unlock_irqrestore(&bypass_lock, flags);
scx_clear_softlockup();
@@ -4192,7 +4194,7 @@ static void scx_root_disable(struct scx_sched *sch)
int cpu;
/* guarantee forward progress and disable all descendants */
- scx_bypass(true);
+ scx_bypass(sch, true);
scx_propagate_disable_and_flush(sch);
switch (scx_set_enable_state(SCX_DISABLING)) {
@@ -4324,7 +4326,7 @@ static void scx_root_disable(struct scx_sched *sch)
WARN_ON_ONCE(scx_set_enable_state(SCX_DISABLED) != SCX_DISABLING);
done:
- scx_bypass(false);
+ scx_bypass(sch, false);
}
static void scx_disable_workfn(struct kthread_work *work)
@@ -4985,7 +4987,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
* scheduling) may not function correctly before all tasks are switched.
* Init in bypass mode to guarantee forward progress.
*/
- scx_bypass(true);
+ scx_bypass(sch, true);
for (i = SCX_OPI_NORMAL_BEGIN; i < SCX_OPI_NORMAL_END; i++)
if (((void (**)(void))ops)[i])
@@ -5096,7 +5098,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
scx_task_iter_stop(&sti);
percpu_up_write(&scx_fork_rwsem);
- scx_bypass(false);
+ scx_bypass(sch, false);
if (!scx_tryset_enable_state(SCX_ENABLED, SCX_ENABLING)) {
WARN_ON_ONCE(atomic_read(&sch->exit_kind) == SCX_EXIT_NONE);
@@ -5800,6 +5802,14 @@ void print_scx_info(const char *log_lvl, struct task_struct *p)
static int scx_pm_handler(struct notifier_block *nb, unsigned long event, void *ptr)
{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+
+ sch = rcu_dereference(scx_root);
+ if (!sch)
+ return NOTIFY_OK;
+
/*
* SCX schedulers often have userspace components which are sometimes
* involved in critial scheduling paths. PM operations involve freezing
@@ -5810,12 +5820,12 @@ static int scx_pm_handler(struct notifier_block *nb, unsigned long event, void *
case PM_HIBERNATION_PREPARE:
case PM_SUSPEND_PREPARE:
case PM_RESTORE_PREPARE:
- scx_bypass(true);
+ scx_bypass(sch, true);
break;
case PM_POST_HIBERNATION:
case PM_POST_SUSPEND:
case PM_POST_RESTORE:
- scx_bypass(false);
+ scx_bypass(sch, false);
break;
}
@@ -6508,7 +6518,7 @@ static void scx_kick_cpu(struct scx_sched *sch, s32 cpu, u64 flags)
* lead to irq_work_queue() malfunction such as infinite busy wait for
* IRQ status update. Suppress kicking.
*/
- if (scx_rq_bypassing(this_rq))
+ if (scx_bypassing(sch, cpu_of(this_rq)))
goto out;
/*
diff --git a/kernel/sched/ext_idle.c b/kernel/sched/ext_idle.c
index ce9c2f345ec5..39732e58ceb6 100644
--- a/kernel/sched/ext_idle.c
+++ b/kernel/sched/ext_idle.c
@@ -768,7 +768,8 @@ void __scx_update_idle(struct rq *rq, bool idle, bool do_notify)
* either enqueue() sees the idle bit or update_idle() sees the task
* that enqueue() queued.
*/
- if (SCX_HAS_OP(sch, update_idle) && do_notify && !scx_rq_bypassing(rq))
+ if (SCX_HAS_OP(sch, update_idle) && do_notify &&
+ !scx_bypassing(sch, cpu_of(rq)))
SCX_CALL_OP(sch, SCX_KF_REST, update_idle, rq, cpu_of(rq), idle);
}
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 154993921c38..083ca14f03e2 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -909,7 +909,13 @@ struct scx_event_stats {
s64 SCX_EV_INSERT_NOT_OWNED;
};
+enum scx_sched_pcpu_flags {
+ SCX_SCHED_PCPU_BYPASSING = 1LLU << 0,
+};
+
struct scx_sched_pcpu {
+ u64 flags; /* protected by rq lock */
+
/*
* The event counters are in a per-CPU variable to minimize the
* accounting overhead. A system-wide view on the event counter is
@@ -934,6 +940,7 @@ struct scx_sched {
struct scx_dispatch_q **global_dsqs;
struct scx_sched_pcpu __percpu *pcpu;
+ s32 bypass_depth;
s32 level;
bool warned_zero_slice:1;
bool warned_deprecated_rq:1;
@@ -1138,9 +1145,10 @@ static inline bool scx_kf_allowed_if_unlocked(void)
return !current->scx.kf_mask;
}
-static inline bool scx_rq_bypassing(struct rq *rq)
+static inline bool scx_bypassing(struct scx_sched *sch, s32 cpu)
{
- return unlikely(rq->scx.flags & SCX_RQ_BYPASSING);
+ return unlikely(per_cpu_ptr(sch->pcpu, cpu)->flags &
+ SCX_SCHED_PCPU_BYPASSING);
}
#ifdef CONFIG_EXT_SUB_SCHED
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index be9745d104f7..add7c0c218d4 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -755,7 +755,6 @@ enum scx_rq_flags {
SCX_RQ_CAN_STOP_TICK = 1 << 1,
SCX_RQ_BAL_PENDING = 1 << 2, /* balance hasn't run yet */
SCX_RQ_BAL_KEEP = 1 << 3, /* balance decided to keep current */
- SCX_RQ_BYPASSING = 1 << 4,
SCX_RQ_CLK_VALID = 1 << 5, /* RQ clock is fresh and valid */
SCX_RQ_IN_WAKEUP = 1 << 16,
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 32/46] sched_ext: Make bypass mode sub-sched aware
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (30 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 31/46] sched_ext: Move bypass_depth into scx_sched Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 33/46] sched_ext: Factor out scx_dispatch_sched() Tejun Heo
` (13 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Bypass mode is used to simplify enable and disable paths and guarantee
forward progress when something goes wrong. When enabled, all tasks skip
BPF scheduling and fall back to simple in-kernel FIFO scheduling. While
this global behavior can be used as-is when dealing with sub-scheds, that
would allow any sub-sched instance to affect the whole system in a
significantly disruptive manner.
Make bypass mode hierarchical instead. An scx_sched bypasses if itself or
any of its ancestors are in the bypass mode.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 38 +++++++++++++++++++++++++++++++-------
1 file changed, 31 insertions(+), 7 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 5f22a79e19ec..44f9cc7f0915 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -3949,6 +3949,7 @@ static void scx_bypass(struct scx_sched *sch, bool bypass)
{
static DEFINE_RAW_SPINLOCK(bypass_lock);
static unsigned long bypass_timestamp;
+ struct scx_sched *pos;
unsigned long flags;
int cpu;
@@ -3970,6 +3971,24 @@ static void scx_bypass(struct scx_sched *sch, bool bypass)
ktime_get_ns() - bypass_timestamp);
}
+ /*
+ * Bypass state is propagated to all descendants - an scx_sched bypasses
+ * if itself or any of its ancestors are in bypass mode.
+ */
+ raw_spin_lock(&scx_sched_lock);
+ scx_for_each_descendant_pre(pos, sch) {
+ if (pos == sch)
+ continue;
+ if (bypass) {
+ pos->bypass_depth++;
+ WARN_ON_ONCE(pos->bypass_depth <= 0);
+ } else {
+ pos->bypass_depth--;
+ WARN_ON_ONCE(pos->bypass_depth < 0);
+ }
+ }
+ raw_spin_unlock(&scx_sched_lock);
+
if (!scx_parent(sch))
atomic_inc(&scx_breather_depth);
@@ -3984,18 +4003,20 @@ static void scx_bypass(struct scx_sched *sch, bool bypass)
*/
for_each_possible_cpu(cpu) {
struct rq *rq = cpu_rq(cpu);
- struct scx_sched_pcpu *pcpu = per_cpu_ptr(sch->pcpu, cpu);
struct task_struct *p, *n;
raw_spin_rq_lock(rq);
- if (bypass) {
- WARN_ON_ONCE(pcpu->flags & SCX_SCHED_PCPU_BYPASSING);
- pcpu->flags |= SCX_SCHED_PCPU_BYPASSING;
- } else {
- WARN_ON_ONCE(!(pcpu->flags & SCX_SCHED_PCPU_BYPASSING));
- pcpu->flags &= ~SCX_SCHED_PCPU_BYPASSING;
+ raw_spin_lock(&scx_sched_lock);
+ scx_for_each_descendant_pre(pos, sch) {
+ struct scx_sched_pcpu *pcpu = per_cpu_ptr(pos->pcpu, cpu);
+
+ if (pos->bypass_depth)
+ pcpu->flags |= SCX_SCHED_PCPU_BYPASSING;
+ else
+ pcpu->flags &= ~SCX_SCHED_PCPU_BYPASSING;
}
+ raw_spin_unlock(&scx_sched_lock);
/*
* We need to guarantee that no tasks are on the BPF scheduler
@@ -4018,6 +4039,9 @@ static void scx_bypass(struct scx_sched *sch, bool bypass)
scx.runnable_node) {
struct sched_enq_and_set_ctx ctx;
+ if (!scx_is_descendant(scx_task_sched(p), sch))
+ continue;
+
/* cycling deq/enq is enough, see the function comment */
sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
sched_enq_and_set_task(&ctx);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 33/46] sched_ext: Factor out scx_dispatch_sched()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (31 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 32/46] sched_ext: Make bypass mode sub-sched aware Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 34/46] sched_ext: When calling ops.dispatch() @prev must be on the same scx_sched Tejun Heo
` (12 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
In preparation of multiple scheduler support, factor out
scx_dispatch_sched() from balance_one(). The function boundary makes
remembering $prev_on_scx and $prev_on_rq less useful. Open code $prev_on_scx
in balance_one() and $prev_on_rq in both balance_one() and
scx_dispatch_sched().
No functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 115 ++++++++++++++++++++++++---------------------
1 file changed, 62 insertions(+), 53 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 44f9cc7f0915..efe01ba84e2d 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2178,14 +2178,68 @@ static void flush_dispatch_buf(struct scx_sched *sch, struct rq *rq)
dspc->cursor = 0;
}
-static int balance_one(struct rq *rq, struct task_struct *prev)
+static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
+ struct task_struct *prev)
{
- struct scx_sched *sch = scx_root;
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
bool prev_on_scx = prev->sched_class == &ext_sched_class;
- bool prev_on_rq = prev->scx.flags & SCX_TASK_QUEUED;
int nr_loops = SCX_DSP_MAX_LOOPS;
+ if (consume_global_dsq(sch, rq))
+ return true;
+
+ if (unlikely(!SCX_HAS_OP(sch, dispatch)) ||
+ scx_bypassing(sch, cpu_of(rq)) || !scx_rq_online(rq))
+ return false;
+
+ dspc->rq = rq;
+
+ /*
+ * The dispatch loop. Because flush_dispatch_buf() may drop the rq lock,
+ * the local DSQ might still end up empty after a successful
+ * ops.dispatch(). If the local DSQ is empty even after ops.dispatch()
+ * produced some tasks, retry. The BPF scheduler may depend on this
+ * looping behavior to simplify its implementation.
+ */
+ do {
+ dspc->nr_tasks = 0;
+
+ SCX_CALL_OP(sch, SCX_KF_DISPATCH, dispatch, rq,
+ cpu_of(rq), prev_on_scx ? prev : NULL);
+
+ flush_dispatch_buf(sch, rq);
+
+ if ((prev->scx.flags & SCX_TASK_QUEUED) && prev->scx.slice) {
+ rq->scx.flags |= SCX_RQ_BAL_KEEP;
+ return true;
+ }
+ if (rq->scx.local_dsq.nr)
+ return true;
+ if (consume_global_dsq(sch, rq))
+ return true;
+
+ /*
+ * ops.dispatch() can trap us in this loop by repeatedly
+ * dispatching ineligible tasks. Break out once in a while to
+ * allow the watchdog to run. As IRQ can't be enabled in
+ * balance(), we want to complete this scheduling cycle and then
+ * start a new one. IOW, we want to call resched_curr() on the
+ * next, most likely idle, task, not the current one. Use
+ * __scx_bpf_kick_cpu() for deferred kicking.
+ */
+ if (unlikely(!--nr_loops)) {
+ scx_kick_cpu(sch, cpu_of(rq), 0);
+ break;
+ }
+ } while (dspc->nr_tasks);
+
+ return false;
+}
+
+static int balance_one(struct rq *rq, struct task_struct *prev)
+{
+ struct scx_sched *sch = scx_root;
+
lockdep_assert_rq_held(rq);
rq->scx.flags |= SCX_RQ_IN_BALANCE;
rq->scx.flags &= ~(SCX_RQ_BAL_PENDING | SCX_RQ_BAL_KEEP);
@@ -2204,7 +2258,7 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
rq->scx.cpu_released = false;
}
- if (prev_on_scx) {
+ if (prev->sched_class == &ext_sched_class) {
update_curr_scx(rq);
/*
@@ -2217,7 +2271,7 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
* See scx_disable_workfn() for the explanation on the bypassing
* test.
*/
- if (prev_on_rq && prev->scx.slice &&
+ if ((prev->scx.flags & SCX_TASK_QUEUED) && prev->scx.slice &&
!scx_bypassing(sch, cpu_of(rq))) {
rq->scx.flags |= SCX_RQ_BAL_KEEP;
goto has_tasks;
@@ -2228,60 +2282,15 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
if (rq->scx.local_dsq.nr)
goto has_tasks;
- if (consume_global_dsq(sch, rq))
+ /* dispatch @sch */
+ if (scx_dispatch_sched(sch, rq, prev))
goto has_tasks;
- if (unlikely(!SCX_HAS_OP(sch, dispatch)) ||
- scx_bypassing(sch, cpu_of(rq)) || !scx_rq_online(rq))
- goto no_tasks;
-
- dspc->rq = rq;
-
- /*
- * The dispatch loop. Because flush_dispatch_buf() may drop the rq lock,
- * the local DSQ might still end up empty after a successful
- * ops.dispatch(). If the local DSQ is empty even after ops.dispatch()
- * produced some tasks, retry. The BPF scheduler may depend on this
- * looping behavior to simplify its implementation.
