[RFC PATCH 2/3] cgroup: support to enable nmi-safe css_rstat_updated

[Shakeel Butt <shakeel.butt@linux.dev>]

Add necessary infrastructure to enable the nmi-safe execution of
css_rstat_updated(). Currently css_rstat_updated() takes a per-cpu
per-css raw spinlock to add the given css in the per-cpu per-css update
tree. However the kernel can not spin in nmi context, so we need to
replace spinning on the raw spinlock with the trylock and on failure,
add the given css to the per-cpu backlog which will be processed when
the context that can spin on raw spinlock can run.

For now, this patch just adds necessary data structures in the css and
ss structures.

Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
---
 include/linux/cgroup-defs.h |  4 ++++
 kernel/cgroup/rstat.c       | 13 +++++++++++--
 2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 560582c4dbeb..f7b680f853ea 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -370,6 +370,9 @@ struct css_rstat_cpu {
 	 */
 	struct cgroup_subsys_state *updated_children;	/* terminated by self cgroup */
 	struct cgroup_subsys_state *updated_next;	/* NULL iff not on the list */
+
+	struct llist_node lnode;			/* lockless backlog node */
+	struct cgroup_subsys_state *owner;		/* back pointer */
 };
 
 /*
@@ -800,6 +803,7 @@ struct cgroup_subsys {
 
 	spinlock_t rstat_ss_lock;
 	raw_spinlock_t __percpu *rstat_ss_cpu_lock;
+	struct llist_head __percpu *lhead; /* lockless backlog list */
 };
 
 extern struct percpu_rw_semaphore cgroup_threadgroup_rwsem;
diff --git a/kernel/cgroup/rstat.c b/kernel/cgroup/rstat.c
index a30bcc4d4f48..d3092b4c85d7 100644
--- a/kernel/cgroup/rstat.c
+++ b/kernel/cgroup/rstat.c
@@ -419,7 +419,8 @@ int css_rstat_init(struct cgroup_subsys_state *css)
 	for_each_possible_cpu(cpu) {
 		struct css_rstat_cpu *rstatc = css_rstat_cpu(css, cpu);
 
-		rstatc->updated_children = css;
+		rstatc->owner = rstatc->updated_children = css;
+		init_llist_node(&rstatc->lnode);
 
 		if (css_is_cgroup(css)) {
 			struct cgroup_rstat_base_cpu *rstatbc;
@@ -484,8 +485,16 @@ int __init ss_rstat_init(struct cgroup_subsys *ss)
 	if (!ss->rstat_ss_cpu_lock)
 		return -ENOMEM;
 
-	for_each_possible_cpu(cpu)
+	ss->lhead = alloc_percpu(struct llist_head);
+	if (!ss->lhead) {
+		free_percpu(ss->rstat_ss_cpu_lock);
+		return -ENOMEM;
+	}
+
+	for_each_possible_cpu(cpu) {
 		raw_spin_lock_init(per_cpu_ptr(ss->rstat_ss_cpu_lock, cpu));
+		init_llist_head(per_cpu_ptr(ss->lhead, cpu));
+	}
 
 	return 0;
 }
-- 
2.47.1