cgroup/for-4.17: Make cgroup_rstat available to controllers

similarity index 100%
rename from kernel/cgroup/stat.c
rename to kernel/cgroup/rstat.c

index 90ede40..02625cf 100644
@@ -259,11 +259,11 @@ struct css_set {
 };
 
 /*
- * cgroup basic resource usage statistics.  Accounting is done per-cpu in
- * cgroup_cpu_stat which is then lazily propagated up the hierarchy on
- * reads.
+ * rstat - cgroup scalable recursive statistics.  Accounting is done
+ * per-cpu in cgroup_rstat_cpu which is then lazily propagated up the
+ * hierarchy on reads.
  *
- * When a stat gets updated, the cgroup_cpu_stat and its ancestors are
+ * When a stat gets updated, the cgroup_rstat_cpu and its ancestors are
  * linked into the updated tree.  On the following read, propagation only
  * considers and consumes the updated tree.  This makes reading O(the
  * number of descendants which have been active since last read) instead of

@@ -274,7 +274,7 @@ struct css_set {
  * become very expensive.  By propagating selectively, increasing reading
  * frequency decreases the cost of each read.
  */
-struct cgroup_cpu_stat {
+struct cgroup_rstat_cpu {
 	/*
 	 * ->sync protects all the current counters.  These are the only
 	 * fields which get updated in the hot path.

@@ -297,7 +297,7 @@ struct cgroup_cpu_stat {
 	 * to the cgroup makes it unnecessary for each per-cpu struct to
 	 * point back to the associated cgroup.
 	 *
-	 * Protected by per-cpu cgroup_cpu_stat_lock.
+	 * Protected by per-cpu cgroup_rstat_cpu_lock.
 	 */
 	struct cgroup *updated_children;	/* terminated by self cgroup */
 	struct cgroup *updated_next;		/* NULL iff not on the list */

@@ -408,8 +408,10 @@ struct cgroup {
 	 */
 	struct cgroup *dom_cgrp;
 
+	/* per-cpu recursive resource statistics */
+	struct cgroup_rstat_cpu __percpu *rstat_cpu;
+
 	/* cgroup basic resource statistics */
-	struct cgroup_cpu_stat __percpu *cpu_stat;
 	struct cgroup_stat pending_stat;	/* pending from children */
 	struct cgroup_stat stat;
 

index b928b27..0927111 100644
@@ -201,13 +201,13 @@ int cgroup_show_path(struct seq_file *sf, struct kernfs_node *kf_node,
 int cgroup_task_count(const struct cgroup *cgrp);
 
 /*
- * stat.c
+ * rstat.c
  */
-void cgroup_stat_flush(struct cgroup *cgrp);
-int cgroup_stat_init(struct cgroup *cgrp);
-void cgroup_stat_exit(struct cgroup *cgrp);
+void cgroup_rstat_flush(struct cgroup *cgrp);
+int cgroup_rstat_init(struct cgroup *cgrp);
+void cgroup_rstat_exit(struct cgroup *cgrp);
 void cgroup_stat_show_cputime(struct seq_file *seq);
-void cgroup_stat_boot(void);
+void cgroup_rstat_boot(void);
 
 /*
  * namespace.c

index ea31ec5..5549a7c 100644
@@ -144,14 +144,14 @@ static struct static_key_true *cgroup_subsys_on_dfl_key[] = {
 };
 #undef SUBSYS
 
-static DEFINE_PER_CPU(struct cgroup_cpu_stat, cgrp_dfl_root_cpu_stat);
+static DEFINE_PER_CPU(struct cgroup_rstat_cpu, cgrp_dfl_root_rstat_cpu);
 
 /*
  * The default hierarchy, reserved for the subsystems that are otherwise
  * unattached - it never has more than a single cgroup, and all tasks are
  * part of that cgroup.
  */
-struct cgroup_root cgrp_dfl_root = { .cgrp.cpu_stat = &cgrp_dfl_root_cpu_stat };
+struct cgroup_root cgrp_dfl_root = { .cgrp.rstat_cpu = &cgrp_dfl_root_rstat_cpu };
 EXPORT_SYMBOL_GPL(cgrp_dfl_root);
 