- */
- do {
- dspc->nr_tasks = 0;
-
- SCX_CALL_OP(sch, SCX_KF_DISPATCH, dispatch, rq,
- cpu_of(rq), prev_on_scx ? prev : NULL);
-
- flush_dispatch_buf(sch, rq);
-
- if (prev_on_rq && prev->scx.slice) {
- rq->scx.flags |= SCX_RQ_BAL_KEEP;
- goto has_tasks;
- }
- if (rq->scx.local_dsq.nr)
- goto has_tasks;
- if (consume_global_dsq(sch, rq))
- goto has_tasks;
-
- /*
- * ops.dispatch() can trap us in this loop by repeatedly
- * dispatching ineligible tasks. Break out once in a while to
- * allow the watchdog to run. As IRQ can't be enabled in
- * balance(), we want to complete this scheduling cycle and then
- * start a new one. IOW, we want to call resched_curr() on the
- * next, most likely idle, task, not the current one. Use
- * scx_kick_cpu() for deferred kicking.
- */
- if (unlikely(!--nr_loops)) {
- scx_kick_cpu(sch, cpu_of(rq), 0);
- break;
- }
- } while (dspc->nr_tasks);
-
-no_tasks:
/*
* Didn't find another task to run. Keep running @prev unless
* %SCX_OPS_ENQ_LAST is in effect.
*/
- if (prev_on_rq &&
+ if ((prev->scx.flags & SCX_TASK_QUEUED) &&
(!(sch->ops.flags & SCX_OPS_ENQ_LAST) ||
scx_bypassing(sch, cpu_of(rq)))) {
rq->scx.flags |= SCX_RQ_BAL_KEEP;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 34/46] sched_ext: When calling ops.dispatch() @prev must be on the same scx_sched
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (32 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 33/46] sched_ext: Factor out scx_dispatch_sched() Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 35/46] sched_ext: Dispatch from all scx_sched instances Tejun Heo
` (11 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The @prev parameter passed into ops.dispatch() is expected to be on the
same sched. Passing in @prev which isn't on the sched can spuriously
trigger failures that can kill the scheduler. Pass in @prev iff it's on
the same sched.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index efe01ba84e2d..714b45a55112 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2182,8 +2182,9 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
struct task_struct *prev)
{
struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
- bool prev_on_scx = prev->sched_class == &ext_sched_class;
int nr_loops = SCX_DSP_MAX_LOOPS;
+ bool prev_on_sch = (prev->sched_class == &ext_sched_class) &&
+ sch == rcu_access_pointer(prev->scx.sched);
if (consume_global_dsq(sch, rq))
return true;
@@ -2205,7 +2206,7 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
dspc->nr_tasks = 0;
SCX_CALL_OP(sch, SCX_KF_DISPATCH, dispatch, rq,
- cpu_of(rq), prev_on_scx ? prev : NULL);
+ cpu_of(rq), prev_on_sch ? prev : NULL);
flush_dispatch_buf(sch, rq);
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 35/46] sched_ext: Dispatch from all scx_sched instances
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (33 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 34/46] sched_ext: When calling ops.dispatch() @prev must be on the same scx_sched Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:58 ` [PATCH 36/46] sched_ext: Move scx_dsp_ctx and scx_dsp_max_batch into scx_sched Tejun Heo
` (10 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The cgroup sub-sched support involves invasive changes to many areas of
sched_ext. The overall scaffolding is now in place and the next step is
implementing sub-sched enable/disable.
To enable partial testing and verification, update balance_one() to
dispatch from all scx_sched instances until it finds a task to run. This
should keep scheduling working when sub-scheds are enabled with tasks on
them. This will be replaced by BPF-driven hierarchical operation.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 714b45a55112..75a4b05fced4 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2239,7 +2239,7 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
static int balance_one(struct rq *rq, struct task_struct *prev)
{
- struct scx_sched *sch = scx_root;
+ struct scx_sched *sch = scx_root, *pos;
lockdep_assert_rq_held(rq);
rq->scx.flags |= SCX_RQ_IN_BALANCE;
@@ -2283,9 +2283,13 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
if (rq->scx.local_dsq.nr)
goto has_tasks;
- /* dispatch @sch */
- if (scx_dispatch_sched(sch, rq, prev))
- goto has_tasks;
+ /*
+ * TEMPORARY - Dispatch all scheds. This will be replaced by BPF-driven
+ * hierarchical operation.
+ */
+ list_for_each_entry_rcu(pos, &scx_sched_all, all)
+ if (scx_dispatch_sched(pos, rq, prev))
+ goto has_tasks;
/*
* Didn't find another task to run. Keep running @prev unless
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 36/46] sched_ext: Move scx_dsp_ctx and scx_dsp_max_batch into scx_sched
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (34 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 35/46] sched_ext: Dispatch from all scx_sched instances Tejun Heo
@ 2025-09-20 0:58 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 37/46] sched_ext: Make watchdog sub-sched aware Tejun Heo
` (9 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:58 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
scx_dsp_ctx and scx_dsp_max_batch are global variables used in the dispatch
path. In prepration for multiple scheduler support, move the former into
scx_sched_pcpu and the latter into scx_sched. No user-visible behavior
changes intended.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 56 ++++++++++---------------------------
kernel/sched/ext_internal.h | 18 ++++++++++++
2 files changed, 33 insertions(+), 41 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 75a4b05fced4..3fcf6cd7fa00 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -102,25 +102,6 @@ static const struct rhashtable_params dsq_hash_params = {
static LLIST_HEAD(dsqs_to_free);
-/* dispatch buf */
-struct scx_dsp_buf_ent {
- struct task_struct *task;
- unsigned long qseq;
- u64 dsq_id;
- u64 enq_flags;
-};
-
-static u32 scx_dsp_max_batch;
-
-struct scx_dsp_ctx {
- struct rq *rq;
- u32 cursor;
- u32 nr_tasks;
- struct scx_dsp_buf_ent buf[];
-};
-
-static struct scx_dsp_ctx __percpu *scx_dsp_ctx;
-
/* string formatting from BPF */
struct scx_bstr_buf {
u64 data[MAX_BPRINTF_VARARGS];
@@ -2164,7 +2145,7 @@ static void finish_dispatch(struct scx_sched *sch, struct rq *rq,
static void flush_dispatch_buf(struct scx_sched *sch, struct rq *rq)
{
- struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
+ struct scx_dsp_ctx *dspc = &this_cpu_ptr(sch->pcpu)->dsp_ctx;
u32 u;
for (u = 0; u < dspc->cursor; u++) {
@@ -2181,7 +2162,7 @@ static void flush_dispatch_buf(struct scx_sched *sch, struct rq *rq)
static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
struct task_struct *prev)
{
- struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
+ struct scx_dsp_ctx *dspc = &this_cpu_ptr(sch->pcpu)->dsp_ctx;
int nr_loops = SCX_DSP_MAX_LOOPS;
bool prev_on_sch = (prev->sched_class == &ext_sched_class) &&
sch == rcu_access_pointer(prev->scx.sched);
@@ -4356,10 +4337,6 @@ static void scx_root_disable(struct scx_sched *sch)
*/
kobject_del(&sch->kobj);
- free_percpu(scx_dsp_ctx);
- scx_dsp_ctx = NULL;
- scx_dsp_max_batch = 0;
-
mutex_unlock(&scx_enable_mutex);
WARN_ON_ONCE(scx_set_enable_state(SCX_DISABLED) != SCX_DISABLING);
@@ -4785,7 +4762,10 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
sch->global_dsqs[node] = dsq;
}
- sch->pcpu = alloc_percpu(struct scx_sched_pcpu);
+ sch->dsp_max_batch = ops->dispatch_max_batch ?: SCX_DSP_DFL_MAX_BATCH;
+ sch->pcpu = __alloc_percpu(struct_size_t(struct scx_sched_pcpu,
+ dsp_ctx.buf, sch->dsp_max_batch),
+ __alignof__(struct scx_sched_pcpu));
if (!sch->pcpu)
goto err_free_gdsqs;
@@ -4999,16 +4979,6 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
if (ret)
goto err_disable;
- WARN_ON_ONCE(scx_dsp_ctx);
- scx_dsp_max_batch = ops->dispatch_max_batch ?: SCX_DSP_DFL_MAX_BATCH;
- scx_dsp_ctx = __alloc_percpu(struct_size_t(struct scx_dsp_ctx, buf,
- scx_dsp_max_batch),
- __alignof__(struct scx_dsp_ctx));
- if (!scx_dsp_ctx) {
- ret = -ENOMEM;
- goto err_disable;
- }
-
if (ops->timeout_ms)
timeout = msecs_to_jiffies(ops->timeout_ms);
else
@@ -5947,7 +5917,7 @@ static bool scx_dsq_insert_preamble(struct scx_sched *sch, struct task_struct *p
static void scx_dsq_insert_commit(struct scx_sched *sch, struct task_struct *p,
u64 dsq_id, u64 enq_flags)
{
- struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
+ struct scx_dsp_ctx *dspc = &this_cpu_ptr(sch->pcpu)->dsp_ctx;
struct task_struct *ddsp_task;
ddsp_task = __this_cpu_read(direct_dispatch_task);
@@ -5956,7 +5926,7 @@ static void scx_dsq_insert_commit(struct scx_sched *sch, struct task_struct *p,
return;
}
- if (unlikely(dspc->cursor >= scx_dsp_max_batch)) {
+ if (unlikely(dspc->cursor >= sch->dsp_max_batch)) {
scx_error(sch, "dispatch buffer overflow");
return;
}
@@ -6204,7 +6174,7 @@ __bpf_kfunc u32 scx_bpf_dispatch_nr_slots(const struct bpf_prog_aux *aux__prog)
if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return 0;
- return scx_dsp_max_batch - __this_cpu_read(scx_dsp_ctx->cursor);
+ return sch->dsp_max_batch - __this_cpu_read(sch->pcpu->dsp_ctx.cursor);
}
/**
@@ -6216,8 +6186,8 @@ __bpf_kfunc u32 scx_bpf_dispatch_nr_slots(const struct bpf_prog_aux *aux__prog)
*/
__bpf_kfunc void scx_bpf_dispatch_cancel(const struct bpf_prog_aux *aux__prog)
{
- struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct scx_sched *sch;
+ struct scx_dsp_ctx *dspc;
guard(rcu)();
@@ -6228,6 +6198,8 @@ __bpf_kfunc void scx_bpf_dispatch_cancel(const struct bpf_prog_aux *aux__prog)
if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return;
+ dspc = &this_cpu_ptr(sch->pcpu)->dsp_ctx;
+
if (dspc->cursor > 0)
dspc->cursor--;
else
@@ -6252,9 +6224,9 @@ __bpf_kfunc void scx_bpf_dispatch_cancel(const struct bpf_prog_aux *aux__prog)
__bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id,
const struct bpf_prog_aux *aux__prog)
{
- struct scx_dsp_ctx *dspc = this_cpu_ptr(scx_dsp_ctx);
struct scx_dispatch_q *dsq;
struct scx_sched *sch;
+ struct scx_dsp_ctx *dspc;
guard(rcu)();
@@ -6265,6 +6237,8 @@ __bpf_kfunc bool scx_bpf_dsq_move_to_local(u64 dsq_id,
if (!scx_kf_allowed(sch, SCX_KF_DISPATCH))
return false;
+ dspc = &this_cpu_ptr(sch->pcpu)->dsp_ctx;
+
flush_dispatch_buf(sch, dspc->rq);
dsq = find_user_dsq(sch, dsq_id);
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 083ca14f03e2..8dbdae910564 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -913,6 +913,21 @@ enum scx_sched_pcpu_flags {
SCX_SCHED_PCPU_BYPASSING = 1LLU << 0,
};
+/* dispatch buf */
+struct scx_dsp_buf_ent {
+ struct task_struct *task;
+ unsigned long qseq;
+ u64 dsq_id;
+ u64 enq_flags;
+};
+
+struct scx_dsp_ctx {
+ struct rq *rq;
+ u32 cursor;
+ u32 nr_tasks;
+ struct scx_dsp_buf_ent buf[];
+};
+
struct scx_sched_pcpu {
u64 flags; /* protected by rq lock */
@@ -922,6 +937,8 @@ struct scx_sched_pcpu {
* constructed when requested by scx_bpf_events().
*/
struct scx_event_stats event_stats;
+
+ struct scx_dsp_ctx dsp_ctx;
};
struct scx_sched {
@@ -941,6 +958,7 @@ struct scx_sched {
struct scx_sched_pcpu __percpu *pcpu;
s32 bypass_depth;
+ u32 dsp_max_batch;
s32 level;
bool warned_zero_slice:1;
bool warned_deprecated_rq:1;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 37/46] sched_ext: Make watchdog sub-sched aware
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (35 preceding siblings ...)
2025-09-20 0:58 ` [PATCH 36/46] sched_ext: Move scx_dsp_ctx and scx_dsp_max_batch into scx_sched Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 38/46] sched_ext: Convert scx_dump_state() spinlock to raw spinlock Tejun Heo
` (8 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Currently, the watchdog checks all tasks as if they are all on scx_root.
Move scx_watchdog_timeout inside scx_sched and make check_rq_for_timeouts()
use the timeout from the scx_sched associated with each task.
refresh_watchdog() is added, which determines the timer interval as half of
the shortest watchdog timeouts of all scheds and arms or disarms it as
necessary. Every scx_sched instance has equivalent or better detection
latency while sharing the same timer.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 74 ++++++++++++++++++++++++-------------
kernel/sched/ext_internal.h | 7 ++++
2 files changed, 56 insertions(+), 25 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 3fcf6cd7fa00..4dc82afb7016 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -66,11 +66,10 @@ static atomic_long_t scx_hotplug_seq = ATOMIC_LONG_INIT(0);
static atomic_long_t scx_enable_seq = ATOMIC_LONG_INIT(0);
/*
- * The maximum amount of time in jiffies that a task may be runnable without
- * being scheduled on a CPU. If this timeout is exceeded, it will trigger
- * scx_error().
