[RFC PATCH 0/9] Add kernel cmdline option for rt_group_sched

[RFC PATCH 0/9] Add kernel cmdline option for rt_group_sched

[Michal Koutný]

Despite RT_GROUP_SCHED is only available on cgroup v1, there are still
some users of this feature. General purpose distros (e.g. [1][2][3][4])
cannot enable CONFIG_RT_GROUP_SCHED easily:
- since it prevents creation of RT tasks unless RT runtime is determined
  and distributed into cgroup tree,
- grouping of RT threads is not what is desired by default on such
  systems,
- it prevents use of cgroup v2 with RT tasks.

This changeset aims at deferring the decision whether to have
CONFIG_RT_GROUP_SCHED or not up until the boot time.
By default RT groups are available as originally but the user can
pass rt_group_sched=0 kernel cmdline parameter that disables the
grouping and behavior is like with !CONFIG_RT_GROUP_SCHED (with certain
runtime overhead).

The series is organized as follows:

1) generic ifdefs cleanup, no functional changes,
2) preparing root_task_group to be used in places that take shortcuts in
   the case of !CONFIG_RT_GROUP_SCHED,
3) boot cmdline option that controls cgroup (v1) attributes,
4) conditional bypass of non-root task groups,
5) checks and comments refresh.

The crux are patches:
  sched: Skip non-root task_groups with disabled RT_GROUP
  sched: Bypass bandwitdh checks with runtime disabled RT_GROUP_SCHED

Futher notes:
- it is not sched_feat() flag because that can be flipped any time
- runtime disablement is not implemented as infinite per-cgroup RT limit
  since that'd still employ group scheduling which is unlike
  !CONFIG_RT_GROUP_SCHED

RFC notes:
- there remain two variants of various functions for
  CONFIG_RT_GROUP_SCHED and !CONFIG_RT_GROUP_SCHED, those could be
  folded into one and runtime evaluated guards in the folded functions
  could be used (I haven't posted it yet due to unclear performance
  benefit)
- I noticed some lockdep issues over rt_runtime_lock but those are also
  in an unpatched kernel (and they seem to have been present since a
  long time without complications)

[1] Debian (https://salsa.debian.org/kernel-team/linux/-/blob/debian/latest/debian/config/kernelarch-x86/config),
[2] ArchLinux (https://gitlab.archlinux.org/archlinux/packaging/packages/linux/-/blob/main/config),
[3] Fedora (https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel-x86_64-fedora.config)
[4] openSUSE TW (https://github.com/SUSE/kernel-source/blob/stable/config/x86_64/default)

Michal Koutný (9):
  sched: Convert CONFIG_RT_GROUP_SCHED macros to code conditions
  sched: Remove unneeed macro wrap
  sched: Always initialize rt_rq's task_group
  sched: Add commadline option for RT_GROUP_SCHED toggling
  sched: Skip non-root task_groups with disabled RT_GROUP_SCHED
  sched: Bypass bandwitdh checks with runtime disabled RT_GROUP_SCHED
  sched: Do not construct nor expose RT_GROUP_SCHED structures if
    disabled
  sched: Add RT_GROUP WARN checks for non-root task_groups
  sched: Add annotations to RT_GROUP_SCHED fields

 .../admin-guide/kernel-parameters.txt         |  5 ++
 init/Kconfig                                  | 11 +++
 kernel/sched/core.c                           | 69 +++++++++++++++----
 kernel/sched/rt.c                             | 51 +++++++++-----
 kernel/sched/sched.h                          | 34 +++++++--
 kernel/sched/syscalls.c                       |  5 +-
 6 files changed, 137 insertions(+), 38 deletions(-)


base-commit: f92f4749861b06fed908d336b4dee1326003291b
-- 
2.47.1

[RFC PATCH 1/9] sched: Convert CONFIG_RT_GROUP_SCHED macros to code conditions

[Michal Koutný]