 /*

@@ -4592,7 +4592,7 @@ static void css_free_work_fn(struct work_struct *work)
 			cgroup_put(cgroup_parent(cgrp));
 			kernfs_put(cgrp->kn);
 			if (cgroup_on_dfl(cgrp))
-				cgroup_stat_exit(cgrp);
+				cgroup_rstat_exit(cgrp);
 			kfree(cgrp);
 		} else {
 			/*

@@ -4638,7 +4638,7 @@ static void css_release_work_fn(struct work_struct *work)
 		trace_cgroup_release(cgrp);
 
 		if (cgroup_on_dfl(cgrp))
-			cgroup_stat_flush(cgrp);
+			cgroup_rstat_flush(cgrp);
 
 		for (tcgrp = cgroup_parent(cgrp); tcgrp;
 		     tcgrp = cgroup_parent(tcgrp))

@@ -4824,7 +4824,7 @@ static struct cgroup *cgroup_create(struct cgroup *parent)
 		goto out_free_cgrp;
 
 	if (cgroup_on_dfl(parent)) {
-		ret = cgroup_stat_init(cgrp);
+		ret = cgroup_rstat_init(cgrp);
 		if (ret)
 			goto out_cancel_ref;
 	}

@@ -4889,7 +4889,7 @@ static struct cgroup *cgroup_create(struct cgroup *parent)
 	cgroup_idr_remove(&root->cgroup_idr, cgrp->id);
 out_stat_exit:
 	if (cgroup_on_dfl(parent))
-		cgroup_stat_exit(cgrp);
+		cgroup_rstat_exit(cgrp);
 out_cancel_ref:
 	percpu_ref_exit(&cgrp->self.refcnt);
 out_free_cgrp:

@@ -5282,7 +5282,7 @@ int __init cgroup_init(void)
 	BUG_ON(cgroup_init_cftypes(NULL, cgroup_base_files));
 	BUG_ON(cgroup_init_cftypes(NULL, cgroup1_base_files));
 
-	cgroup_stat_boot();
+	cgroup_rstat_boot();
 
 	/*
 	 * The latency of the synchronize_sched() is too high for cgroups,

374 lines not on this page · raw message

index 02625cf..cf9db7b 100644
@@ -258,6 +258,10 @@ struct css_set {
 	struct rcu_head rcu_head;
 };
 
+struct cgroup_base_stat {
+	struct task_cputime cputime;
+};
+
 /*
  * rstat - cgroup scalable recursive statistics.  Accounting is done
  * per-cpu in cgroup_rstat_cpu which is then lazily propagated up the

@@ -273,20 +277,24 @@ struct css_set {
  * aren't active and stat may be read frequently.  The combination can
  * become very expensive.  By propagating selectively, increasing reading
  * frequency decreases the cost of each read.
+ *
+ * This struct hosts both the fields which implement the above -
+ * updated_children and updated_next - and the fields which track basic
+ * resource statistics on top of it - bsync, bstat and last_bstat.
  */
 struct cgroup_rstat_cpu {
 	/*
-	 * ->sync protects all the current counters.  These are the only
-	 * fields which get updated in the hot path.
+	 * ->bsync protects ->bstat.  These are the only fields which get
+	 * updated in the hot path.
 	 */
-	struct u64_stats_sync sync;
-	struct task_cputime cputime;
+	struct u64_stats_sync bsync;
+	struct cgroup_base_stat bstat;
 
 	/*
 	 * Snapshots at the last reading.  These are used to calculate the
 	 * deltas to propagate to the global counters.
 	 */
-	struct task_cputime last_cputime;
+	struct cgroup_base_stat last_bstat;
 
 	/*
 	 * Child cgroups with stat updates on this cpu since the last read

@@ -303,12 +311,6 @@ struct cgroup_rstat_cpu {
 	struct cgroup *updated_next;		/* NULL iff not on the list */
 };
 