+ * Watchdog interval. All scx_sched's share a single watchdog timer and the
+ * interval is half of the shortest sch->watchdog_timeout.
*/
-static unsigned long scx_watchdog_timeout;
+static unsigned long scx_watchdog_interval;
/*
* The last time the delayed work was run. This delayed work relies on
@@ -2761,10 +2760,11 @@ static bool check_rq_for_timeouts(struct rq *rq)
goto out_unlock;
list_for_each_entry(p, &rq->scx.runnable_list, scx.runnable_node) {
+ struct scx_sched *sch = scx_task_sched(p);
unsigned long last_runnable = p->scx.runnable_at;
if (unlikely(time_after(jiffies,
- last_runnable + scx_watchdog_timeout))) {
+ last_runnable + sch->watchdog_timeout))) {
u32 dur_ms = jiffies_to_msecs(jiffies - last_runnable);
scx_exit(sch, SCX_EXIT_ERROR_STALL, 0,
@@ -2781,6 +2781,7 @@ static bool check_rq_for_timeouts(struct rq *rq)
static void scx_watchdog_workfn(struct work_struct *work)
{
+ unsigned long intv;
int cpu;
WRITE_ONCE(scx_watchdog_timestamp, jiffies);
@@ -2791,28 +2792,31 @@ static void scx_watchdog_workfn(struct work_struct *work)
cond_resched();
}
- queue_delayed_work(system_unbound_wq, to_delayed_work(work),
- scx_watchdog_timeout / 2);
+
+ intv = READ_ONCE(scx_watchdog_interval);
+ if (intv < ULONG_MAX)
+ queue_delayed_work(system_unbound_wq, to_delayed_work(work),
+ intv);
}
void scx_tick(struct rq *rq)
{
- struct scx_sched *sch;
+ struct scx_sched *root;
unsigned long last_check;
if (!scx_enabled())
return;
- sch = rcu_dereference_bh(scx_root);
- if (unlikely(!sch))
+ root = rcu_dereference_bh(scx_root);
+ if (unlikely(!root))
return;
last_check = READ_ONCE(scx_watchdog_timestamp);
if (unlikely(time_after(jiffies,
- last_check + READ_ONCE(scx_watchdog_timeout)))) {
+ last_check + READ_ONCE(root->watchdog_timeout)))) {
u32 dur_ms = jiffies_to_msecs(jiffies - last_check);
- scx_exit(sch, SCX_EXIT_ERROR_STALL, 0,
+ scx_exit(root, SCX_EXIT_ERROR_STALL, 0,
"watchdog failed to check in for %u.%03us",
dur_ms / 1000, dur_ms % 1000);
}
@@ -4108,6 +4112,26 @@ static const char *scx_exit_reason(enum scx_exit_kind kind)
}
}
+static void refresh_watchdog(void)
+{
+ struct scx_sched *sch;
+ unsigned long intv = ULONG_MAX;
+
+ /* take the shortest timeout and use its half for watchdog interval */
+ rcu_read_lock();
+ list_for_each_entry_rcu(sch, &scx_sched_all, all)
+ intv = max(min(intv, sch->watchdog_timeout / 2), 1);
+ rcu_read_unlock();
+
+ WRITE_ONCE(scx_watchdog_timestamp, jiffies);
+ WRITE_ONCE(scx_watchdog_interval, intv);
+
+ if (intv < ULONG_MAX)
+ mod_delayed_work(system_unbound_wq, &scx_watchdog_work, intv);
+ else
+ cancel_delayed_work_sync(&scx_watchdog_work);
+}
+
#ifdef CONFIG_EXT_SUB_SCHED
static DECLARE_WAIT_QUEUE_HEAD(scx_unlink_waitq);
@@ -4159,6 +4183,8 @@ static void scx_sub_disable(struct scx_sched *sch)
list_del_rcu(&sch->all);
raw_spin_unlock_irq(&scx_sched_lock);
+ refresh_watchdog();
+
mutex_unlock(&scx_enable_mutex);
/*
@@ -4316,12 +4342,12 @@ static void scx_root_disable(struct scx_sched *sch)
if (sch->ops.exit)
SCX_CALL_OP(sch, SCX_KF_UNLOCKED, exit, NULL, ei);
- cancel_delayed_work_sync(&scx_watchdog_work);
-
raw_spin_lock_irq(&scx_sched_lock);
list_del_rcu(&sch->all);
raw_spin_unlock_irq(&scx_sched_lock);
+ refresh_watchdog();
+
/*
* scx_root clearing must be inside cpus_read_lock(). See
* handle_hotplug().
@@ -4780,6 +4806,11 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
sch->ancestors[level] = sch;
sch->level = level;
+ if (ops->timeout_ms)
+ sch->watchdog_timeout = msecs_to_jiffies(ops->timeout_ms);
+ else
+ sch->watchdog_timeout = SCX_WATCHDOG_MAX_TIMEOUT;
+
atomic_set(&sch->exit_kind, SCX_EXIT_NONE);
init_irq_work(&sch->error_irq_work, scx_error_irq_workfn);
kthread_init_work(&sch->disable_work, scx_disable_workfn);
@@ -4899,7 +4930,6 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
struct scx_sched *sch;
struct scx_task_iter sti;
struct task_struct *p;
- unsigned long timeout;
int i, cpu, ret;
if (!cpumask_equal(housekeeping_cpumask(HK_TYPE_DOMAIN),
@@ -4953,6 +4983,8 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
list_add_tail_rcu(&sch->all, &scx_sched_all);
raw_spin_unlock_irq(&scx_sched_lock);
+ refresh_watchdog();
+
scx_idle_enable(ops);
if (sch->ops.init) {
@@ -4979,16 +5011,6 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
if (ret)
goto err_disable;
- if (ops->timeout_ms)
- timeout = msecs_to_jiffies(ops->timeout_ms);
- else
- timeout = SCX_WATCHDOG_MAX_TIMEOUT;
-
- WRITE_ONCE(scx_watchdog_timeout, timeout);
- WRITE_ONCE(scx_watchdog_timestamp, jiffies);
- queue_delayed_work(system_unbound_wq, &scx_watchdog_work,
- scx_watchdog_timeout / 2);
-
/*
* Once __scx_enabled is set, %current can be switched to SCX anytime.
* This can lead to stalls as some BPF schedulers (e.g. userspace
@@ -5215,6 +5237,8 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
list_add_tail_rcu(&sch->all, &scx_sched_all);
raw_spin_unlock_irq(&scx_sched_lock);
+ refresh_watchdog();
+
if (sch->level >= SCX_SUB_MAX_DEPTH) {
scx_error(sch, "max nesting depth %d violated",
SCX_SUB_MAX_DEPTH);
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 8dbdae910564..4399c003c15f 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -974,6 +974,13 @@ struct scx_sched {
struct kset *sub_kset;
#endif /* CONFIG_EXT_SUB_SCHED */
+ /*
+ * The maximum amount of time in jiffies that a task may be runnable
+ * without being scheduled on a CPU. If this timeout is exceeded, it
+ * will trigger scx_error().
+ */
+ unsigned long watchdog_timeout;
+
atomic_t exit_kind;
struct scx_exit_info *exit_info;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 38/46] sched_ext: Convert scx_dump_state() spinlock to raw spinlock
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (36 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 37/46] sched_ext: Make watchdog sub-sched aware Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 39/46] sched_ext: Support dumping multiple schedulers and add scheduler identification Tejun Heo
` (7 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The scx_dump_state() function uses a regular spinlock to serialize
access. In a subsequent patch, this function will be called while
holding scx_sched_lock, which is a raw spinlock, creating a lock
nesting violation.
Convert the dump_lock to a raw spinlock and use the guard macro for
cleaner lock management.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 4dc82afb7016..fafffe3ee812 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4565,7 +4565,7 @@ static void scx_dump_task(struct seq_buf *s, struct scx_dump_ctx *dctx,
static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
{
- static DEFINE_SPINLOCK(dump_lock);
+ static DEFINE_RAW_SPINLOCK(dump_lock);
static const char trunc_marker[] = "\n\n~~~~ TRUNCATED ~~~~\n";
struct scx_sched *sch = scx_root;
struct scx_dump_ctx dctx = {
@@ -4577,11 +4577,10 @@ static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
};
struct seq_buf s;
struct scx_event_stats events;
- unsigned long flags;
char *buf;
int cpu;
- spin_lock_irqsave(&dump_lock, flags);
+ guard(raw_spinlock_irqsave)(&dump_lock);
seq_buf_init(&s, ei->dump, dump_len);
@@ -4705,8 +4704,6 @@ static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
if (seq_buf_has_overflowed(&s) && dump_len >= sizeof(trunc_marker))
memcpy(ei->dump + dump_len - sizeof(trunc_marker),
trunc_marker, sizeof(trunc_marker));
-
- spin_unlock_irqrestore(&dump_lock, flags);
}
static void scx_error_irq_workfn(struct irq_work *irq_work)
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 39/46] sched_ext: Support dumping multiple schedulers and add scheduler identification
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (37 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 38/46] sched_ext: Convert scx_dump_state() spinlock to raw spinlock Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 40/46] sched_ext: Implement cgroup sub-sched enabling and disabling Tejun Heo
` (6 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Extend scx_dump_state() to support multiple schedulers and improve task
identification in dumps. The function now takes a specific scheduler to
dump and can optionally filter tasks by scheduler.
scx_dump_task() now displays which scheduler each task belongs to, using
"*" to mark tasks owned by the scheduler being dumped. Sub-schedulers
are identified with their level and cgroup ID.
The SysRq-D handler now iterates through all active schedulers under
scx_sched_lock and dumps each one separately. For SysRq-D dumps, only
tasks owned by each scheduler are dumped to avoid redundancy since all
schedulers are being dumped. Error-triggered dumps continue to dump all
tasks since only that specific scheduler is being dumped.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 54 ++++++++++++++++++++++++++++++++++++----------
1 file changed, 43 insertions(+), 11 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index fafffe3ee812..1a48510e6f98 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4520,22 +4520,34 @@ static void ops_dump_exit(void)
scx_dump_data.cpu = -1;
}
-static void scx_dump_task(struct seq_buf *s, struct scx_dump_ctx *dctx,
+static void scx_dump_task(struct scx_sched *sch,
+ struct seq_buf *s, struct scx_dump_ctx *dctx,
struct task_struct *p, char marker)
{
static unsigned long bt[SCX_EXIT_BT_LEN];
- struct scx_sched *sch = scx_root;
+ struct scx_sched *task_sch = scx_task_sched(p);
+ const char *own_marker;
+ char sch_id_buf[32];
char dsq_id_buf[19] = "(n/a)";
unsigned long ops_state = atomic_long_read(&p->scx.ops_state);
unsigned int bt_len = 0;
+ own_marker = task_sch == sch ? "*" : "";
+
+ if (task_sch->level == 0)
+ scnprintf(sch_id_buf, sizeof(sch_id_buf), "root");
+ else
+ scnprintf(sch_id_buf, sizeof(sch_id_buf), "sub%d-%llu",
+ task_sch->level, task_sch->ops.sub_cgroup_id);
+
if (p->scx.dsq)
scnprintf(dsq_id_buf, sizeof(dsq_id_buf), "0x%llx",
(unsigned long long)p->scx.dsq->id);
dump_newline(s);
- dump_line(s, " %c%c %s[%d] %+ldms",
+ dump_line(s, " %c%c %s[%d] %s%s %+ldms",
marker, task_state_to_char(p), p->comm, p->pid,
+ own_marker, sch_id_buf,
jiffies_delta_msecs(p->scx.runnable_at, dctx->at_jiffies));
dump_line(s, " scx_state/flags=%u/0x%x dsq_flags=0x%x ops_state/qseq=%lu/%lu",
scx_get_task_state(p), p->scx.flags & ~SCX_TASK_STATE_MASK,
@@ -4563,11 +4575,18 @@ static void scx_dump_task(struct seq_buf *s, struct scx_dump_ctx *dctx,
}
}
-static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
+/*
+ * Dump scheduler state. If @dump_all_tasks is true, dump all tasks regardless
+ * of which scheduler they belong to. If false, only dump tasks owned by @sch.
+ * For SysRq-D dumps, @dump_all_tasks=false since all schedulers are dumped
+ * separately. For error dumps, @dump_all_tasks=true since only the failing
+ * scheduler is dumped.
+ */
+static void scx_dump_state(struct scx_sched *sch, struct scx_exit_info *ei,
+ size_t dump_len, bool dump_all_tasks)
{
static DEFINE_RAW_SPINLOCK(dump_lock);
static const char trunc_marker[] = "\n\n~~~~ TRUNCATED ~~~~\n";
- struct scx_sched *sch = scx_root;
struct scx_dump_ctx dctx = {
.kind = ei->kind,
.exit_code = ei->exit_code,
@@ -4584,6 +4603,14 @@ static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
seq_buf_init(&s, ei->dump, dump_len);
+#ifdef CONFIG_EXT_SUB_SCHED
+ if (sch->level == 0)
+ dump_line(&s, "%s: root", sch->ops.name);
+ else
+ dump_line(&s, "%s: sub%d-%llu %s",
+ sch->ops.name, sch->level, sch->ops.sub_cgroup_id,
+ sch->cgrp_path);
+#endif
if (ei->kind == SCX_EXIT_NONE) {
dump_line(&s, "Debug dump triggered by %s", ei->reason);
} else {
@@ -4676,11 +4703,13 @@ static void scx_dump_state(struct scx_exit_info *ei, size_t dump_len)
seq_buf_set_overflow(&s);
}
- if (rq->curr->sched_class == &ext_sched_class)
- scx_dump_task(&s, &dctx, rq->curr, '*');
+ if (rq->curr->sched_class == &ext_sched_class &&
+ (dump_all_tasks || scx_task_sched(rq->curr) == sch))
+ scx_dump_task(sch, &s, &dctx, rq->curr, '*');
list_for_each_entry(p, &rq->scx.runnable_list, scx.runnable_node)
- scx_dump_task(&s, &dctx, p, ' ');
+ if (dump_all_tasks || scx_task_sched(p) == sch)
+ scx_dump_task(sch, &s, &dctx, p, ' ');
next:
rq_unlock(rq, &rf);
}
@@ -4712,7 +4741,7 @@ static void scx_error_irq_workfn(struct irq_work *irq_work)
struct scx_exit_info *ei = sch->exit_info;
if (ei->kind >= SCX_EXIT_ERROR)
- scx_dump_state(ei, sch->ops.exit_dump_len);
+ scx_dump_state(sch, ei, sch->ops.exit_dump_len, true);
kthread_queue_work(sch->helper, &sch->disable_work);
}
@@ -5662,9 +5691,12 @@ static const struct sysrq_key_op sysrq_sched_ext_reset_op = {
static void sysrq_handle_sched_ext_dump(u8 key)
{
struct scx_exit_info ei = { .kind = SCX_EXIT_NONE, .reason = "SysRq-D" };
+ struct scx_sched *sch;
- if (scx_enabled())
- scx_dump_state(&ei, 0);
+ guard(raw_spinlock_irqsave)(&scx_sched_lock);
+
+ list_for_each_entry_rcu(sch, &scx_sched_all, all)
+ scx_dump_state(sch, &ei, 0, false);
}
static const struct sysrq_key_op sysrq_sched_ext_dump_op = {
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 40/46] sched_ext: Implement cgroup sub-sched enabling and disabling
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (38 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 39/46] sched_ext: Support dumping multiple schedulers and add scheduler identification Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 41/46] HACK_NOT_FOR_UPSTREAM: sched_ext: Work around @aux__prog prototype mismatch Tejun Heo
` (5 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
The preceding changes implemented the framework to support cgroup
sub-scheds and updated scheduling paths and kfuncs so that they have
minimal but working support for sub-scheds. However, actual sub-sched
enabling/disabling hasn't been implemented yet and all tasks stayed on
scx_root.