Convert the blocks guarded by macros to regular code so that the RT
group code gets more compile validation. Reasoning is in
Documentation/process/coding-style.rst 21) Conditional Compilation.
With that, no functional change is expected.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/rt.c       | 10 ++++------
 kernel/sched/syscalls.c |  2 +-
 2 files changed, 5 insertions(+), 7 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index bd66a46b06aca..6ea46c7219634 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1068,13 +1068,12 @@ inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
 {
 	struct rq *rq = rq_of_rt_rq(rt_rq);
 
-#ifdef CONFIG_RT_GROUP_SCHED
 	/*
 	 * Change rq's cpupri only if rt_rq is the top queue.
 	 */
-	if (&rq->rt != rt_rq)
+	if (IS_ENABLED(CONFIG_RT_GROUP_SCHED) && &rq->rt != rt_rq)
 		return;
-#endif
+
 	if (rq->online && prio < prev_prio)
 		cpupri_set(&rq->rd->cpupri, rq->cpu, prio);
 }
@@ -1084,13 +1083,12 @@ dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio)
 {
 	struct rq *rq = rq_of_rt_rq(rt_rq);
 
-#ifdef CONFIG_RT_GROUP_SCHED
 	/*
 	 * Change rq's cpupri only if rt_rq is the top queue.
 	 */
-	if (&rq->rt != rt_rq)
+	if (IS_ENABLED(CONFIG_RT_GROUP_SCHED) && &rq->rt != rt_rq)
 		return;
-#endif
+
 	if (rq->online && rt_rq->highest_prio.curr != prev_prio)
 		cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr);
 }
diff --git a/kernel/sched/syscalls.c b/kernel/sched/syscalls.c
index ff0e5ab4e37cb..77d0d4a2b68da 100644
--- a/kernel/sched/syscalls.c
+++ b/kernel/sched/syscalls.c
@@ -650,7 +650,7 @@ int __sched_setscheduler(struct task_struct *p,
 			retval = -EPERM;
 			goto unlock;
 		}
-#endif
+#endif /* CONFIG_RT_GROUP_SCHED */
 #ifdef CONFIG_SMP
 		if (dl_bandwidth_enabled() && dl_policy(policy) &&
 				!(attr->sched_flags & SCHED_FLAG_SUGOV)) {
-- 
2.47.1

[RFC PATCH 2/9] sched: Remove unneeed macro wrap

[Michal Koutný]

rt_entity_is_task has split definitions based on CONFIG_RT_GROUP_SCHED,
therefore we can use it always. No functional change intended.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/rt.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 6ea46c7219634..1940301c40f7d 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1257,11 +1257,9 @@ static void __delist_rt_entity(struct sched_rt_entity *rt_se, struct rt_prio_arr
 static inline struct sched_statistics *
 __schedstats_from_rt_se(struct sched_rt_entity *rt_se)
 {
-#ifdef CONFIG_RT_GROUP_SCHED
 	/* schedstats is not supported for rt group. */
 	if (!rt_entity_is_task(rt_se))
 		return NULL;
-#endif
 
 	return &rt_task_of(rt_se)->stats;
 }
-- 
2.47.1

[RFC PATCH 3/9] sched: Always initialize rt_rq's task_group

[Michal Koutný]

rt_rq->tg may be NULL which denotes the root task_group.
Store the pointer to root_task_group directly so that callers may use
rt_rq->tg homogenously.

root_task_group exists always with CONFIG_CGROUPS_SCHED,
CONFIG_RT_GROUP_SCHED depends on that.

This changes root level rt_rq's default limit from infinity to the
value of (originally) global RT throttling.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/rt.c    | 7 ++-----
 kernel/sched/sched.h | 2 ++
 2 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 1940301c40f7d..41fed8865cb09 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -89,6 +89,7 @@ void init_rt_rq(struct rt_rq *rt_rq)
 	rt_rq->rt_throttled = 0;
 	rt_rq->rt_runtime = 0;
 	raw_spin_lock_init(&rt_rq->rt_runtime_lock);
+	rt_rq->tg = &root_task_group;
 #endif
 }
 
@@ -484,9 +485,6 @@ static inline bool rt_task_fits_capacity(struct task_struct *p, int cpu)
 
 static inline u64 sched_rt_runtime(struct rt_rq *rt_rq)
 {
-	if (!rt_rq->tg)
-		return RUNTIME_INF;
-
 	return rt_rq->rt_runtime;
 }
 
