cgroup, perf_event: make perf_event work on v2 hierarchy

index aae8c94..0af8dc4 100644
@@ -450,6 +450,19 @@ struct cgroup_subsys {
 	bool early_init:1;
 
 	/*
+	 * If %true, the controller, on the default hierarchy, doesn't show
+	 * up in "cgroup.controllers" or "cgroup.subtree_control", is
+	 * implicitly enabled on all cgroups on the default hierarchy, and
+	 * bypasses the "no internal process" constraint.  This is for
+	 * utility type controllers which is transparent to userland.
+	 *
+	 * An implicit controller can be stolen from the default hierarchy
+	 * anytime and thus must be okay with offline csses from previous
+	 * hierarchies coexisting with csses for the current one.
+	 */
+	bool implicit_on_dfl:1;
+
+	/*
 	 * If %false, this subsystem is properly hierarchical -
 	 * configuration, resource accounting and restriction on a parent
 	 * cgroup cover those of its children.  If %true, hierarchy support

index a44c123..0f17c65 100644
@@ -186,6 +186,9 @@ static u16 cgroup_no_v1_mask;
 /* some controllers are not supported in the default hierarchy */
 static u16 cgrp_dfl_inhibit_ss_mask;
 
+/* some controllers are implicitly enabled on the default hierarchy */
+static unsigned long cgrp_dfl_implicit_ss_mask;
+
 /* The list of hierarchy roots */
 
 static LIST_HEAD(cgroup_roots);

@@ -359,8 +362,8 @@ static u16 cgroup_control(struct cgroup *cgrp)
 		return parent->subtree_control;
 
 	if (cgroup_on_dfl(cgrp))
-		root_ss_mask &= ~cgrp_dfl_inhibit_ss_mask;
-
+		root_ss_mask &= ~(cgrp_dfl_inhibit_ss_mask |
+				  cgrp_dfl_implicit_ss_mask);
 	return root_ss_mask;
 }
 

@@ -1326,6 +1329,8 @@ static u16 cgroup_calc_subtree_ss_mask(u16 subtree_control, u16 this_ss_mask)
 
 	lockdep_assert_held(&cgroup_mutex);
 
+	cur_ss_mask |= cgrp_dfl_implicit_ss_mask;
+
 	while (true) {
 		u16 new_ss_mask = cur_ss_mask;
 

@@ -1511,8 +1516,13 @@ static int rebind_subsystems(struct cgroup_root *dst_root, u16 ss_mask)
 	lockdep_assert_held(&cgroup_mutex);
 
 	do_each_subsys_mask(ss, ssid, ss_mask) {
-		/* if @ss has non-root csses attached to it, can't move */
-		if (css_next_child(NULL, cgroup_css(&ss->root->cgrp, ss)))
+		/*
+		 * If @ss has non-root csses attached to it, can't move.
+		 * If @ss is an implicit controller, it is exempt from this
+		 * rule and can be stolen.
+		 */
+		if (css_next_child(NULL, cgroup_css(&ss->root->cgrp, ss)) &&
+		    !ss->implicit_on_dfl)
 			return -EBUSY;
 
 		/* can't move between two non-dummy roots either */

@@ -3032,6 +3042,18 @@ static void cgroup_restore_control(struct cgroup *cgrp)
 	}
 }
 
+static bool css_visible(struct cgroup_subsys_state *css)
+{
+	struct cgroup_subsys *ss = css->ss;
+	struct cgroup *cgrp = css->cgroup;
+
+	if (cgroup_control(cgrp) & (1 << ss->id))
+		return true;
+	if (!(cgroup_ss_mask(cgrp) & (1 << ss->id)))
+		return false;
+	return cgroup_on_dfl(cgrp) && ss->implicit_on_dfl;
+}
+
 /**
  * cgroup_apply_control_enable - enable or show csses according to control
  * @cgrp: root of the target subtree

@@ -3067,7 +3089,7 @@ static int cgroup_apply_control_enable(struct cgroup *cgrp)
 					return PTR_ERR(css);
 			}
 
-			if (cgroup_control(dsct) & (1 << ss->id)) {
+			if (css_visible(css)) {
 				ret = css_populate_dir(css);
 				if (ret)
 					return ret;

@@ -3110,7 +3132,7 @@ static void cgroup_apply_control_disable(struct cgroup *cgrp)
 			if (css->parent &&
 			    !(cgroup_ss_mask(dsct) & (1 << ss->id))) {
 				kill_css(css);
-			} else if (!(cgroup_control(dsct) & (1 << ss->id))) {
+			} else if (!css_visible(css)) {
 				css_clear_dir(css);
 				if (ss->css_reset)
 					ss->css_reset(css);

@@ -5440,7 +5462,9 @@ int __init cgroup_init(void)
 
 		cgrp_dfl_root.subsys_mask |= 1 << ss->id;
 