-struct cgroup_stat {
-	/* per-cpu statistics are collected into the folowing global counters */
-	struct task_cputime cputime;
-	struct prev_cputime prev_cputime;
-};
-
 struct cgroup {
 	/* self css with NULL ->ss, points back to this cgroup */
 	struct cgroup_subsys_state self;

@@ -412,8 +414,9 @@ struct cgroup {
 	struct cgroup_rstat_cpu __percpu *rstat_cpu;
 
 	/* cgroup basic resource statistics */
-	struct cgroup_stat pending_stat;	/* pending from children */
-	struct cgroup_stat stat;
+	struct cgroup_base_stat pending_bstat;	/* pending from children */
+	struct cgroup_base_stat bstat;
+	struct prev_cputime prev_cputime;	/* for printing out cputime */
 
 	/*
 	 * list of pidlists, up to two for each namespace (one for procs, one

index 0927111..aab4d0a 100644
@@ -206,7 +206,7 @@ int cgroup_task_count(const struct cgroup *cgrp);
 void cgroup_rstat_flush(struct cgroup *cgrp);
 int cgroup_rstat_init(struct cgroup *cgrp);
 void cgroup_rstat_exit(struct cgroup *cgrp);
-void cgroup_stat_show_cputime(struct seq_file *seq);
+void cgroup_base_stat_cputime_show(struct seq_file *seq);
 void cgroup_rstat_boot(void);
 
 /*

index 5549a7c..0d3d093 100644
@@ -54,6 +54,7 @@
 #include <linux/proc_ns.h>
 #include <linux/nsproxy.h>
 #include <linux/file.h>
+#include <linux/sched/cputime.h>
 #include <net/sock.h>
 
 #define CREATE_TRACE_POINTS

@@ -1859,6 +1860,7 @@ static void init_cgroup_housekeeping(struct cgroup *cgrp)
 	cgrp->dom_cgrp = cgrp;
 	cgrp->max_descendants = INT_MAX;
 	cgrp->max_depth = INT_MAX;
+	prev_cputime_init(&cgrp->prev_cputime);
 
 	for_each_subsys(ss, ssid)
 		INIT_LIST_HEAD(&cgrp->e_csets[ssid]);

@@ -3396,7 +3398,7 @@ static int cpu_stat_show(struct seq_file *seq, void *v)
 	struct cgroup __maybe_unused *cgrp = seq_css(seq)->cgroup;
 	int ret = 0;
 
-	cgroup_stat_show_cputime(seq);
+	cgroup_base_stat_cputime_show(seq);
 #ifdef CONFIG_CGROUP_SCHED
 	ret = cgroup_extra_stat_show(seq, cgrp, cpu_cgrp_id);
 #endif

156 lines not on this page · raw message

index aab4d0a..2bf6fb4 100644
@@ -206,8 +206,8 @@ int cgroup_task_count(const struct cgroup *cgrp);
 void cgroup_rstat_flush(struct cgroup *cgrp);
 int cgroup_rstat_init(struct cgroup *cgrp);
 void cgroup_rstat_exit(struct cgroup *cgrp);
-void cgroup_base_stat_cputime_show(struct seq_file *seq);
 void cgroup_rstat_boot(void);
+void cgroup_base_stat_cputime_show(struct seq_file *seq);
 
 /*
  * namespace.c

203 lines not on this page · raw message

index 473e0c0..5c6018f 100644
@@ -690,11 +690,18 @@ static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
 	char *buf, size_t buflen) {}
 #endif /* !CONFIG_CGROUPS */
 
+#ifdef CONFIG_CGROUPS
 /*
- * Basic resource stats.
+ * cgroup scalable recursive statistics.
  */
-#ifdef CONFIG_CGROUPS
+void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
+void cgroup_rstat_flush(struct cgroup *cgrp);
+void cgroup_rstat_flush_hold(struct cgroup *cgrp);
+void cgroup_rstat_flush_release(void);
 
+/*
+ * Basic resource stats.
+ */
 #ifdef CONFIG_CGROUP_CPUACCT
 void cpuacct_charge(struct task_struct *tsk, u64 cputime);
 void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);

index 2bf6fb4..b68e1a7 100644
@@ -203,7 +203,6 @@ int cgroup_task_count(const struct cgroup *cgrp);
 /*
  * rstat.c
  */
-void cgroup_rstat_flush(struct cgroup *cgrp);
 int cgroup_rstat_init(struct cgroup *cgrp);
 void cgroup_rstat_exit(struct cgroup *cgrp);
 void cgroup_rstat_boot(void);