@@ -1156,8 +1154,7 @@ inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
 	if (rt_se_boosted(rt_se))
 		rt_rq->rt_nr_boosted++;
 
-	if (rt_rq->tg)
-		start_rt_bandwidth(&rt_rq->tg->rt_bandwidth);
+	start_rt_bandwidth(&rt_rq->tg->rt_bandwidth);
 }
 
 static void
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 76f5f53a645fc..38325bd32a0e0 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -820,6 +820,8 @@ struct rt_rq {
 	unsigned int		rt_nr_boosted;
 
 	struct rq		*rq;
+#endif
+#ifdef CONFIG_CGROUP_SCHED
 	struct task_group	*tg;
 #endif
 };
-- 
2.47.1

[RFC PATCH 4/9] sched: Add commadline option for RT_GROUP_SCHED toggling

[Michal Koutný]

Only simple implementation with a static key wrapper, it will be wired
in later.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 .../admin-guide/kernel-parameters.txt         |  5 ++++
 init/Kconfig                                  | 11 ++++++++
 kernel/sched/core.c                           | 25 +++++++++++++++++++
 kernel/sched/sched.h                          | 17 +++++++++++++
 4 files changed, 58 insertions(+)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 3872bc6ec49d6..1c890c9ad8716 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -5937,6 +5937,11 @@
 			Memory area to be used by remote processor image,
 			managed by CMA.
 
+	rt_group_sched=	[KNL] Enable or disable SCHED_RR/FIFO group scheduling
+			when CONFIG_RT_GROUP_SCHED=y. Defaults to
+			!CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED.
+			Format: <bool>
+
 	rw		[KNL] Mount root device read-write on boot
 
 	S		[KNL] Run init in single mode
diff --git a/init/Kconfig b/init/Kconfig
index a20e6efd3f0fb..7823e5ac0311d 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1076,6 +1076,17 @@ config RT_GROUP_SCHED
 	  realtime bandwidth for them.
 	  See Documentation/scheduler/sched-rt-group.rst for more information.
 
+config RT_GROUP_SCHED_DEFAULT_DISABLED
+	bool "Require boot parameter to enable group scheduling for SCHED_RR/FIFO"
+	depends on RT_GROUP_SCHED
+	default n
+	help
+	  When set, the RT group scheduling is disabled by default. The option
+	  is in inverted form so that mere RT_GROUP_SCHED enables the group
+	  scheduling.
+
+	  Say N if unsure.
+
 config EXT_GROUP_SCHED
 	bool
 	depends on SCHED_CLASS_EXT && CGROUP_SCHED
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index c6d8232ad9eea..47898f895a5a3 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9844,6 +9844,31 @@ static struct cftype cpu_legacy_files[] = {
 	{ }	/* Terminate */
 };
 
+#ifdef CONFIG_RT_GROUP_SCHED
+# ifdef RT_GROUP_SCHED_DEFAULT_DISABLED
+DEFINE_STATIC_KEY_FALSE(rt_group_sched);
+# else
+DEFINE_STATIC_KEY_TRUE(rt_group_sched);
+# endif
+
+static int __init setup_rt_group_sched(char *str)
+{
+	long val;
+
+	if (kstrtol(str, 0, &val) || val < 0 || val > 1) {
+		pr_warn("Unable to set rt_group_sched\n");
+		return 1;
+	}
+	if (val)
+		static_branch_enable(&rt_group_sched);
+	else
+		static_branch_disable(&rt_group_sched);
+
+	return 1;
+}
+__setup("rt_group_sched=", setup_rt_group_sched);
+#endif /* CONFIG_RT_GROUP_SCHED */
+
 static int cpu_extra_stat_show(struct seq_file *sf,
 			       struct cgroup_subsys_state *css)
 {
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 38325bd32a0e0..1c457dc1472a3 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1501,6 +1501,23 @@ static inline bool sched_group_cookie_match(struct rq *rq,
 }
 