Implement cgroup sub-sched enabling and disabling to actually activate
sub-scheds:
- Both enable and disable operations bypass only the tasks in the subtree
of the child being enabled or disabled to limit disruptions.
- When enabling, all candidate tasks are first initialized for the child
sched. Once that succeeds, the tasks are exited for the parent and then
switched over to the child. This adds a bit of complication but
guarantees that child scheduler failures are always contained.
- Disabling works the same way in the other direction. However, when the
parent may fail to initialize a task, disabling is propagated up to the
parent. While this means that a parent sched fail due to a child sched
event, the failure can only originate from the parent itself (its
ops.init_task()). The only effect a malfunctioning child can have on the
parent is attempting to move the tasks back to the parent.
After this change, although not all the necessary mechanisms are in place
yet, sub-scheds can take control of their tasks and schedule them.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/sched/ext.h | 1 +
kernel/sched/ext.c | 281 +++++++++++++++++++++++++++++++++++++-
2 files changed, 276 insertions(+), 6 deletions(-)
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index 73f9df0759e2..df1111d245bc 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -74,6 +74,7 @@ enum scx_ent_flags {
SCX_TASK_QUEUED = 1 << 0, /* on ext runqueue */
SCX_TASK_RESET_RUNNABLE_AT = 1 << 2, /* runnable_at should be reset */
SCX_TASK_DEQD_FOR_SLEEP = 1 << 3, /* last dequeue was for SLEEP */
+ SCX_TASK_SUB_INIT = 1 << 4, /* task being initialized for a sub sched */
SCX_TASK_STATE_SHIFT = 8, /* bit 8 and 9 are used to carry scx_task_state */
SCX_TASK_STATE_BITS = 2,
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 1a48510e6f98..eff5f6894f14 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -58,6 +58,15 @@ DEFINE_STATIC_KEY_FALSE(__scx_switched_all);
static atomic_long_t scx_nr_rejected = ATOMIC_LONG_INIT(0);
static atomic_long_t scx_hotplug_seq = ATOMIC_LONG_INIT(0);
+#ifdef CONFIG_EXT_SUB_SCHED
+/*
+ * The sub sched being enabled. Used by scx_disable_and_exit_task() to exit
+ * tasks for the sub-sched being enabled. Use a global variable instead of a
+ * per-task field as all enables are serialized.
+ */
+static struct scx_sched *scx_enabling_sub_sched = NULL;
+#endif /* CONFIG_EXT_SUB_SCHED */
+
/*
* A monotically increasing sequence number that is incremented every time a
* scheduler is enabled. This can be used by to check if any custom sched_ext
@@ -3047,6 +3056,17 @@ static void scx_disable_and_exit_task(struct scx_sched *sch,
{
__scx_disable_and_exit_task(sch, p);
+ /*
+ * If set, @p exited between __scx_init_task() and scx_enable_task() in
+ * scx_sub_enable() and is initialized for both the associated sched and
+ * its parent. Disable and exit for the child too.
+ */
+ if ((p->scx.flags & SCX_TASK_SUB_INIT) &&
+ !WARN_ON_ONCE(!scx_enabling_sub_sched)) {
+ __scx_disable_and_exit_task(scx_enabling_sub_sched, p);
+ p->scx.flags &= ~SCX_TASK_SUB_INIT;
+ }
+
scx_set_task_sched(p, NULL);
scx_set_task_state(p, SCX_TASK_NONE);
}
@@ -3082,9 +3102,11 @@ int scx_fork(struct task_struct *p, struct kernel_clone_args *kargs)
percpu_rwsem_assert_held(&scx_fork_rwsem);
if (scx_init_task_enabled) {
- ret = scx_init_task(scx_root, p, true);
+ struct scx_sched *sch = kargs->cset->dfl_cgrp->scx_sched;
+
+ ret = scx_init_task(sch, p, true);
if (!ret)
- scx_set_task_sched(p, scx_root);
+ scx_set_task_sched(p, sch);
return ret;
}
@@ -4004,9 +4026,9 @@ static void scx_bypass(struct scx_sched *sch, bool bypass)
struct rq *rq = cpu_rq(cpu);
struct task_struct *p, *n;
+ raw_spin_lock(&scx_sched_lock); /* nests outside rq_lock */
raw_spin_rq_lock(rq);
- raw_spin_lock(&scx_sched_lock);
scx_for_each_descendant_pre(pos, sch) {
struct scx_sched_pcpu *pcpu = per_cpu_ptr(pos->pcpu, cpu);
@@ -4015,6 +4037,7 @@ static void scx_bypass(struct scx_sched *sch, bool bypass)
else
pcpu->flags &= ~SCX_SCHED_PCPU_BYPASSING;
}
+
raw_spin_unlock(&scx_sched_lock);
/*
@@ -4161,23 +4184,141 @@ static void scx_propagate_disable_and_flush(struct scx_sched *sch)
wait_event(scx_unlink_waitq, list_empty(&sch->children));
}
+static void scx_fail_parent(struct scx_sched *sch,
+ struct task_struct *failed, int fail_code)
+{
+ struct scx_sched *parent = scx_parent(sch);
+ struct scx_task_iter sti;
+ struct task_struct *p;
+ struct sched_enq_and_set_ctx ctx;
+
+ scx_error(parent, "ops.init_task() failed (%d) for %s[%d] while disabling a sub-scheduler",
+ fail_code, failed->comm, failed->pid);
+
+ /*
+ * Once $parent is bypassed, it's safe to put SCX_TASK_NONE tasks into
+ * it. This may cause downstream failures on the BPF side but $parent is
+ * dying anyway.
+ */
+ scx_bypass(parent, true);
+
+ scx_task_iter_start(&sti, sch->cgrp);
+ while ((p = scx_task_iter_next_locked(&sti))) {
+ if (scx_task_sched(p) == parent)
+ continue;
+
+ sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
+ scx_disable_and_exit_task(sch, p);
+ rcu_assign_pointer(p->scx.sched, parent);
+ sched_enq_and_set_task(&ctx);
+ }
+ scx_task_iter_stop(&sti);
+}
+
static void scx_sub_disable(struct scx_sched *sch)
{
struct scx_sched *parent = scx_parent(sch);
+ struct scx_task_iter sti;
+ struct task_struct *p;
+ int ret;
+ /*
+ * Guarantee forward progress and disable all descendants. To limit
+ * disruptions, $parent is not bypassed. Tasks are fully prepped and
+ * then inserted back into $parent.
+ */
+ scx_bypass(sch, true);
scx_propagate_disable_and_flush(sch);
+ /*
+ * Here, every runnable task is guaranteed to make forward progress and
+ * we can safely use blocking synchronization constructs. Actually
+ * disable ops.
+ */
mutex_lock(&scx_enable_mutex);
percpu_down_write(&scx_fork_rwsem);
scx_cgroup_lock();
set_cgroup_sched(sch->cgrp, parent);
- /* TODO - perform actual disabling here */
+ scx_task_iter_start(&sti, sch->cgrp);
+ while ((p = scx_task_iter_next_locked(&sti))) {
+ struct rq *rq;
+ struct rq_flags rf;
+ struct sched_enq_and_set_ctx ctx;
+
+ /* filter out duplicate visits */
+ if (scx_task_sched(p) == parent)
+ continue;
+
+ /*
+ * By the time control reaches here, all descendant schedulers
+ * should already have been disabled.
+ */
+ WARN_ON_ONCE(scx_task_sched(p) != sch);
+
+ /*
+ * If $p is about to be freed, nothing prevents $sch from
+ * unloading before $p reaches sched_ext_free(). Disable and
+ * exit $p right away.
+ */
+ if (!tryget_task_struct(p)) {
+ scx_disable_and_exit_task(sch, p);
+ continue;
+ }
+
+ scx_task_iter_unlock(&sti);
+
+ /*
+ * $p is READY or ENABLED on @sch. Initialize for $parent,
+ * disable and exit from @sch, and then switch over to $parent.
+ *
+ * If a task fails to initialize for $parent, the only available
+ * action is disabling $parent too. While this allows disabling
+ * of a child sched to cause the parent scheduler to fail, the
+ * failure can only originate from ops.init_task() of the
+ * parent. A child can't directly affect the parent through its
+ * own failures.
+ */
+ ret = __scx_init_task(parent, p, false);
+ if (ret) {
+ scx_fail_parent(parent, p, ret);
+ put_task_struct(p);
+ break;
+ }
+
+ rq = task_rq_lock(p, &rf);
+ sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
+
+ /*
+ * $p is initialized for $parent and still attached to @sch.
+ * Disable and exit for @sch, switch over to $parent, override
+ * the state to READY to account for $p having already been
+ * initialized, and then enable.
+ */
+ scx_disable_and_exit_task(sch, p);
+ scx_set_task_state(p, SCX_TASK_INIT);
+ rcu_assign_pointer(p->scx.sched, parent);
+ scx_set_task_state(p, SCX_TASK_READY);
+ scx_enable_task(parent, p);
+
+ sched_enq_and_set_task(&ctx);
+ task_rq_unlock(rq, p, &rf);
+
+ put_task_struct(p);
+ }
+ scx_task_iter_stop(&sti);
scx_cgroup_unlock();
percpu_up_write(&scx_fork_rwsem);
+ /*
+ * All tasks are moved off of @sch but there may still be on-going
+ * operations (e.g. ops.select_cpu()). Drain them by flushing RCU. Use
+ * the expedited version as ancestors may be waiting in bypass mode.
+ */
+ synchronize_rcu_expedited();
+
raw_spin_lock_irq(&scx_sched_lock);
list_del_init(&sch->sibling);
list_del_rcu(&sch->all);
@@ -5222,11 +5363,29 @@ static struct scx_sched *find_parent_sched(struct cgroup *cgrp)
return parent;
}
+static bool assert_task_ready_or_enabled(struct task_struct *p)
+{
+ enum scx_task_state state = scx_get_task_state(p);
+
+ switch (state) {
+ case SCX_TASK_READY:
+ case SCX_TASK_ENABLED:
+ return true;
+ default:
+ WARN_ONCE(true, "sched_ext: Invalid task state %d for %s[%d] during enabling sub sched",
+ state, p->comm, p->pid);
+ return false;
+ }
+}
+
static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
{
struct cgroup *cgrp;
struct scx_sched *parent, *sch;
- int ret;
+ struct scx_task_iter sti;
+ struct task_struct *p;
+ struct sched_enq_and_set_ctx ctx;
+ int i, ret;
mutex_lock(&scx_enable_mutex);
@@ -5316,6 +5475,12 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
}
sch->sub_attached = true;
+ scx_bypass(sch, true);
+
+ for (i = SCX_OPI_BEGIN; i < SCX_OPI_END; i++)
+ if (((void (**)(void))ops)[i])
+ set_bit(i, sch->has_op);
+
percpu_down_write(&scx_fork_rwsem);
scx_cgroup_lock();
@@ -5329,16 +5494,119 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
goto err_unlock_and_disable;
}
- /* TODO - perform actual enabling here */
+ /*
+ * Initialize tasks for the new child $sch without exiting them for
+ * $parent so that the tasks can always be reverted back to $parent
+ * sched on child init failure.
+ */
+ WARN_ON_ONCE(scx_enabling_sub_sched);
+ scx_enabling_sub_sched = sch;
+
+ scx_task_iter_start(&sti, sch->cgrp);
+ while ((p = scx_task_iter_next_locked(&sti))) {
+ struct rq *rq;
+ struct rq_flags rf;
+
+ /*
+ * Task iteration may visit the same task twice when racing
+ * against exiting. Use %SCX_TASK_SUB_INIT to mark tasks which
+ * finished __scx_init_task() and skip if set.
+ *
+ * A task may exit and get freed between __scx_init_task()
+ * completion and scx_enable_task(). In such cases,
+ * scx_disable_and_exit_task() must exit the task for both the
+ * parent and child scheds.
+ */
+ if (p->scx.flags & SCX_TASK_SUB_INIT)
+ continue;
+
+ /* see scx_root_enable() */
+ if (!tryget_task_struct(p))
+ continue;
+
+ if (!assert_task_ready_or_enabled(p)) {
+ scx_task_iter_stop(&sti);
+ goto exit_tasks;
+ }
+
+ scx_task_iter_unlock(&sti);
+
+ /*
+ * As $p is still on $parent, it can't be transitioned to INIT.
+ * Let's worry about task state later. Use __scx_init_task().