85 lines not on this page · raw message

index 473e0c0..5c6018f 100644
@@ -690,11 +690,18 @@ static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
 	char *buf, size_t buflen) {}
 #endif /* !CONFIG_CGROUPS */
 
+#ifdef CONFIG_CGROUPS
 /*
- * Basic resource stats.
+ * cgroup scalable recursive statistics.
  */
-#ifdef CONFIG_CGROUPS
+void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
+void cgroup_rstat_flush(struct cgroup *cgrp);
+void cgroup_rstat_flush_hold(struct cgroup *cgrp);
+void cgroup_rstat_flush_release(void);
 
+/*
+ * Basic resource stats.
+ */
 #ifdef CONFIG_CGROUP_CPUACCT
 void cpuacct_charge(struct task_struct *tsk, u64 cputime);
 void cpuacct_account_field(struct task_struct *tsk, int index, u64 val);

index 2bf6fb4..b68e1a7 100644
@@ -203,7 +203,6 @@ int cgroup_task_count(const struct cgroup *cgrp);
 /*
  * rstat.c
  */
-void cgroup_rstat_flush(struct cgroup *cgrp);
 int cgroup_rstat_init(struct cgroup *cgrp);
 void cgroup_rstat_exit(struct cgroup *cgrp);
 void cgroup_rstat_boot(void);

92 lines not on this page · raw message

index 5c6018f..c9fdf6f 100644
@@ -696,6 +696,7 @@ static inline void cgroup_path_from_kernfs_id(const union kernfs_node_id *id,
  */
 void cgroup_rstat_updated(struct cgroup *cgrp, int cpu);
 void cgroup_rstat_flush(struct cgroup *cgrp);
+void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
 void cgroup_rstat_flush_hold(struct cgroup *cgrp);
 void cgroup_rstat_flush_release(void);
 

105 lines not on this page · raw message

index cf9db7b..8b8dd17 100644
@@ -130,6 +130,9 @@ struct cgroup_subsys_state {
 	struct list_head sibling;
 	struct list_head children;
 
+	/* flush target list anchored at cgrp->rstat_css_list */
+	struct list_head rstat_css_node;
+
 	/*
 	 * PI: Subsys-unique ID.  0 is unused and root is always 1.  The
 	 * matching css can be looked up using css_from_id().

@@ -412,6 +415,7 @@ struct cgroup {
 
 	/* per-cpu recursive resource statistics */
 	struct cgroup_rstat_cpu __percpu *rstat_cpu;
+	struct list_head rstat_css_list;
 
 	/* cgroup basic resource statistics */
 	struct cgroup_base_stat pending_bstat;	/* pending from children */

@@ -577,6 +581,7 @@ struct cgroup_subsys {
 	void (*css_released)(struct cgroup_subsys_state *css);
 	void (*css_free)(struct cgroup_subsys_state *css);
 	void (*css_reset)(struct cgroup_subsys_state *css);
+	void (*css_rstat_flush)(struct cgroup_subsys_state *css, int cpu);
 	int (*css_extra_stat_show)(struct seq_file *seq,
 				   struct cgroup_subsys_state *css);
 

index 0d3d093..f29f84e 100644
@@ -1860,6 +1860,7 @@ static void init_cgroup_housekeeping(struct cgroup *cgrp)
 	cgrp->dom_cgrp = cgrp;
 	cgrp->max_descendants = INT_MAX;
 	cgrp->max_depth = INT_MAX;
+	INIT_LIST_HEAD(&cgrp->rstat_css_list);
 	prev_cputime_init(&cgrp->prev_cputime);
 
 	for_each_subsys(ss, ssid)