 #endif /* !CONFIG_SCHED_CORE */
+#ifdef CONFIG_RT_GROUP_SCHED
+# ifdef RT_GROUP_SCHED_DEFAULT_DISABLED
+DECLARE_STATIC_KEY_FALSE(rt_group_sched);
+static inline bool rt_group_sched_enabled(void)
+{
+	return static_branch_unlikely(&rt_group_sched);
+}
+# else
+DECLARE_STATIC_KEY_TRUE(rt_group_sched);
+static inline bool rt_group_sched_enabled(void)
+{
+	return static_branch_likely(&rt_group_sched);
+}
+# endif /* RT_GROUP_SCHED_DEFAULT_DISABLED */
+#else
+# define rt_group_sched_enabled()	false
+#endif /* CONFIG_RT_GROUP_SCHED */
 
 static inline void lockdep_assert_rq_held(struct rq *rq)
 {
-- 
2.47.1

[RFC PATCH 5/9] sched: Skip non-root task_groups with disabled RT_GROUP_SCHED

[Michal Koutný]

First, we want to prevent placement of RT tasks on non-root rt_rqs which
we achieve in the task migration code that'd fall back to
root_task_group's rt_rq.

Second, we want to work with only root_task_group's rt_rq when iterating
all "real" rt_rqs when RT_GROUP is disabled. To achieve this we keep
root_task_group as the first one on the task_groups and break out
quickly.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/core.c  | 2 +-
 kernel/sched/rt.c    | 9 ++++++---
 kernel/sched/sched.h | 7 +++++++
 3 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 47898f895a5a3..dfd2778622b8b 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -8958,7 +8958,7 @@ void sched_online_group(struct task_group *tg, struct task_group *parent)
 	unsigned long flags;
 
 	spin_lock_irqsave(&task_group_lock, flags);
-	list_add_rcu(&tg->list, &task_groups);
+	list_add_tail_rcu(&tg->list, &task_groups);
 
 	/* Root should already exist: */
 	WARN_ON(!parent);
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 41fed8865cb09..923ec978ff756 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -497,6 +497,9 @@ typedef struct task_group *rt_rq_iter_t;
 
 static inline struct task_group *next_task_group(struct task_group *tg)
 {
+	if (!rt_group_sched_enabled())
+		return NULL;
+
 	do {
 		tg = list_entry_rcu(tg->list.next,
 			typeof(struct task_group), list);
@@ -509,9 +512,9 @@ static inline struct task_group *next_task_group(struct task_group *tg)
 }
 
 #define for_each_rt_rq(rt_rq, iter, rq)					\
-	for (iter = container_of(&task_groups, typeof(*iter), list);	\
-		(iter = next_task_group(iter)) &&			\
-		(rt_rq = iter->rt_rq[cpu_of(rq)]);)
+	for (iter = &root_task_group;					\
+		iter && (rt_rq = iter->rt_rq[cpu_of(rq)]);		\
+		iter = next_task_group(iter))
 
 #define for_each_sched_rt_entity(rt_se) \
 	for (; rt_se; rt_se = rt_se->parent)
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 1c457dc1472a3..d8d28c3d1ac5f 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2148,6 +2148,13 @@ static inline void set_task_rq(struct task_struct *p, unsigned int cpu)
 #endif
 
 #ifdef CONFIG_RT_GROUP_SCHED
+	/*
+	 * p->rt.rt_rq is NULL initially and it is easier to assign
+	 * root_task_group's rt_rq than switching in rt_rq_of_se()
+	 * Clobbers tg(!)
+	 */
+	if (!rt_group_sched_enabled())
+		tg = &root_task_group;
 	p->rt.rt_rq  = tg->rt_rq[cpu];
 	p->rt.parent = tg->rt_se[cpu];
 #endif
-- 
2.47.1

[RFC PATCH 6/9] sched: Bypass bandwitdh checks with runtime disabled RT_GROUP_SCHED

[Michal Koutný]

When RT_GROUPs are compiled but not exposed, their bandwidth cannot
be configured (and it is not initialized for non-root task_groups neither).
Therefore bypass any checks of task vs task_group bandwidth.