+ */
+ ret = __scx_init_task(sch, p, false);
+ if (ret) {
+ put_task_struct(p);
+ scx_task_iter_stop(&sti);
+ goto exit_tasks;
+ }
+
+ rq = task_rq_lock(p, &rf);
+ p->scx.flags |= SCX_TASK_SUB_INIT;
+ task_rq_unlock(rq, p, &rf);
+
+ put_task_struct(p);
+ }
+ scx_task_iter_stop(&sti);
+
+ /*
+ * All tasks are prepped. Disable/exit tasks for $parent and enable for
+ * the new @sch.
+ */
+ scx_task_iter_start(&sti, sch->cgrp);
+ while ((p = scx_task_iter_next_locked(&sti))) {
+ /*
+ * Use clearing of %SCX_TASK_SUB_INIT to detect and skip
+ * duplicate iterations.
+ */
+ if (!(p->scx.flags & SCX_TASK_SUB_INIT))
+ continue;
+
+ sched_deq_and_put_task(p, DEQUEUE_SAVE | DEQUEUE_MOVE, &ctx);
+
+ /*
+ * $p must be either READY or ENABLED. If ENABLED,
+ * __scx_disabled_and_exit_task() first disables and makes it
+ * READY. However, after exiting $p, it will leave $p as READY.
+ */
+ assert_task_ready_or_enabled(p);
+ __scx_disable_and_exit_task(parent, p);
+
+ /*
+ * $p is now only initialized for @sch and READY, which is what
+ * we want. Assign it to @sch and enable.
+ */
+ rcu_assign_pointer(p->scx.sched, sch);
+ scx_enable_task(sch, p);
+
+ p->scx.flags &= ~SCX_TASK_SUB_INIT;
+
+ sched_enq_and_set_task(&ctx);
+ }
+ scx_task_iter_stop(&sti);
+
+ scx_enabling_sub_sched = NULL;
scx_cgroup_unlock();
percpu_up_write(&scx_fork_rwsem);
+ scx_bypass(sch, false);
+
pr_info("sched_ext: BPF sub-scheduler \"%s\" enabled\n", sch->ops.name);
kobject_uevent(&sch->kobj, KOBJ_ADD);
ret = 0;
goto out_unlock;
+exit_tasks:
+ scx_task_iter_start(&sti, sch->cgrp);
+ while ((p = scx_task_iter_next_locked(&sti))) {
+ if (p->scx.flags & SCX_TASK_SUB_INIT) {
+ __scx_disable_and_exit_task(sch, p);
+ p->scx.flags &= ~SCX_TASK_SUB_INIT;
+ }
+ }
+ scx_task_iter_stop(&sti);
out_put_cgrp:
cgroup_put(cgrp);
out_unlock:
@@ -5346,6 +5614,7 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
return ret;
err_unlock_and_disable:
+ /* we'll soon enter disable path, keep bypass on */
scx_cgroup_unlock();
percpu_up_write(&scx_fork_rwsem);
err_disable:
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 41/46] HACK_NOT_FOR_UPSTREAM: sched_ext: Work around @aux__prog prototype mismatch
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (39 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 40/46] sched_ext: Implement cgroup sub-sched enabling and disabling Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 42/46] sched_ext: Wrap global DSQs in per-node structure Tejun Heo
` (4 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Hide them using macros. This breaks binary compatibility. There should be a
better way.
NOT_SIGNED_OFF
---
tools/sched_ext/include/scx/common.bpf.h | 88 ++++++++++++++++--------
tools/sched_ext/include/scx/compat.bpf.h | 7 +-
tools/sched_ext/scx_qmap.bpf.c | 6 +-
3 files changed, 69 insertions(+), 32 deletions(-)
diff --git a/tools/sched_ext/include/scx/common.bpf.h b/tools/sched_ext/include/scx/common.bpf.h
index 06e2551033cb..6bf75f970237 100644
--- a/tools/sched_ext/include/scx/common.bpf.h
+++ b/tools/sched_ext/include/scx/common.bpf.h
@@ -58,54 +58,86 @@ static inline void ___vmlinux_h_sanity_check___(void)
"bpftool generated vmlinux.h is missing high bits for 64bit enums, upgrade clang and pahole");
}
-s32 scx_bpf_create_dsq(u64 dsq_id, s32 node) __ksym;
+/*
+ * XXX TEMPORARY - Each macro that has the same name as the preceding kfunc is
+ * to work around aux__prog issue. This causes a whole bunch of other problems.
+ * This is to work around just enough for testing and verification for now.
+ */
+s32 scx_bpf_create_dsq(u64 dsq_id, s32 node, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_create_dsq(dsq_id, node) scx_bpf_create_dsq((dsq_id), (node), NULL)
s32 scx_bpf_select_cpu_dfl(struct task_struct *p, s32 prev_cpu, u64 wake_flags, bool *is_idle) __ksym;
s32 scx_bpf_select_cpu_and(struct task_struct *p, s32 prev_cpu, u64 wake_flags,
const struct cpumask *cpus_allowed, u64 flags) __ksym __weak;
-void scx_bpf_dsq_insert(struct task_struct *p, u64 dsq_id, u64 slice, u64 enq_flags) __ksym __weak;
+void scx_bpf_dsq_insert(struct task_struct *p, u64 dsq_id, u64 slice, u64 enq_flags,
+ const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_dsq_insert(p, dsq_id, slice, enq_flags) scx_bpf_dsq_insert((p), (dsq_id), (slice), (enq_flags), NULL)
void scx_bpf_dsq_insert_vtime(struct task_struct *p, u64 dsq_id, u64 slice, u64 vtime, u64 enq_flags) __ksym __weak;
-u32 scx_bpf_dispatch_nr_slots(void) __ksym;
-void scx_bpf_dispatch_cancel(void) __ksym;
-bool scx_bpf_dsq_move_to_local(u64 dsq_id) __ksym __weak;
+u32 scx_bpf_dispatch_nr_slots(const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_dispatch_nr_slots() scx_bpf_dispatch_nr_slots(NULL)
+void scx_bpf_dispatch_cancel(const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_dispatch_cancel() scx_bpf_dispatch_cancel(NULL)
+bool scx_bpf_dsq_move_to_local(u64 dsq_id, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_dsq_move_to_local(dsq_id) scx_bpf_dsq_move_to_local(dsq_id, NULL)
void scx_bpf_dsq_move_set_slice(struct bpf_iter_scx_dsq *it__iter, u64 slice) __ksym __weak;
void scx_bpf_dsq_move_set_vtime(struct bpf_iter_scx_dsq *it__iter, u64 vtime) __ksym __weak;
bool scx_bpf_dsq_move(struct bpf_iter_scx_dsq *it__iter, struct task_struct *p, u64 dsq_id, u64 enq_flags) __ksym __weak;
bool scx_bpf_dsq_move_vtime(struct bpf_iter_scx_dsq *it__iter, struct task_struct *p, u64 dsq_id, u64 enq_flags) __ksym __weak;
-u32 scx_bpf_reenqueue_local(void) __ksym;
-void scx_bpf_kick_cpu(s32 cpu, u64 flags) __ksym;
+u32 scx_bpf_reenqueue_local(const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_reenqueue_local() scx_bpf_reenqueue_local(NULL)
+void scx_bpf_kick_cpu(s32 cpu, u64 flags, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_kick_cpu(cpu, flags) scx_bpf_kick_cpu((cpu), (flags), NULL)
s32 scx_bpf_dsq_nr_queued(u64 dsq_id) __ksym;
void scx_bpf_destroy_dsq(u64 dsq_id) __ksym;
-int bpf_iter_scx_dsq_new(struct bpf_iter_scx_dsq *it, u64 dsq_id, u64 flags) __ksym __weak;
+int bpf_iter_scx_dsq_new(struct bpf_iter_scx_dsq *it, u64 dsq_id, u64 flags, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+//#define bpf_iter_scx_dsq_new(it, dsq_id, flags) bpf_iter_scx_dsq_new((it), (dsq_id), (flags))
struct task_struct *bpf_iter_scx_dsq_next(struct bpf_iter_scx_dsq *it) __ksym __weak;
void bpf_iter_scx_dsq_destroy(struct bpf_iter_scx_dsq *it) __ksym __weak;
-void scx_bpf_exit_bstr(s64 exit_code, char *fmt, unsigned long long *data, u32 data__sz) __ksym __weak;
-void scx_bpf_error_bstr(char *fmt, unsigned long long *data, u32 data_len) __ksym;
-void scx_bpf_dump_bstr(char *fmt, unsigned long long *data, u32 data_len) __ksym __weak;
-u32 scx_bpf_cpuperf_cap(s32 cpu) __ksym __weak;
-u32 scx_bpf_cpuperf_cur(s32 cpu) __ksym __weak;
-void scx_bpf_cpuperf_set(s32 cpu, u32 perf) __ksym __weak;
+void scx_bpf_exit_bstr(s64 exit_code, char *fmt, unsigned long long *data, u32 data__sz, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_exit_bstr(exit_code, fmt, data, data__sz) scx_bpf_exit_bstr((exit_code), (fmt), (data), (data__sz), NULL)
+void scx_bpf_error_bstr(char *fmt, unsigned long long *data, u32 data__sz, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_error_bstr(fmt, data, data__sz) scx_bpf_error_bstr((fmt), (data), (data__sz), NULL)
+void scx_bpf_dump_bstr(char *fmt, unsigned long long *data, u32 data__sz, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_dump_bstr(fmt, data, data__sz) scx_bpf_dump_bstr((fmt), (data), (data__sz), NULL)
+u32 scx_bpf_cpuperf_cap(s32 cpu, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_cpuperf_cap(cpu) scx_bpf_cpuperf_cap((cpu), NULL)
+u32 scx_bpf_cpuperf_cur(s32 cpu, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_cpuperf_cur(cpu) scx_bpf_cpuperf_cur((cpu), NULL)
+void scx_bpf_cpuperf_set(s32 cpu, u32 perf, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_cpuperf_set(cpu, perf) scx_bpf_cpuperf_set((cpu), (perf), NULL)
u32 scx_bpf_nr_node_ids(void) __ksym __weak;
u32 scx_bpf_nr_cpu_ids(void) __ksym __weak;
-int scx_bpf_cpu_node(s32 cpu) __ksym __weak;
+int scx_bpf_cpu_node(s32 cpu, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_cpu_node(cpu) scx_bpf_cpu_node((cpu), NULL)
const struct cpumask *scx_bpf_get_possible_cpumask(void) __ksym __weak;
const struct cpumask *scx_bpf_get_online_cpumask(void) __ksym __weak;
void scx_bpf_put_cpumask(const struct cpumask *cpumask) __ksym __weak;
-const struct cpumask *scx_bpf_get_idle_cpumask_node(int node) __ksym __weak;
-const struct cpumask *scx_bpf_get_idle_cpumask(void) __ksym;
-const struct cpumask *scx_bpf_get_idle_smtmask_node(int node) __ksym __weak;
-const struct cpumask *scx_bpf_get_idle_smtmask(void) __ksym;
-void scx_bpf_put_idle_cpumask(const struct cpumask *cpumask) __ksym;
-bool scx_bpf_test_and_clear_cpu_idle(s32 cpu) __ksym;
-s32 scx_bpf_pick_idle_cpu_node(const cpumask_t *cpus_allowed, int node, u64 flags) __ksym __weak;
-s32 scx_bpf_pick_idle_cpu(const cpumask_t *cpus_allowed, u64 flags) __ksym;
-s32 scx_bpf_pick_any_cpu_node(const cpumask_t *cpus_allowed, int node, u64 flags) __ksym __weak;
-s32 scx_bpf_pick_any_cpu(const cpumask_t *cpus_allowed, u64 flags) __ksym;
+const struct cpumask *scx_bpf_get_idle_cpumask_node(int node, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_get_idle_cpumask_node(node) scx_bpf_get_idle_cpumask_node((node), NULL)
+const struct cpumask *scx_bpf_get_idle_cpumask(const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_get_idle_cpumask() scx_bpf_get_idle_cpumask(NULL)
+const struct cpumask *scx_bpf_get_idle_smtmask_node(int node, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_get_idle_smtmask_node(node) scx_bpf_get_idle_smtmask_node((node))
+const struct cpumask *scx_bpf_get_idle_smtmask(const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_get_idle_smtmask() scx_bpf_get_idle_smtmask(NULL)
+void scx_bpf_put_idle_cpumask(const struct cpumask *cpumask, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_put_idle_cpumask(cpumask) scx_bpf_put_idle_cpumask((cpumask), NULL)
+bool scx_bpf_test_and_clear_cpu_idle(s32 cpu, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_test_and_clear_cpu_idle(cpu) scx_bpf_test_and_clear_cpu_idle((cpu), NULL)
+s32 scx_bpf_pick_idle_cpu_node(const cpumask_t *cpus_allowed, int node, u64 flags, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_pick_idle_cpu_node(cpus_allowed, node, flags) scx_bpf_pick_idle_cpu_node((cpus_allowed), (node), (flags), NULL)
+s32 scx_bpf_pick_idle_cpu(const cpumask_t *cpus_allowed, u64 flags, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_pick_idle_cpu(cpus_allowed, flags) scx_bpf_pick_idle_cpu((cpus_allowed), (flags), NULL)
+s32 scx_bpf_pick_any_cpu_node(const cpumask_t *cpus_allowed, int node, u64 flags, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_pick_any_cpu_node(cpus_allowed, node, flags) scx_bpf_pick_any_cpu_node((cpus_allowed), (node), (flags), NULL)
+s32 scx_bpf_pick_any_cpu(const cpumask_t *cpus_allowed, u64 flags, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_pick_any_cpu(cpus_allowed, flags) scx_bpf_pick_any_cpu((cpus_allowed), (flags), NULL)
bool scx_bpf_task_running(const struct task_struct *p) __ksym;
s32 scx_bpf_task_cpu(const struct task_struct *p) __ksym;
-struct rq *scx_bpf_cpu_rq(s32 cpu) __ksym;
-struct rq *scx_bpf_locked_rq(void) __ksym;
+struct rq *scx_bpf_cpu_rq(s32 cpu, const struct bpf_prog_aux *aux__prog) __ksym;
+#define scx_bpf_cpu_rq(cpu) scx_bpf_cpu_rq((cpu), NULL)
struct task_struct *scx_bpf_cpu_curr(s32 cpu) __ksym __weak;
-struct cgroup *scx_bpf_task_cgroup(struct task_struct *p) __ksym __weak;
+struct cgroup *scx_bpf_task_cgroup(struct task_struct *p, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_task_cgroup(p) scx_bpf_task_cgroup((p), NULL)
u64 scx_bpf_now(void) __ksym __weak;
void scx_bpf_events(struct scx_event_stats *events, size_t events__sz) __ksym __weak;
diff --git a/tools/sched_ext/include/scx/compat.bpf.h b/tools/sched_ext/include/scx/compat.bpf.h
index dd9144624dc9..b84a3d42d2aa 100644
--- a/tools/sched_ext/include/scx/compat.bpf.h
+++ b/tools/sched_ext/include/scx/compat.bpf.h
@@ -40,6 +40,11 @@ bool scx_bpf_dispatch_from_dsq___compat(struct bpf_iter_scx_dsq *it__iter, struc
bool scx_bpf_dispatch_vtime_from_dsq___compat(struct bpf_iter_scx_dsq *it__iter, struct task_struct *p, u64 dsq_id, u64 enq_flags) __ksym __weak;
int bpf_cpumask_populate(struct cpumask *dst, void *src, size_t src__sz) __ksym __weak;
+/*
+ * XXX TEMPORARY - The followings conflict with prog__aux wrappers. Comment them
+ * out for now.
