diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index ed128fed..9dc2263 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -544,31 +544,107 @@
 
 #ifdef CONFIG_SOCK_CGROUP_DATA
 
+/*
+ * sock_cgroup_data is embedded at sock->sk_cgrp_data and contains
+ * per-socket cgroup information except for memcg association.
+ *
+ * On legacy hierarchies, net_prio and net_cls controllers directly set
+ * attributes on each sock which can then be tested by the network layer.
+ * On the default hierarchy, each sock is associated with the cgroup it was
+ * created in and the networking layer can match the cgroup directly.
+ *
+ * To avoid carrying all three cgroup related fields separately in sock,
+ * sock_cgroup_data overloads (prioidx, classid) and the cgroup pointer.
+ * On boot, sock_cgroup_data records the cgroup that the sock was created
+ * in so that cgroup2 matches can be made; however, once either net_prio or
+ * net_cls starts being used, the area is overriden to carry prioidx and/or
+ * classid.  The two modes are distinguished by whether the lowest bit is
+ * set.  Clear bit indicates cgroup pointer while set bit prioidx and
+ * classid.
+ *
+ * While userland may start using net_prio or net_cls at any time, once
+ * either is used, cgroup2 matching no longer works.  There is no reason to
+ * mix the two and this is in line with how legacy and v2 compatibility is
+ * handled.  On mode switch, cgroup references which are already being
+ * pointed to by socks may be leaked.  While this can be remedied by adding
+ * synchronization around sock_cgroup_data, given that the number of leaked
+ * cgroups is bound and highly unlikely to be high, this seems to be the
+ * better trade-off.
+ */
 struct sock_cgroup_data {
-	u16	prioidx;
-	u32	classid;
+	union {
+#ifdef __LITTLE_ENDIAN
+		struct {
+			u8	is_data;
+			u8	padding;
+			u16	prioidx;
+			u32	classid;
+		} __packed;
+#else
+		struct {
+			u32	classid;
+			u16	prioidx;
+			u8	padding;
+			u8	is_data;
+		} __packed;
+#endif
+		u64		val;
+	};
 };
 
+/*
+ * There's a theoretical window where the following accessors race with
+ * updaters and return part of the previous pointer as the prioidx or
+ * classid.  Such races are short-lived and the result isn't critical.
+ */
 static inline u16 sock_cgroup_prioidx(struct sock_cgroup_data *skcd)
 {
-	return skcd->prioidx;
+	/* fallback to 1 which is always the ID of the root cgroup */
+	return (skcd->is_data & 1) ? skcd->prioidx : 1;
 }
 
 static inline u32 sock_cgroup_classid(struct sock_cgroup_data *skcd)
 {
-	return skcd->classid;
+	/* fallback to 0 which is the unconfigured default classid */
+	return (skcd->is_data & 1) ? skcd->classid : 0;
 }
 
+/*
+ * If invoked concurrently, the updaters may clobber each other.  The
+ * caller is responsible for synchronization.
+ */
 static inline void sock_cgroup_set_prioidx(struct sock_cgroup_data *skcd,
 					   u16 prioidx)
 {
-	skcd->prioidx = prioidx;
+	struct sock_cgroup_data skcd_buf = { .val = READ_ONCE(skcd->val) };
+
+	if (sock_cgroup_prioidx(&skcd_buf) == prioidx)
+		return;
+
+	if (!(skcd_buf.is_data & 1)) {
+		skcd_buf.val = 0;
+		skcd_buf.is_data = 1;
+	}
+
+	skcd_buf.prioidx = prioidx;
+	WRITE_ONCE(skcd->val, skcd_buf.val);	/* see sock_cgroup_ptr() */
 }
 
 static inline void sock_cgroup_set_classid(struct sock_cgroup_data *skcd,
 					   u32 classid)
 {
-	skcd->classid = classid;
+	struct sock_cgroup_data skcd_buf = { .val = READ_ONCE(skcd->val) };
+
+	if (sock_cgroup_classid(&skcd_buf) == classid)
+		return;
+
+	if (!(skcd_buf.is_data & 1)) {
+		skcd_buf.val = 0;
+		skcd_buf.is_data = 1;
+	}
+
+	skcd_buf.classid = classid;
+	WRITE_ONCE(skcd->val, skcd_buf.val);	/* see sock_cgroup_ptr() */
 }
 
 #else	/* CONFIG_SOCK_CGROUP_DATA */