This will achieve behavior very similar to setups that have
!CONFIG_RT_GROUP_SCHED and attach cpu controller to cgroup v2 hierarchy.
(On a related note, this may allow having RT tasks with
CONFIG_RT_GROUP_SCHED and cgroup v2 hierarchy.)

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/core.c     | 6 +++++-
 kernel/sched/rt.c       | 2 +-
 kernel/sched/syscalls.c | 3 ++-
 3 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index dfd2778622b8b..6e21e0885557d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9158,11 +9158,15 @@ static int cpu_cgroup_can_attach(struct cgroup_taskset *tset)
 	struct task_struct *task;
 	struct cgroup_subsys_state *css;
 
+	if (!rt_group_sched_enabled())
+		goto scx_check;
+
 	cgroup_taskset_for_each(task, css, tset) {
 		if (!sched_rt_can_attach(css_tg(css), task))
 			return -EINVAL;
 	}
-#endif
+scx_check:
+#endif /* CONFIG_RT_GROUP_SCHED */
 	return scx_cgroup_can_attach(tset);
 }
 
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 923ec978ff756..161d91f7479b4 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -2866,7 +2866,7 @@ static int sched_rt_global_constraints(void)
 int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
 {
 	/* Don't accept real-time tasks when there is no way for them to run */
-	if (rt_task(tsk) && tg->rt_bandwidth.rt_runtime == 0)
+	if (rt_group_sched_enabled() && rt_task(tsk) && tg->rt_bandwidth.rt_runtime == 0)
 		return 0;
 
 	return 1;
diff --git a/kernel/sched/syscalls.c b/kernel/sched/syscalls.c
index 77d0d4a2b68da..80382f5d53a44 100644
--- a/kernel/sched/syscalls.c
+++ b/kernel/sched/syscalls.c
@@ -644,7 +644,8 @@ int __sched_setscheduler(struct task_struct *p,
 		 * Do not allow real-time tasks into groups that have no runtime
 		 * assigned.
 		 */
-		if (rt_bandwidth_enabled() && rt_policy(policy) &&
+		if (rt_group_sched_enabled() &&
+				rt_bandwidth_enabled() && rt_policy(policy) &&
 				task_group(p)->rt_bandwidth.rt_runtime == 0 &&
 				!task_group_is_autogroup(task_group(p))) {
 			retval = -EPERM;
-- 
2.47.1

[RFC PATCH 7/9] sched: Do not construct nor expose RT_GROUP_SCHED structures if disabled

[Michal Koutný]

Thanks to kernel cmdline being available early, before any
cgroup hierarchy exists, we can achieve the RT_GROUP_SCHED boottime
disabling goal by simply skipping any creation (and destruction) of
RT_GROUP data and its exposure via RT attributes.

We can do this thanks to previously placed runtime guards that would
redirect all operations to root_task_group's data when RT_GROUP_SCHED
disabled.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/core.c | 36 ++++++++++++++++++++++++------------
 kernel/sched/rt.c   |  9 +++++++++
 2 files changed, 33 insertions(+), 12 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 6e21e0885557d..300a1a83e1a3c 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9819,18 +9819,6 @@ static struct cftype cpu_legacy_files[] = {
 		.seq_show = cpu_cfs_local_stat_show,
 	},
 #endif
-#ifdef CONFIG_RT_GROUP_SCHED
-	{
-		.name = "rt_runtime_us",
-		.read_s64 = cpu_rt_runtime_read,
-		.write_s64 = cpu_rt_runtime_write,
-	},
-	{
-		.name = "rt_period_us",
-		.read_u64 = cpu_rt_period_read_uint,
-		.write_u64 = cpu_rt_period_write_uint,
-	},
-#endif
 #ifdef CONFIG_UCLAMP_TASK_GROUP
 	{
 		.name = "uclamp.min",
@@ -9849,6 +9837,20 @@ static struct cftype cpu_legacy_files[] = {
 };
 