-		if (!ss->dfl_cftypes)
+		if (ss->implicit_on_dfl)
+			cgrp_dfl_implicit_ss_mask |= 1 << ss->id;
+		else if (!ss->dfl_cftypes)
 			cgrp_dfl_inhibit_ss_mask |= 1 << ss->id;
 
 		if (ss->dfl_cftypes == ss->legacy_cftypes) {

index c095741..e77fa4f 100644
@@ -9416,5 +9416,11 @@ struct cgroup_subsys perf_event_cgrp_subsys = {
 	.css_alloc	= perf_cgroup_css_alloc,
 	.css_free	= perf_cgroup_css_free,
 	.attach		= perf_cgroup_attach,
+	/*
+	 * Implicitly enable on dfl hierarchy so that perf events can
+	 * always be filtered by cgroup2 path as long as perf_event
+	 * controller is not mounted on a legacy hierarchy.
+	 */
+	.implicit_on_dfl = true,
 };
 #endif /* CONFIG_CGROUP_PERF */

index 90aa1b4..5df6235 100644
@@ -12,8 +12,8 @@ cgroupfs_find_mountpoint(char *buf, size_t maxlen)
 {
 	FILE *fp;
 	char mountpoint[PATH_MAX + 1], tokens[PATH_MAX + 1], type[PATH_MAX + 1];
+	char path_v1[PATH_MAX + 1], path_v2[PATH_MAX + 2], *path;
 	char *token, *saved_ptr = NULL;
-	int found = 0;
 
 	fp = fopen("/proc/mounts", "r");
 	if (!fp)

@@ -24,31 +24,43 @@ cgroupfs_find_mountpoint(char *buf, size_t maxlen)
 	 * and inspect every cgroupfs mount point to find one that has
 	 * perf_event subsystem
 	 */
+	path_v1[0] = '\0';
+	path_v2[0] = '\0';
+
 	while (fscanf(fp, "%*s %"STR(PATH_MAX)"s %"STR(PATH_MAX)"s %"
 				STR(PATH_MAX)"s %*d %*d\n",
 				mountpoint, type, tokens) == 3) {
 
-		if (!strcmp(type, "cgroup")) {
+		if (!path_v1[0] && !strcmp(type, "cgroup")) {
 
 			token = strtok_r(tokens, ",", &saved_ptr);
 
 			while (token != NULL) {
 				if (!strcmp(token, "perf_event")) {
-					found = 1;
+					strcpy(path_v1, mountpoint);
 					break;
 				}
 				token = strtok_r(NULL, ",", &saved_ptr);
 			}
 		}
-		if (found)
+
+		if (!path_v2[0] && !strcmp(type, "cgroup2"))
+			strcpy(path_v2, mountpoint);
+
+		if (path_v1[0] && path_v2[0])
 			break;
 	}
 	fclose(fp);
-	if (!found)
+
+	if (path_v1[0])
+		path = path_v1;
+	else if (path_v2[0])
+		path = path_v2;
+	else
 		return -1;
 
-	if (strlen(mountpoint) < maxlen) {
-		strcpy(buf, mountpoint);
+	if (strlen(path) < maxlen) {
+		strcpy(buf, path);
 		return 0;
 	}
 	return -1;

@@ -10792,5 +10792,11 @@ struct cgroup_subsys perf_event_cgrp_sub
 	.css_alloc	= perf_cgroup_css_alloc,
 	.css_free	= perf_cgroup_css_free,
 	.attach		= perf_cgroup_attach,
+	/*
+	 * Implicitly enable on dfl hierarchy so that perf events can
+	 * always be filtered by cgroup2 path as long as perf_event
+	 * controller is not mounted on a legacy hierarchy.
+	 */
+	.implicit_on_dfl = true,
 };
 #endif /* CONFIG_CGROUP_PERF */

@@ -12,8 +12,8 @@ cgroupfs_find_mountpoint(char *buf, size
 {
 	FILE *fp;
 	char mountpoint[PATH_MAX + 1], tokens[PATH_MAX + 1], type[PATH_MAX + 1];
+	char path_v1[PATH_MAX + 1], path_v2[PATH_MAX + 2], *path;
 	char *token, *saved_ptr = NULL;
-	int found = 0;
 
 	fp = fopen("/proc/mounts", "r");
 	if (!fp)

@@ -24,31 +24,43 @@ cgroupfs_find_mountpoint(char *buf, size
 	 * and inspect every cgroupfs mount point to find one that has
 	 * perf_event subsystem
 	 */
+	path_v1[0] = '\0';
+	path_v2[0] = '\0';
+
 	while (fscanf(fp, "%*s %"STR(PATH_MAX)"s %"STR(PATH_MAX)"s %"
 				STR(PATH_MAX)"s %*d %*d\n",
 				mountpoint, type, tokens) == 3) {
 
-		if (!strcmp(type, "cgroup")) {
+		if (!path_v1[0] && !strcmp(type, "cgroup")) {
 
 			token = strtok_r(tokens, ",", &saved_ptr);
 
 			while (token != NULL) {
 				if (!strcmp(token, "perf_event")) {
-					found = 1;
+					strcpy(path_v1, mountpoint);
 					break;
 				}
 				token = strtok_r(NULL, ",", &saved_ptr);
 			}
 		}
-		if (found)
+
+		if (!path_v2[0] && !strcmp(type, "cgroup2"))
+			strcpy(path_v2, mountpoint);
+
+		if (path_v1[0] && path_v2[0])
 			break;
 	}
 	fclose(fp);
-	if (!found)
+
+	if (path_v1[0])
+		path = path_v1;
+	else if (path_v2[0])
+		path = path_v2;
+	else
 		return -1;
 
-	if (strlen(mountpoint) < maxlen) {
-		strcpy(buf, mountpoint);
+	if (strlen(path) < maxlen) {
+		strcpy(buf, path);
 		return 0;
 	}
 	return -1;