+ */
+/*
#define scx_bpf_dsq_insert(p, dsq_id, slice, enq_flags) \
(bpf_ksym_exists(scx_bpf_dsq_insert) ? \
scx_bpf_dsq_insert((p), (dsq_id), (slice), (enq_flags)) : \
@@ -54,7 +59,7 @@ int bpf_cpumask_populate(struct cpumask *dst, void *src, size_t src__sz) __ksym
(bpf_ksym_exists(scx_bpf_dsq_move_to_local) ? \
scx_bpf_dsq_move_to_local((dsq_id)) : \
scx_bpf_consume___compat((dsq_id)))
-
+*/
#define __COMPAT_scx_bpf_dsq_move_set_slice(it__iter, slice) \
(bpf_ksym_exists(scx_bpf_dsq_move_set_slice) ? \
scx_bpf_dsq_move_set_slice((it__iter), (slice)) : \
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 9631e4d04d88..15e15cb234dc 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -312,7 +312,7 @@ static bool dispatch_highpri(bool from_timer)
s32 this_cpu = bpf_get_smp_processor_id();
/* scan SHARED_DSQ and move highpri tasks to HIGHPRI_DSQ */
- bpf_for_each(scx_dsq, p, SHARED_DSQ, 0) {
+ bpf_for_each(scx_dsq, p, SHARED_DSQ, 0, NULL) {
static u64 highpri_seq;
struct task_ctx *tctx;
@@ -334,7 +334,7 @@ static bool dispatch_highpri(bool from_timer)
* Scan HIGHPRI_DSQ and dispatch until a task that can run on this CPU
* is found.
*/
- bpf_for_each(scx_dsq, p, HIGHPRI_DSQ, 0) {
+ bpf_for_each(scx_dsq, p, HIGHPRI_DSQ, 0, NULL) {
bool dispatched = false;
s32 cpu;
@@ -801,7 +801,7 @@ static void dump_shared_dsq(void)
bpf_printk("Dumping %d tasks in SHARED_DSQ in reverse order", nr);
bpf_rcu_read_lock();
- bpf_for_each(scx_dsq, p, SHARED_DSQ, SCX_DSQ_ITER_REV)
+ bpf_for_each(scx_dsq, p, SHARED_DSQ, SCX_DSQ_ITER_REV, NULL)
bpf_printk("%s[%d]", p->comm, p->pid);
bpf_rcu_read_unlock();
}
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 42/46] sched_ext: Wrap global DSQs in per-node structure
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (40 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 41/46] HACK_NOT_FOR_UPSTREAM: sched_ext: Work around @aux__prog prototype mismatch Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 43/46] sched_ext: Add bypass DSQ for sub-schedulers Tejun Heo
` (3 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Currently, global DSQs are stored as an array of pointers to
scx_dispatch_q structs, one per NUMA node. This is limiting for future
enhancements that may require additional per-node data structures.
Introduce scx_sched_pnode structure to wrap the global DSQ. This change
replaces the global_dsqs array with a pnode array of scx_sched_pnode
structs and updates all references to access global_dsq through the
pnode wrapper.
This refactoring maintains NUMA-aware allocation and provides foundation
for adding more per-node data structures with no functional impact.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 33 ++++++++++++++++-----------------
kernel/sched/ext_internal.h | 6 +++++-
2 files changed, 21 insertions(+), 18 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index eff5f6894f14..a74ae955c489 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -274,7 +274,7 @@ static bool scx_is_descendant(struct scx_sched *sch, struct scx_sched *ancestor)
static struct scx_dispatch_q *find_global_dsq(struct scx_sched *sch,
struct task_struct *p)
{
- return sch->global_dsqs[cpu_to_node(task_cpu(p))];
+ return &sch->pnode[cpu_to_node(task_cpu(p))]->global_dsq;
}
static struct scx_dispatch_q *find_user_dsq(struct scx_sched *sch, u64 dsq_id)
@@ -1965,7 +1965,7 @@ static bool consume_global_dsq(struct scx_sched *sch, struct rq *rq)
{
int node = cpu_to_node(cpu_of(rq));
- return consume_dispatch_q(sch, rq, sch->global_dsqs[node]);
+ return consume_dispatch_q(sch, rq, &sch->pnode[node]->global_dsq);
}
/**
@@ -3733,8 +3733,8 @@ static void scx_sched_free_rcu_work(struct work_struct *work)
free_percpu(sch->pcpu);
for_each_node_state(node, N_POSSIBLE)
- kfree(sch->global_dsqs[node]);
- kfree(sch->global_dsqs);
+ kfree(sch->pnode[node]);
+ kfree(sch->pnode);
rhashtable_walk_enter(&sch->dsq_hash, &rht_iter);
do {
@@ -4935,24 +4935,23 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
if (ret < 0)
goto err_free_ei;
- sch->global_dsqs = kcalloc(nr_node_ids, sizeof(sch->global_dsqs[0]),
- GFP_KERNEL);
- if (!sch->global_dsqs) {
+ sch->pnode = kcalloc(nr_node_ids, sizeof(sch->pnode[0]), GFP_KERNEL);
+ if (!sch->pnode) {
ret = -ENOMEM;
goto err_free_hash;
}
for_each_node_state(node, N_POSSIBLE) {
- struct scx_dispatch_q *dsq;
+ struct scx_sched_pnode *pnode;
- dsq = kzalloc_node(sizeof(*dsq), GFP_KERNEL, node);
- if (!dsq) {
+ pnode = kzalloc_node(sizeof(*pnode), GFP_KERNEL, node);
+ if (!pnode) {
ret = -ENOMEM;
- goto err_free_gdsqs;
+ goto err_free_pnode;
}
- init_dsq(dsq, SCX_DSQ_GLOBAL, sch);
- sch->global_dsqs[node] = dsq;
+ init_dsq(&pnode->global_dsq, SCX_DSQ_GLOBAL, sch);
+ sch->pnode[node] = pnode;
}
sch->dsp_max_batch = ops->dispatch_max_batch ?: SCX_DSP_DFL_MAX_BATCH;
@@ -4960,7 +4959,7 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
dsp_ctx.buf, sch->dsp_max_batch),
__alignof__(struct scx_sched_pcpu));
if (!sch->pcpu)
- goto err_free_gdsqs;
+ goto err_free_pnode;
sch->helper = kthread_run_worker(0, "sched_ext_helper");
if (!sch->helper)
@@ -5031,10 +5030,10 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
kthread_stop(sch->helper->task);
err_free_pcpu:
free_percpu(sch->pcpu);
-err_free_gdsqs:
+err_free_pnode:
for_each_node_state(node, N_POSSIBLE)
- kfree(sch->global_dsqs[node]);
- kfree(sch->global_dsqs);
+ kfree(sch->pnode[node]);
+ kfree(sch->pnode);
err_free_hash:
rhashtable_free_and_destroy(&sch->dsq_hash, NULL, NULL);
err_free_ei:
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 4399c003c15f..846855ea5948 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -941,6 +941,10 @@ struct scx_sched_pcpu {
struct scx_dsp_ctx dsp_ctx;
};
+struct scx_sched_pnode {
+ struct scx_dispatch_q global_dsq;
+};
+
struct scx_sched {
struct sched_ext_ops ops;
DECLARE_BITMAP(has_op, SCX_OPI_END);
@@ -954,7 +958,7 @@ struct scx_sched {
* per-node split isn't sufficient, it can be further split.
*/
struct rhashtable dsq_hash;
- struct scx_dispatch_q **global_dsqs;
+ struct scx_sched_pnode **pnode;
struct scx_sched_pcpu __percpu *pcpu;
s32 bypass_depth;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 43/46] sched_ext: Add bypass DSQ for sub-schedulers
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (41 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 42/46] sched_ext: Wrap global DSQs in per-node structure Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 44/46] sched_ext: Factor out scx_link_sched() and scx_unlink_sched() Tejun Heo
` (2 subsequent siblings)
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
When sub-schedulers bypass, we want to isolate their impact on the
system while ensuring forward progress. Add a bypass DSQ to route tasks
from bypassing sub-schedulers to their nearest non-bypassing ancestor.
Each per-node structure gets a bypass DSQ (SCX_DSQ_SUB_BYPASS). Tasks
from bypassing descendants are collected there and consumed with a
simple alternating policy between regular dispatch and bypass DSQ.
Add SCX_EV_SUB_BYPASS_DISPATCH event counter to track bypass DSQ usage
for observability.
In the future, we can add ops flags and kfuncs to allow BPF schedulers
to control bypass scheduling if needed.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
include/linux/sched/ext.h | 2 +
kernel/sched/ext.c | 91 +++++++++++++++++++++++++++++++++----
kernel/sched/ext_internal.h | 9 ++++
kernel/sched/sched.h | 3 ++
4 files changed, 95 insertions(+), 10 deletions(-)
diff --git a/include/linux/sched/ext.h b/include/linux/sched/ext.h
index df1111d245bc..2f46de38966d 100644
--- a/include/linux/sched/ext.h
+++ b/include/linux/sched/ext.h
@@ -48,6 +48,8 @@ enum scx_dsq_id_flags {
SCX_DSQ_LOCAL = SCX_DSQ_FLAG_BUILTIN | 2,
SCX_DSQ_LOCAL_ON = SCX_DSQ_FLAG_BUILTIN | SCX_DSQ_FLAG_LOCAL_ON,
SCX_DSQ_LOCAL_CPU_MASK = 0xffffffffLLU,
+
+ SCX_DSQ_SUB_BYPASS = SCX_DSQ_FLAG_BUILTIN | 3,
};
/*
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index a74ae955c489..4558bec72508 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -282,6 +282,42 @@ static struct scx_dispatch_q *find_user_dsq(struct scx_sched *sch, u64 dsq_id)
return rhashtable_lookup(&sch->dsq_hash, &dsq_id, dsq_hash_params);
}
+static struct scx_dispatch_q *find_bypass_dsq(struct scx_sched *sch,
+ struct rq *rq,
+ struct task_struct *p)
+{
+#ifdef CONFIG_EXT_SUB_SCHED
+ struct scx_sched *pos;
+
+ /*
+ * If @sch is a sub-sched which is bypassing, its tasks should go into
+ * the bypass_dsq of the nearest ancestor which is not bypassing. The
+ * not-bypassing sched is responsible for scheduling all tasks from
+ * bypassing sub-trees. If all ancestors including root are bypassing,
+ * there is no one home and @p should go to the root's global_dsq.
+ *
+ * Whenever a sched starts bypassing, all runnable tasks in its subtree
+ * are re-enqueued after scx_bypassing() is turned on, guaranteeing that
+ * all tasks are transferred to the right DSQ.
+ */
+ pos = scx_parent(sch);
+ if (pos) {
+ while (true) {
+ if (!scx_bypassing(pos, cpu_of(rq))) {
+ int node = cpu_to_node(task_cpu(p));
+ return &pos->pnode[node]->sub_bypass_dsq;
+ }
+ if (pos->level == 0) {
+ sch = pos;
+ break;
+ }
+ pos = scx_parent(pos);
+ }
+ }
+#endif
+ return find_global_dsq(sch, p);
+}
+
/*
* scx_kf_mask enforcement. Some kfuncs can only be called from specific SCX
* ops. When invoking SCX ops, SCX_CALL_OP[_RET]() should be used to indicate
@@ -1353,6 +1389,7 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
int sticky_cpu)
{
struct scx_sched *sch = scx_task_sched(p);
+ struct scx_dispatch_q *fallback_dsq;
struct task_struct **ddsp_taskp;
unsigned long qseq;
@@ -1372,7 +1409,8 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
if (scx_bypassing(sch, cpu_of(rq))) {
__scx_add_event(sch, SCX_EV_BYPASS_DISPATCH, 1);
- goto global;
+ fallback_dsq = find_bypass_dsq(sch, rq, p);
+ goto fallback;
}
if (p->scx.ddsp_dsq_id != SCX_DSQ_INVALID)
@@ -1392,8 +1430,10 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
goto local;
}
- if (unlikely(!SCX_HAS_OP(sch, enqueue)))
- goto global;
+ if (unlikely(!SCX_HAS_OP(sch, enqueue))) {
+ fallback_dsq = find_global_dsq(sch, p);
+ goto fallback;
+ }
/* DSQ bypass didn't trigger, enqueue on the BPF scheduler */
qseq = rq->scx.ops_qseq++ << SCX_OPSS_QSEQ_SHIFT;
@@ -1434,10 +1474,10 @@ static void do_enqueue_task(struct rq *rq, struct task_struct *p, u64 enq_flags,
dispatch_enqueue(sch, &rq->scx.local_dsq, p, enq_flags);
return;
-global:
+fallback:
touch_core_sched(rq, p); /* see the comment in local: */
refill_task_slice_dfl(sch, p);
- dispatch_enqueue(sch, find_global_dsq(sch, p), p, enq_flags);
+ dispatch_enqueue(sch, fallback_dsq, p, enq_flags);
}
static bool task_runnable(const struct task_struct *p)
@@ -1961,11 +2001,36 @@ static bool consume_dispatch_q(struct scx_sched *sch, struct rq *rq,
return false;
}
-static bool consume_global_dsq(struct scx_sched *sch, struct rq *rq)
+static bool consume_fallback_dsqs(struct scx_sched *sch, struct rq *rq)
{
- int node = cpu_to_node(cpu_of(rq));
+ struct scx_sched_pnode *pnode = sch->pnode[cpu_to_node(cpu_of(rq))];
- return consume_dispatch_q(sch, rq, &sch->pnode[node]->global_dsq);
+ if (consume_dispatch_q(sch, rq, &pnode->global_dsq))
+ return true;
+
+#ifdef CONFIG_EXT_SUB_SCHED
+ /*
+ * @sch must ensure forward progress of its bypassing descendants that
+ * dump their tasks into @sch's bypass DSQs. The following implements a
+ * simple built-in behavior - let each CPU toggle between regular
+ * dispatch and the bypass DSQ.