@@ -4630,6 +4631,11 @@ static void css_release_work_fn(struct work_struct *work)
 
 	if (ss) {
 		/* css release path */
+		if (!list_empty(&css->rstat_css_node)) {
+			cgroup_rstat_flush(cgrp);
+			list_del_rcu(&css->rstat_css_node);
+		}
+
 		cgroup_idr_replace(&ss->css_idr, NULL, css->id);
 		if (ss->css_released)
 			ss->css_released(css);

@@ -4690,6 +4696,7 @@ static void init_and_link_css(struct cgroup_subsys_state *css,
 	css->id = -1;
 	INIT_LIST_HEAD(&css->sibling);
 	INIT_LIST_HEAD(&css->children);
+	INIT_LIST_HEAD(&css->rstat_css_node);
 	css->serial_nr = css_serial_nr_next++;
 	atomic_set(&css->online_cnt, 0);
 

@@ -4698,6 +4705,9 @@ static void init_and_link_css(struct cgroup_subsys_state *css,
 		css_get(css->parent);
 	}
 
+	if (cgroup_on_dfl(cgrp) && ss->css_rstat_flush)
+		list_add_rcu(&css->rstat_css_node, &cgrp->rstat_css_list);
+
 	BUG_ON(cgroup_css(cgrp, ss));
 }
 

@@ -4799,6 +4809,7 @@ static struct cgroup_subsys_state *css_create(struct cgroup *cgrp,
 err_list_del:
 	list_del_rcu(&css->sibling);
 err_free_css:
+	list_del_rcu(&css->rstat_css_node);
 	call_rcu(&css->rcu_head, css_free_rcu_fn);
 	return ERR_PTR(err);
 }

index a5f9338..18b464a 100644
@@ -143,8 +143,17 @@ static void cgroup_rstat_flush_locked(struct cgroup *cgrp, bool may_sleep)
 		struct cgroup *pos = NULL;
 
 		raw_spin_lock(cpu_lock);
-		while ((pos = cgroup_rstat_cpu_pop_updated(pos, cgrp, cpu)))
+		while ((pos = cgroup_rstat_cpu_pop_updated(pos, cgrp, cpu))) {
+			struct cgroup_subsys_state *css;
+
 			cgroup_base_stat_flush(pos, cpu);
+
+			rcu_read_lock();
+			list_for_each_entry_rcu(css, &pos->rstat_css_list,
+						rstat_css_node)
+				css->ss->css_rstat_flush(css, cpu);
+			rcu_read_unlock();
+		}
 		raw_spin_unlock(cpu_lock);
 
 		/* if @may_sleep, play nice and yield if necessary */

index 18b464a..662d7ae 100644
@@ -28,9 +28,12 @@ void cgroup_rstat_updated(struct cgroup *cgrp, int cpu)
 	unsigned long flags;
 
 	/*
-	 * Speculative already-on-list test.  This may race leading to
-	 * temporary inaccuracies, which is fine.
-	 *
+	 * Paired with the one in cgroup_rstat_cpu_pop_upated().  Either we
+	 * see NULL updated_next or they see our updated stat.
+	 */
+	smp_mb();
+
+	/*
 	 * Because @parent's updated_children is terminated with @parent
 	 * instead of NULL, we can tell whether @cgrp is on the list by
 	 * testing the next pointer for NULL.

@@ -125,6 +128,13 @@ static struct cgroup *cgroup_rstat_cpu_pop_updated(struct cgroup *pos,
 
 		*nextp = rstatc->updated_next;
 		rstatc->updated_next = NULL;
+
+		/*
+		 * Paired with the one in cgroup_rstat_cpu_updated().
+		 * Either they see NULL updated_next or we see their
+		 * updated stat.
+		 */
+		smp_mb();
 	}
 
 	return pos;

index 0d1d5fc..9a30b51 100644
@@ -27,6 +27,10 @@ void cgroup_rstat_updated(struct cgroup *cgrp, int cpu)
 	struct cgroup *parent;
 	unsigned long flags;
 
+	/* nothing to do for root */
+	if (!cgroup_parent(cgrp))
+		return;
+
 	/*
 	 * Paired with the one in cgroup_rstat_cpu_pop_upated().  Either we
 	 * see NULL updated_next or they see our updated stat.