strparser: pad sk_skb_cb to avoid straddling cachelines

sk_skb_cb lives within skb->cb[]. skb->cb[] straddles
2 cache lines, each containing 24B of data.
The first cache line does not contain much interesting
information for users of strparser, so pad things a little.
Previously strp_msg->full_len would live in the first cache
line and strp_msg->offset in the second.

We need to reorder the 8 byte temp_reg with struct tls_msg
to prevent a 4B hole which would push the struct over 48B.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

include/net/strparser.h

@@ -65,15 +65,19 @@ struct _strp_msg {
 struct sk_skb_cb {
 #define SK_SKB_CB_PRIV_LEN 20
 	unsigned char data[SK_SKB_CB_PRIV_LEN];
+	/* align strp on cache line boundary within skb->cb[] */
+	unsigned char pad[4];
 	struct _strp_msg strp;
-	/* temp_reg is a temporary register used for bpf_convert_data_end_access
-	 * when dst_reg == src_reg.
-	 */
-	u64 temp_reg;
+
+	/* strp users' data follows */
 	struct tls_msg {
 		u8 control;
 		u8 decrypted;
 	} tls;
+	/* temp_reg is a temporary register used for bpf_convert_data_end_access
+	 * when dst_reg == src_reg.
+	 */
+	u64 temp_reg;
 };
 
 static inline struct strp_msg *strp_msg(struct sk_buff *skb)

net/strparser/strparser.c

@@ -533,6 +533,9 @@ EXPORT_SYMBOL_GPL(strp_check_rcv);
 
 static int __init strp_dev_init(void)
 {
+	BUILD_BUG_ON(sizeof(struct sk_skb_cb) >
+		     sizeof_field(struct sk_buff, cb));
+
 	strp_wq = create_singlethread_workqueue("kstrp");
 	if (unlikely(!strp_wq))
 		return -ENOMEM;