+ *
+ * Later, if necessary, we can add an ops flag to suppress the
+ * auto-consumption and a kfunc to consume the bypass DSQ and, so that
+ * the BPF scheduler can fully control scheduling of bypassed tasks.
+ */
+ if (rq->scx.dsp_sub_bypass_toggle) {
+ if (consume_dispatch_q(sch, rq, &pnode->sub_bypass_dsq)) {
+ __scx_add_event(sch, SCX_EV_SUB_BYPASS_DISPATCH, 1);
+ rq->scx.dsp_sub_bypass_toggle = false;
+ return true;
+ }
+ } else {
+ rq->scx.dsp_sub_bypass_toggle = true;
+ }
+#endif /* CONFIG_EXT_SUB_SCHED */
+
+ return false;
}
/**
@@ -2175,7 +2240,7 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
bool prev_on_sch = (prev->sched_class == &ext_sched_class) &&
sch == rcu_access_pointer(prev->scx.sched);
- if (consume_global_dsq(sch, rq))
+ if (consume_fallback_dsqs(sch, rq))
return true;
if (unlikely(!SCX_HAS_OP(sch, dispatch)) ||
@@ -2205,7 +2270,7 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
}
if (rq->scx.local_dsq.nr)
return true;
- if (consume_global_dsq(sch, rq))
+ if (consume_fallback_dsqs(sch, rq))
return true;
/*
@@ -3789,6 +3854,7 @@ static ssize_t scx_attr_events_show(struct kobject *kobj,
at += scx_attr_event_show(buf, at, &events, SCX_EV_BYPASS_DISPATCH);
at += scx_attr_event_show(buf, at, &events, SCX_EV_BYPASS_ACTIVATE);
at += scx_attr_event_show(buf, at, &events, SCX_EV_INSERT_NOT_OWNED);
+ at += scx_attr_event_show(buf, at, &events, SCX_EV_SUB_BYPASS_DISPATCH);
return at;
}
SCX_ATTR(events);
@@ -4870,6 +4936,7 @@ static void scx_dump_state(struct scx_sched *sch, struct scx_exit_info *ei,
scx_dump_event(s, &events, SCX_EV_BYPASS_DISPATCH);
scx_dump_event(s, &events, SCX_EV_BYPASS_ACTIVATE);
scx_dump_event(s, &events, SCX_EV_INSERT_NOT_OWNED);
+ scx_dump_event(s, &events, SCX_EV_SUB_BYPASS_DISPATCH);
if (seq_buf_has_overflowed(&s) && dump_len >= sizeof(trunc_marker))
memcpy(ei->dump + dump_len - sizeof(trunc_marker),
@@ -4951,6 +5018,9 @@ static struct scx_sched *scx_alloc_and_add_sched(struct sched_ext_ops *ops,
}
init_dsq(&pnode->global_dsq, SCX_DSQ_GLOBAL, sch);
+#ifdef CONFIG_EXT_SUB_SCHED
+ init_dsq(&pnode->sub_bypass_dsq, SCX_DSQ_SUB_BYPASS, sch);
+#endif /* CONFIG_EXT_SUB_SCHED */
sch->pnode[node] = pnode;
}
@@ -7623,6 +7693,7 @@ static void scx_read_events(struct scx_sched *sch, struct scx_event_stats *event
scx_agg_event(events, e_cpu, SCX_EV_BYPASS_DISPATCH);
scx_agg_event(events, e_cpu, SCX_EV_BYPASS_ACTIVATE);
scx_agg_event(events, e_cpu, SCX_EV_INSERT_NOT_OWNED);
+ scx_agg_event(events, e_cpu, SCX_EV_SUB_BYPASS_DISPATCH);
}
}
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 846855ea5948..f6d5867230bd 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -907,6 +907,12 @@ struct scx_event_stats {
* scheduler.
*/
s64 SCX_EV_INSERT_NOT_OWNED;
+
+ /*
+ * The number of times tasks from bypassing descendants are scheduled
+ * from sub_bypass_dsq's.
+ */
+ s64 SCX_EV_SUB_BYPASS_DISPATCH;
};
enum scx_sched_pcpu_flags {
@@ -943,6 +949,9 @@ struct scx_sched_pcpu {
struct scx_sched_pnode {
struct scx_dispatch_q global_dsq;
+#ifdef CONFIG_EXT_SUB_SCHED
+ struct scx_dispatch_q sub_bypass_dsq;
+#endif /* CONFIG_EXT_SUB_SCHED */
};
struct scx_sched {
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index add7c0c218d4..cd6bdcdf9314 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -780,6 +780,9 @@ struct scx_rq {
struct balance_callback deferred_bal_cb;
struct irq_work deferred_irq_work;
struct irq_work kick_cpus_irq_work;
+#ifdef CONFIG_EXT_SUB_SCHED
+ bool dsp_sub_bypass_toggle;
+#endif /* CONFIG_EXT_SUB_SCHED */
};
#endif /* CONFIG_SCHED_CLASS_EXT */
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 44/46] sched_ext: Factor out scx_link_sched() and scx_unlink_sched()
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (42 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 43/46] sched_ext: Add bypass DSQ for sub-schedulers Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 45/46] sched_ext: Add rhashtable lookup for sub-schedulers Tejun Heo
2025-09-20 0:59 ` [PATCH 46/46] sched_ext: Add basic building blocks for nested sub-scheduler dispatching Tejun Heo
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Factor out scx_link_sched() and scx_unlink_sched() functions to reduce
code duplication in the scheduler enable/disable paths.
No functional change.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 53 +++++++++++++++++++++++++++-------------------
1 file changed, 31 insertions(+), 22 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 4558bec72508..058315fc524b 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -4221,6 +4221,33 @@ static void refresh_watchdog(void)
cancel_delayed_work_sync(&scx_watchdog_work);
}
+static void scx_link_sched(struct scx_sched *sch)
+{
+ scoped_guard(raw_spinlock_irq, &scx_sched_lock) {
+#ifdef CONFIG_EXT_SUB_SCHED
+ struct scx_sched *parent = scx_parent(sch);
+ if (parent)
+ list_add_tail(&sch->sibling, &parent->children);
+#endif /* CONFIG_EXT_SUB_SCHED */
+ list_add_tail_rcu(&sch->all, &scx_sched_all);
+ }
+
+ refresh_watchdog();
+}
+
+static void scx_unlink_sched(struct scx_sched *sch)
+{
+ scoped_guard(raw_spinlock_irq, &scx_sched_lock) {
+#ifdef CONFIG_EXT_SUB_SCHED
+ if (scx_parent(sch))
+ list_del_init(&sch->sibling);
+#endif /* CONFIG_EXT_SUB_SCHED */
+ list_del_rcu(&sch->all);
+ }
+
+ refresh_watchdog();
+}
+
#ifdef CONFIG_EXT_SUB_SCHED
static DECLARE_WAIT_QUEUE_HEAD(scx_unlink_waitq);
@@ -4385,12 +4412,7 @@ static void scx_sub_disable(struct scx_sched *sch)
*/
synchronize_rcu_expedited();
- raw_spin_lock_irq(&scx_sched_lock);
- list_del_init(&sch->sibling);
- list_del_rcu(&sch->all);
- raw_spin_unlock_irq(&scx_sched_lock);
-
- refresh_watchdog();
+ scx_unlink_sched(sch);
mutex_unlock(&scx_enable_mutex);
@@ -4549,11 +4571,7 @@ static void scx_root_disable(struct scx_sched *sch)
if (sch->ops.exit)
SCX_CALL_OP(sch, SCX_KF_UNLOCKED, exit, NULL, ei);
- raw_spin_lock_irq(&scx_sched_lock);
- list_del_rcu(&sch->all);
- raw_spin_unlock_irq(&scx_sched_lock);
-
- refresh_watchdog();
+ scx_unlink_sched(sch);
/*
* scx_root clearing must be inside cpus_read_lock(). See
@@ -5215,11 +5233,7 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
*/
rcu_assign_pointer(scx_root, sch);
- raw_spin_lock_irq(&scx_sched_lock);
- list_add_tail_rcu(&sch->all, &scx_sched_all);
- raw_spin_unlock_irq(&scx_sched_lock);
-
- refresh_watchdog();
+ scx_link_sched(sch);
scx_idle_enable(ops);
@@ -5486,12 +5500,7 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
goto out_put_cgrp;
}
- raw_spin_lock_irq(&scx_sched_lock);
- list_add_tail(&sch->sibling, &parent->children);
- list_add_tail_rcu(&sch->all, &scx_sched_all);
- raw_spin_unlock_irq(&scx_sched_lock);
-
- refresh_watchdog();
+ scx_link_sched(sch);
if (sch->level >= SCX_SUB_MAX_DEPTH) {
scx_error(sch, "max nesting depth %d violated",
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 45/46] sched_ext: Add rhashtable lookup for sub-schedulers
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (43 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 44/46] sched_ext: Factor out scx_link_sched() and scx_unlink_sched() Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
2025-09-20 0:59 ` [PATCH 46/46] sched_ext: Add basic building blocks for nested sub-scheduler dispatching Tejun Heo
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
Add rhashtable-based lookup for sub-schedulers indexed by cgroup_id to
enable efficient scheduler discovery in preparation for multiple scheduler
support. The hash table allows quick lookup of the appropriate scheduler
instance when processing tasks from different cgroups.
This extends scx_link_sched() to register sub-schedulers in the hash table
and scx_unlink_sched() to remove them. A new scx_find_sub_sched() function
provides the lookup interface.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 54 +++++++++++++++++++++++++++++++++----
kernel/sched/ext_internal.h | 2 ++
2 files changed, 51 insertions(+), 5 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 058315fc524b..0d865e017115 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -27,6 +27,16 @@ struct scx_sched __rcu *scx_root;
*/
static LIST_HEAD(scx_sched_all);
+#ifdef CONFIG_EXT_SUB_SCHED
+static const struct rhashtable_params scx_sched_hash_params = {
+ .key_len = sizeof_field(struct scx_sched, ops.sub_cgroup_id),
+ .key_offset = offsetof(struct scx_sched, ops.sub_cgroup_id),
+ .head_offset = offsetof(struct scx_sched, hash_node),
+};
+
+static struct rhashtable scx_sched_hash;
+#endif
+
/*
* - We want to visit and perform sleepable operations on every task.
*
@@ -201,6 +211,12 @@ static struct scx_sched *scx_parent(struct scx_sched *sch)
}
#ifdef CONFIG_EXT_SUB_SCHED
+static struct scx_sched *scx_find_sub_sched(u64 cgroup_id)
+{
+ return rhashtable_lookup(&scx_sched_hash, &cgroup_id,
+ scx_sched_hash_params);
+}
+
static void scx_set_task_sched(struct task_struct *p, struct scx_sched *sch)
{
rcu_assign_pointer(p->scx.sched, sch);
@@ -233,6 +249,11 @@ static struct scx_sched *scx_next_descendant_pre(struct scx_sched *pos,
return NULL;
}
#else /* CONFIG_EXT_SUB_SCHED */
+static struct scx_sched *scx_find_sub_sched(u64 cgroup_id)
+{
+ return NULL;
+}
+
static void scx_set_task_sched(struct task_struct *p, struct scx_sched *sch)
{
}
@@ -4221,26 +4242,41 @@ static void refresh_watchdog(void)
cancel_delayed_work_sync(&scx_watchdog_work);
}
-static void scx_link_sched(struct scx_sched *sch)
+static int scx_link_sched(struct scx_sched *sch)
{
scoped_guard(raw_spinlock_irq, &scx_sched_lock) {
#ifdef CONFIG_EXT_SUB_SCHED
struct scx_sched *parent = scx_parent(sch);
- if (parent)
+ int ret;
+
+ if (parent) {
+ ret = rhashtable_lookup_insert_fast(&scx_sched_hash,
+ &sch->hash_node, scx_sched_hash_params);
+ if (ret) {
+ scx_error(sch, "failed to insert into scx_sched_hash (%d)", ret);
+ return ret;
+ }
+
list_add_tail(&sch->sibling, &parent->children);
+ }
#endif /* CONFIG_EXT_SUB_SCHED */
+
list_add_tail_rcu(&sch->all, &scx_sched_all);
}
refresh_watchdog();
+ return 0;
}
static void scx_unlink_sched(struct scx_sched *sch)
{
scoped_guard(raw_spinlock_irq, &scx_sched_lock) {
#ifdef CONFIG_EXT_SUB_SCHED
- if (scx_parent(sch))
+ if (scx_parent(sch)) {
+ rhashtable_remove_fast(&scx_sched_hash, &sch->hash_node,
+ scx_sched_hash_params);
list_del_init(&sch->sibling);
+ }
#endif /* CONFIG_EXT_SUB_SCHED */
list_del_rcu(&sch->all);
}
@@ -5233,7 +5269,9 @@ static int scx_root_enable(struct sched_ext_ops *ops, struct bpf_link *link)
*/
rcu_assign_pointer(scx_root, sch);
- scx_link_sched(sch);
+ ret = scx_link_sched(sch);
+ if (ret)
+ goto err_disable;
scx_idle_enable(ops);
@@ -5500,7 +5538,9 @@ static int scx_sub_enable(struct sched_ext_ops *ops, struct bpf_link *link)
goto out_put_cgrp;
}
- scx_link_sched(sch);
+ ret = scx_link_sched(sch);
+ if (ret)
+ goto err_disable;
if (sch->level >= SCX_SUB_MAX_DEPTH) {
scx_error(sch, "max nesting depth %d violated",
@@ -6287,6 +6327,10 @@ void __init init_sched_ext_class(void)
register_sysrq_key('S', &sysrq_sched_ext_reset_op);
register_sysrq_key('D', &sysrq_sched_ext_dump_op);
INIT_DELAYED_WORK(&scx_watchdog_work, scx_watchdog_workfn);
+
+#ifdef CONFIG_EXT_SUB_SCHED
+ BUG_ON(rhashtable_init(&scx_sched_hash, &scx_sched_hash_params));
+#endif /* CONFIG_EXT_SUB_SCHED */
}
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index f6d5867230bd..7a26c7c89e1f 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -980,6 +980,8 @@ struct scx_sched {
struct list_head all;
#ifdef CONFIG_EXT_SUB_SCHED
+ struct rhash_head hash_node;
+
struct list_head children;
struct list_head sibling;
struct cgroup *cgrp;
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread* [PATCH 46/46] sched_ext: Add basic building blocks for nested sub-scheduler dispatching
2025-09-20 0:58 [PATCHSET RFC] sched_ext: Implement cgroup sub-scheduler support Tejun Heo
` (44 preceding siblings ...)