 #ifdef CONFIG_RT_GROUP_SCHED
+static struct cftype rt_group_files[] = {
+	{
+		.name = "rt_runtime_us",
+		.read_s64 = cpu_rt_runtime_read,
+		.write_s64 = cpu_rt_runtime_write,
+	},
+	{
+		.name = "rt_period_us",
+		.read_u64 = cpu_rt_period_read_uint,
+		.write_u64 = cpu_rt_period_write_uint,
+	},
+	{ }	/* Terminate */
+};
+
 # ifdef RT_GROUP_SCHED_DEFAULT_DISABLED
 DEFINE_STATIC_KEY_FALSE(rt_group_sched);
 # else
@@ -9871,6 +9873,16 @@ static int __init setup_rt_group_sched(char *str)
 	return 1;
 }
 __setup("rt_group_sched=", setup_rt_group_sched);
+
+static int __init cpu_rt_group_init(void)
+{
+	if (!rt_group_sched_enabled())
+		return 0;
+
+	WARN_ON(cgroup_add_legacy_cftypes(&cpu_cgrp_subsys, rt_group_files));
+	return 0;
+}
+subsys_initcall(cpu_rt_group_init);
 #endif /* CONFIG_RT_GROUP_SCHED */
 
 static int cpu_extra_stat_show(struct seq_file *sf,
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 161d91f7479b4..db7cdc82003bd 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -195,6 +195,9 @@ static inline struct rq *rq_of_rt_se(struct sched_rt_entity *rt_se)
 
 void unregister_rt_sched_group(struct task_group *tg)
 {
+	if (!rt_group_sched_enabled())
+		return;
+
 	if (tg->rt_se)
 		destroy_rt_bandwidth(&tg->rt_bandwidth);
 }
@@ -203,6 +206,9 @@ void free_rt_sched_group(struct task_group *tg)
 {
 	int i;
 
+	if (!rt_group_sched_enabled())
+		return;
+
 	for_each_possible_cpu(i) {
 		if (tg->rt_rq)
 			kfree(tg->rt_rq[i]);
@@ -247,6 +253,9 @@ int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
 	struct sched_rt_entity *rt_se;
 	int i;
 
+	if (!rt_group_sched_enabled())
+		return 1;
+
 	tg->rt_rq = kcalloc(nr_cpu_ids, sizeof(rt_rq), GFP_KERNEL);
 	if (!tg->rt_rq)
 		goto err;
-- 
2.47.1

[RFC PATCH 8/9] sched: Add RT_GROUP WARN checks for non-root task_groups

[Michal Koutný]

With CONFIG_RT_GROUP_SCHED but runtime disabling of RT_GROUPs we expect
the existence of the root task_group only and all rt_sched_entity'ies
should be queued on root's rt_rq.

If we get a non-root RT_GROUP something went wrong.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/rt.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index db7cdc82003bd..deacd46e27823 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -178,11 +178,14 @@ static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se)
 
 static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq)
 {
+	/* Cannot fold with non-CONFIG_RT_GROUP_SCHED version, layout */
+	WARN_ON(!rt_group_sched_enabled() && rt_rq->tg != &root_task_group);
 	return rt_rq->rq;
 }
 
 static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se)
 {
+	WARN_ON(!rt_group_sched_enabled() && rt_se->rt_rq->tg != &root_task_group);
 	return rt_se->rt_rq;
 }
 
@@ -190,6 +193,7 @@ static inline struct rq *rq_of_rt_se(struct sched_rt_entity *rt_se)
 {
 	struct rt_rq *rt_rq = rt_se->rt_rq;
 
+	WARN_ON(!rt_group_sched_enabled() && rt_rq->tg != &root_task_group);
 	return rt_rq->rq;
 }
 
@@ -506,8 +510,10 @@ typedef struct task_group *rt_rq_iter_t;
 
 static inline struct task_group *next_task_group(struct task_group *tg)
 {
-	if (!rt_group_sched_enabled())
+	if (!rt_group_sched_enabled()) {
+		WARN_ON(tg != &root_task_group);
 		return NULL;
+	}
 
 	do {
 		tg = list_entry_rcu(tg->list.next,
@@ -2609,8 +2615,9 @@ static int task_is_throttled_rt(struct task_struct *p, int cpu)
 {
 	struct rt_rq *rt_rq;
 
-#ifdef CONFIG_RT_GROUP_SCHED
+#ifdef CONFIG_RT_GROUP_SCHED // XXX maybe add task_rt_rq(), see also sched_rt_period_rt_rq
 	rt_rq = task_group(p)->rt_rq[cpu];
+	WARN_ON(!rt_group_sched_enabled() && rt_rq->tg != &root_task_group);
 #else
 	rt_rq = &cpu_rq(cpu)->rt;
 #endif
@@ -2720,6 +2727,9 @@ static int tg_rt_schedulable(struct task_group *tg, void *data)
 	    tg->rt_bandwidth.rt_runtime && tg_has_rt_tasks(tg))
 		return -EBUSY;
 
+	if (WARN_ON(!rt_group_sched_enabled() && tg != &root_task_group))
+		return -EBUSY;
+
 	total = to_ratio(period, runtime);
 
 	/*
-- 
2.47.1

[RFC PATCH 9/9] sched: Add annotations to RT_GROUP_SCHED fields

[Michal Koutný]

Update comments to ease RT throttling understanding.

Signed-off-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/sched/sched.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index d8d28c3d1ac5f..5c32c23915810 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -812,17 +812,17 @@ struct rt_rq {
 
 #ifdef CONFIG_RT_GROUP_SCHED
 	int			rt_throttled;
-	u64			rt_time;
-	u64			rt_runtime;
+	u64			rt_time; /* consumed RT time, goes up in update_curr_rt */
+	u64			rt_runtime; /* allotted RT time, "slice" from rt_bandwidth, RT sharing/balancing */
 	/* Nests inside the rq lock: */
 	raw_spinlock_t		rt_runtime_lock;
 
 	unsigned int		rt_nr_boosted;
 
-	struct rq		*rq;
+	struct rq		*rq; /* this is always top-level rq, cache? */
 #endif
 #ifdef CONFIG_CGROUP_SCHED
-	struct task_group	*tg;
+	struct task_group	*tg; /* this tg has "this" rt_rq on given CPU for runnable entities */
 #endif
 };
 
-- 
2.47.1

Re: [RFC PATCH 0/9] Add kernel cmdline option for rt_group_sched

[Michal Koutný]

On Mon, Dec 16, 2024 at 09:12:56PM +0100, Michal Koutný <mkoutny@suse.com> wrote:
> The series is organized as follows:

(I saw no replies, this may have slipped through the turn of the year
period.)

> RFC notes:

So I wonder if there any initial comments on this change.

Thanks,
Michal

Re: [RFC PATCH 0/9] Add kernel cmdline option for rt_group_sched

[Peter Zijlstra]

On Mon, Dec 16, 2024 at 09:12:56PM +0100, Michal Koutný wrote:
> Despite RT_GROUP_SCHED is only available on cgroup v1, there are still
> some users of this feature. General purpose distros (e.g. [1][2][3][4])
> cannot enable CONFIG_RT_GROUP_SCHED easily:

We all hate this thing and want it to go away. So not being able to use
it is a pro from where I'm at.

Sadly the replacement isn't there yet either, which makes it all really
difficult.

Re: [RFC PATCH 0/9] Add kernel cmdline option for rt_group_sched

[Michal Koutný]

[-- Attachment #1: Type: text/plain, Size: 860 bytes --]

On Tue, Jan 07, 2025 at 08:41:06PM +0100, Peter Zijlstra <peterz@infradead.org> wrote:
> We all hate this thing and want it to go away. So not being able to use
> it is a pro from where I'm at.

I understand and to some extent am not a fan of it neither (we had
disabled it in SUSE quite some time ago). I'd consider the remaining
existing users legacy.

> Sadly the replacement isn't there yet either, which makes it all really
> difficult.

Exactly. Thus the runtime switch is meant as a bridge for general
purpose distros where a kernel is shipped pre-configured (i.e. one
config where the default is non-grouped not to hinder the majority use
cases).

Considering the legacy usecases on distribution kernels do you oppose
the chosen approach? I can work on changes if you have comments on the
implementation itself.

Thanks,
Michal


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]