2025-09-20 0:59 ` [PATCH 45/46] sched_ext: Add rhashtable lookup for sub-schedulers Tejun Heo
@ 2025-09-20 0:59 ` Tejun Heo
45 siblings, 0 replies; 47+ messages in thread
From: Tejun Heo @ 2025-09-20 0:59 UTC (permalink / raw)
To: void, arighi, multics69; +Cc: linux-kernel, sched-ext, memxor, bpf, Tejun Heo
This is an early-stage partial implementation that demonstrates the core
building blocks for nested sub-scheduler dispatching. While significant
work remains in the enqueue path and other areas, this patch establishes
the fundamental mechanisms needed for hierarchical scheduler operation.
The key building blocks introduced include:
- Private stack support for ops.dispatch() to prevent stack overflow when
walking down nested schedulers during dispatch operations
- scx_bpf_sub_dispatch() kfunc that allows parent schedulers to trigger
dispatch operations on their direct child schedulers
- Proper parent-child relationship validation to ensure dispatch requests
are only made to legitimate child schedulers
- Updated scx_dispatch_sched() to handle both nested and non-nested
invocations with appropriate kf_mask handling
The qmap scheduler is updated to demonstrate the functionality by calling
scx_bpf_sub_dispatch() on registered child schedulers when it has no
tasks in its own queues.
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/sched/ext.c | 116 ++++++++++++++++++++---
kernel/sched/sched.h | 3 +
tools/sched_ext/include/scx/common.bpf.h | 2 +
tools/sched_ext/scx_qmap.bpf.c | 37 +++++++-
4 files changed, 145 insertions(+), 13 deletions(-)
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index 0d865e017115..99462d0da543 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -2253,8 +2253,14 @@ static void flush_dispatch_buf(struct scx_sched *sch, struct rq *rq)
dspc->cursor = 0;
}
-static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
- struct task_struct *prev)
+/*
+ * One user of this function is scx_bpf_dispatch() which can be called
+ * recursively as sub-sched dispatches nest. Always inline to reduce stack usage
+ * from the call frame.
+ */
+static __always_inline bool
+scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
+ struct task_struct *prev, bool nested)
{
struct scx_dsp_ctx *dspc = &this_cpu_ptr(sch->pcpu)->dsp_ctx;
int nr_loops = SCX_DSP_MAX_LOOPS;
@@ -2280,8 +2286,23 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
do {
dspc->nr_tasks = 0;
- SCX_CALL_OP(sch, SCX_KF_DISPATCH, dispatch, rq,
- cpu_of(rq), prev_on_sch ? prev : NULL);
+ if (nested) {
+ /*
+ * If nested, don't update kf_mask as the originating
+ * invocation would already have set it up.
+ */
+ SCX_CALL_OP(sch, 0, dispatch, rq,
+ cpu_of(rq), prev_on_sch ? prev : NULL);
+ } else {
+ /*
+ * If not nested, stash @prev so that nested invocations
+ * can access it.
+ */
+ rq->scx.sub_dispatch_prev = prev;
+ SCX_CALL_OP(sch, SCX_KF_DISPATCH, dispatch, rq,
+ cpu_of(rq), prev_on_sch ? prev : NULL);
+ rq->scx.sub_dispatch_prev = NULL;
+ }
flush_dispatch_buf(sch, rq);
@@ -2314,7 +2335,7 @@ static bool scx_dispatch_sched(struct scx_sched *sch, struct rq *rq,
static int balance_one(struct rq *rq, struct task_struct *prev)
{
- struct scx_sched *sch = scx_root, *pos;
+ struct scx_sched *sch = scx_root;
lockdep_assert_rq_held(rq);
rq->scx.flags |= SCX_RQ_IN_BALANCE;
@@ -2358,13 +2379,8 @@ static int balance_one(struct rq *rq, struct task_struct *prev)
if (rq->scx.local_dsq.nr)
goto has_tasks;
- /*
- * TEMPORARY - Dispatch all scheds. This will be replaced by BPF-driven
- * hierarchical operation.
- */
- list_for_each_entry_rcu(pos, &scx_sched_all, all)
- if (scx_dispatch_sched(pos, rq, prev))
- goto has_tasks;
+ if (scx_dispatch_sched(sch, rq, prev, false))
+ goto has_tasks;
/*
* Didn't find another task to run. Keep running @prev unless
@@ -5883,6 +5899,20 @@ static int bpf_scx_init_member(const struct btf_type *t,
return 0;
}
+#ifdef CONFIG_EXT_SUB_SCHED
+static void scx_pstack_recursion_on_dispatch(struct bpf_prog *prog)
+{
+ struct scx_sched *sch;
+
+ guard(rcu)();
+ sch = scx_prog_sched(prog->aux);
+ if (unlikely(!sch))
+ return;
+
+ scx_error(sch, "dispatch recursion detected");
+}
+#endif /* CONFIG_EXT_SUB_SCHED */
+
static int bpf_scx_check_member(const struct btf_type *t,
const struct btf_member *member,
const struct bpf_prog *prog)
@@ -5908,6 +5938,22 @@ static int bpf_scx_check_member(const struct btf_type *t,
return -EINVAL;
}
+#ifdef CONFIG_EXT_SUB_SCHED
+ /*
+ * Enable private stack for operations that can nest along the
+ * hierarchy.
+ *
+ * XXX - Ideally, we should only do this for scheds that allow
+ * sub-scheds and sub-scheds themselves but I don't know how to access
+ * struct_ops from here.
+ */
+ switch (moff) {
+ case offsetof(struct sched_ext_ops, dispatch):
+ prog->aux->priv_stack_requested = true;
+ prog->aux->recursion_detected = scx_pstack_recursion_on_dispatch;
+ }
+#endif /* CONFIG_EXT_SUB_SCHED */
+
return 0;
}
@@ -6799,6 +6845,49 @@ __bpf_kfunc bool scx_bpf_dsq_move_vtime(struct bpf_iter_scx_dsq *it__iter,
p, dsq_id, enq_flags | SCX_ENQ_DSQ_PRIQ);
}
+#ifdef CONFIG_EXT_SUB_SCHED
+/**
+ * scx_bpf_sub_dispatch - Trigger dispatching on a child scheduler
+ * @cgroup_id: cgroup ID of the child scheduler to dispatch
+ * @aux__prog: magic BPF argument to access bpf_prog_aux hidden from BPF progs
+ *
+ * Allows a parent scheduler to trigger dispatching on one of its direct
+ * child schedulers. The child scheduler runs its dispatch operation to
+ * move tasks from dispatch queues to the local runqueue.
+ *
+ * Returns: true on success, false if cgroup_id is invalid, not a direct
+ * child, or caller lacks dispatch permission.
+ */
+__bpf_kfunc bool scx_bpf_sub_dispatch(u64 cgroup_id,
+ const struct bpf_prog_aux *aux__prog)
+{
+ struct rq *this_rq = this_rq();
+ struct scx_sched *parent, *child;
+
+ guard(rcu)();
+ parent = scx_prog_sched(aux__prog);
+ if (unlikely(!parent))
+ return false;
+
+ if (!scx_kf_allowed(parent, SCX_KF_DISPATCH))
+ return false;
+
+ child = scx_find_sub_sched(cgroup_id);
+
+ if (unlikely(!child))
+ return false;
+
+ if (unlikely(scx_parent(child) != parent)) {
+ scx_error(parent, "trying to dispatch a distant sub-sched on cgroup %llu",
+ cgroup_id);
+ return false;
+ }
+
+ return scx_dispatch_sched(child, this_rq, this_rq->scx.sub_dispatch_prev,
+ true);
+}
+#endif /* CONFIG_EXT_SUB_SCHED */
+
__bpf_kfunc_end_defs();
BTF_KFUNCS_START(scx_kfunc_ids_dispatch)
@@ -6809,6 +6898,9 @@ BTF_ID_FLAGS(func, scx_bpf_dsq_move_set_slice)
BTF_ID_FLAGS(func, scx_bpf_dsq_move_set_vtime)
BTF_ID_FLAGS(func, scx_bpf_dsq_move, KF_RCU)
BTF_ID_FLAGS(func, scx_bpf_dsq_move_vtime, KF_RCU)
+#ifdef CONFIG_EXT_SUB_SCHED
+BTF_ID_FLAGS(func, scx_bpf_sub_dispatch)
+#endif
BTF_KFUNCS_END(scx_kfunc_ids_dispatch)
static const struct btf_kfunc_id_set scx_kfunc_set_dispatch = {
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index cd6bdcdf9314..0aa0caa84308 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -777,6 +777,9 @@ struct scx_rq {
cpumask_var_t cpus_to_preempt;
cpumask_var_t cpus_to_wait;
unsigned long pnt_seq;
+
+ struct task_struct *sub_dispatch_prev;
+
struct balance_callback deferred_bal_cb;
struct irq_work deferred_irq_work;
struct irq_work kick_cpus_irq_work;
diff --git a/tools/sched_ext/include/scx/common.bpf.h b/tools/sched_ext/include/scx/common.bpf.h
index 6bf75f970237..a1890790d58c 100644
--- a/tools/sched_ext/include/scx/common.bpf.h
+++ b/tools/sched_ext/include/scx/common.bpf.h
@@ -140,6 +140,8 @@ struct cgroup *scx_bpf_task_cgroup(struct task_struct *p, const struct bpf_prog_
#define scx_bpf_task_cgroup(p) scx_bpf_task_cgroup((p), NULL)
u64 scx_bpf_now(void) __ksym __weak;
void scx_bpf_events(struct scx_event_stats *events, size_t events__sz) __ksym __weak;
+bool scx_bpf_sub_dispatch(u64 cgroup_id, const struct bpf_prog_aux *aux__prog) __ksym __weak;
+#define scx_bpf_sub_dispatch(cgroup_id) scx_bpf_sub_dispatch((cgroup_id), NULL)
/*
* Use the following as @it__iter when calling scx_bpf_dsq_move[_vtime]() from
diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 15e15cb234dc..9927cd1064e7 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -48,6 +48,9 @@ const volatile bool suppress_dump;
u64 nr_highpri_queued;
u32 test_error_cnt;
+#define MAX_SUB_SCHEDS 8
+u64 sub_sched_cgroup_ids[MAX_SUB_SCHEDS];
+
UEI_DEFINE(uei);
struct qmap {
@@ -452,6 +455,12 @@ void BPF_STRUCT_OPS(qmap_dispatch, s32 cpu, struct task_struct *prev)
cpuc->dsp_cnt = 0;
}
+ for (i = 0; i < MAX_SUB_SCHEDS; i++) {
+ if (sub_sched_cgroup_ids[i] &&
+ scx_bpf_sub_dispatch(sub_sched_cgroup_ids[i]))
+ return;
+ }
+
/*
* No other tasks. @prev will keep running. Update its core_sched_seq as
* if the task were enqueued and dispatched immediately.
@@ -877,7 +886,32 @@ void BPF_STRUCT_OPS(qmap_exit, struct scx_exit_info *ei)
s32 BPF_STRUCT_OPS(qmap_sub_attach, struct scx_sub_attach_args *args)
{
- return 0;
+ int i;
+
+ for (i = 0; i < MAX_SUB_SCHEDS; i++) {
+ if (!sub_sched_cgroup_ids[i]) {
+ sub_sched_cgroup_ids[i] = args->ops->sub_cgroup_id;
+ bpf_printk("attaching sub-sched[%d] on %s",
+ i, args->cgroup_path);
+ return 0;
+ }
+ }
+
+ return -ENOSPC;
+}
+
+void BPF_STRUCT_OPS(qmap_sub_detach, struct scx_sub_detach_args *args)
+{
+ int i;
+
+ for (i = 0; i < MAX_SUB_SCHEDS; i++) {
+ if (sub_sched_cgroup_ids[i] == args->ops->sub_cgroup_id) {
+ sub_sched_cgroup_ids[i] = 0;
+ bpf_printk("detaching sub-sched[%d] on %s",
+ i, args->cgroup_path);
+ break;
+ }
+ }
}
SCX_OPS_DEFINE(qmap_ops,
@@ -896,6 +930,7 @@ SCX_OPS_DEFINE(qmap_ops,
.cgroup_set_weight = (void *)qmap_cgroup_set_weight,
.cgroup_set_bandwidth = (void *)qmap_cgroup_set_bandwidth,
.sub_attach = (void *)qmap_sub_attach,
+ .sub_detach = (void *)qmap_sub_detach,
.cpu_online = (void *)qmap_cpu_online,
.cpu_offline = (void *)qmap_cpu_offline,
.init = (void *)qmap_init,
--
2.51.0
^ permalink raw reply related [flat|nested] 47+ messages in thread