net: switch secpath to use skb extension infrastructure

Remove skb->sp and allocate secpath storage via extension
infrastructure.  This also reduces sk_buff by 8 bytes on x86_64.

Total size of allyesconfig kernel is reduced slightly, as there is
less inlined code (one conditional atomic op instead of two on
skb_clone).

No differences in throughput in following ipsec performance tests:
- transport mode with aes on 10GB link
- tunnel mode between two network namespaces with aes and null cipher

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 59 insertions, 83 deletions

diff --git a/Documentation/networking/xfrm_device.txt b/Documentation/networking/xfrm_device.txt
index 267f55b5f54a..a1c904dc70dc 100644
--- a/Documentation/networking/xfrm_device.txt
+++ b/Documentation/networking/xfrm_device.txt
@@ -111,9 +111,10 @@ the stack in xfrm_input().
 		xfrm_state_hold(xs);
 
 	store the state information into the skb
-		skb->sp = secpath_dup(skb->sp);
-		skb->sp->xvec[skb->sp->len++] = xs;
-		skb->sp->olen++;
+		sp = secpath_set(skb);
+		if (!sp) return;
+		sp->xvec[sp->len++] = xs;
+		sp->olen++;
 
 	indicate the success and/or error status of the offload
 		xo = xfrm_offload(skb);
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index d0f254a016bf..3f741b04e55d 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -714,9 +714,6 @@ struct sk_buff {
 		struct list_head	tcp_tsorted_anchor;
 	};
 
-#ifdef CONFIG_XFRM
-	struct	sec_path	*sp;
-#endif
 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 	unsigned long		 _nfct;
 #endif
@@ -3908,6 +3905,9 @@ enum skb_ext_id {
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 	SKB_EXT_BRIDGE_NF,
 #endif
+#ifdef CONFIG_XFRM
+	SKB_EXT_SEC_PATH,
+#endif
 	SKB_EXT_NUM, /* must be last */
 };
 
@@ -4069,7 +4069,7 @@ static inline void skb_init_secmark(struct sk_buff *skb)
 static inline int secpath_exists(const struct sk_buff *skb)
 {
 #ifdef CONFIG_XFRM
-	return skb->sp != NULL;
+	return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
 #else
 	return 0;
 #endif
@@ -4127,7 +4127,7 @@ static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
 {
 #ifdef CONFIG_XFRM
-	return skb->sp;
+	return skb_ext_find(skb, SKB_EXT_SEC_PATH);
 #else
 	return NULL;
 #endif
diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index 31220edcce95..38c232861a64 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -1096,7 +1096,6 @@ struct xfrm_offload {
 };
 
 struct sec_path {
-	refcount_t		refcnt;
 	int			len;
 	int			olen;
 
@@ -1104,32 +1103,13 @@ struct sec_path {
 	struct xfrm_offload	ovec[XFRM_MAX_OFFLOAD_DEPTH];
 };
 
-static inline struct sec_path *
-secpath_get(struct sec_path *sp)
-{
-	if (sp)
-		refcount_inc(&sp->refcnt);
-	return sp;
-}
-
-void __secpath_destroy(struct sec_path *sp);
-
-static inline void
-secpath_put(struct sec_path *sp)
-{
-	if (sp && refcount_dec_and_test(&sp->refcnt))
-		__secpath_destroy(sp);
-}
-
-struct sec_path *secpath_dup(struct sec_path *src);
 struct sec_path *secpath_set(struct sk_buff *skb);
 
 static inline void
 secpath_reset(struct sk_buff *skb)
 {
 #ifdef CONFIG_XFRM
-	secpath_put(skb->sp);
-	skb->sp = NULL;
+	skb_ext_del(skb, SKB_EXT_SEC_PATH);
 #endif
 }
 
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 0c65723591d7..cb0bf4215745 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -609,7 +609,6 @@ fastpath:
 void skb_release_head_state(struct sk_buff *skb)
 {
 	skb_dst_drop(skb);
-	secpath_reset(skb);
 	if (skb->destructor) {
 		WARN_ON(in_irq());
 		skb->destructor(skb);
@@ -798,9 +797,6 @@ static void __copy_skb_header(struct sk_buff *new, const struct sk_buff *old)
 	memcpy(new->cb, old->cb, sizeof(old->cb));
 	skb_dst_copy(new, old);
 	__skb_ext_copy(new, old);
-#ifdef CONFIG_XFRM
-	new->sp			= secpath_get(old->sp);
-#endif
 	__nf_copy(new, old, false);
 
 	/* Note : this field could be in headers_start/headers_end section
@@ -3912,6 +3908,9 @@ static const u8 skb_ext_type_len[] = {
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 	[SKB_EXT_BRIDGE_NF] = SKB_EXT_CHUNKSIZEOF(struct nf_bridge_info),
 #endif
+#ifdef CONFIG_XFRM
+	[SKB_EXT_SEC_PATH] = SKB_EXT_CHUNKSIZEOF(struct sec_path),
+#endif
 };
 
 static __always_inline unsigned int skb_ext_total_length(void)
@@ -3920,6 +3919,9 @@ static __always_inline unsigned int skb_ext_total_length(void)
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 		skb_ext_type_len[SKB_EXT_BRIDGE_NF] +
 #endif
+#ifdef CONFIG_XFRM
+		skb_ext_type_len[SKB_EXT_SEC_PATH] +
+#endif
 		0;
 }
 
@@ -5610,7 +5612,8 @@ static struct skb_ext *skb_ext_alloc(void)
 	return new;
 }
 
-static struct skb_ext *skb_ext_maybe_cow(struct skb_ext *old)
+static struct skb_ext *skb_ext_maybe_cow(struct skb_ext *old,
+					 unsigned int old_active)
 {
 	struct skb_ext *new;
 
@@ -5624,6 +5627,15 @@ static struct skb_ext *skb_ext_maybe_cow(struct skb_ext *old)
 	memcpy(new, old, old->chunks * SKB_EXT_ALIGN_VALUE);
 	refcount_set(&new->refcnt, 1);
 
+#ifdef CONFIG_XFRM
+	if (old_active & (1 << SKB_EXT_SEC_PATH)) {
+		struct sec_path *sp = skb_ext_get_ptr(old, SKB_EXT_SEC_PATH);
+		unsigned int i;
+
+		for (i = 0; i < sp->len; i++)
+			xfrm_state_hold(sp->xvec[i]);
+	}
+#endif
 	__skb_ext_put(old);
 	return new;
 }
@@ -5650,7 +5662,7 @@ void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id)
 	if (skb->active_extensions) {
 		old = skb->extensions;
 
-		new = skb_ext_maybe_cow(old);
+		new = skb_ext_maybe_cow(old, skb->active_extensions);
 		if (!new)
 			return NULL;
 
@@ -5679,6 +5691,16 @@ set_active:
 }
 EXPORT_SYMBOL(skb_ext_add);
 
+#ifdef CONFIG_XFRM
+static void skb_ext_put_sp(struct sec_path *sp)
+{
+	unsigned int i;
+
+	for (i = 0; i < sp->len; i++)
+		xfrm_state_put(sp->xvec[i]);
+}
+#endif
+
 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
 {
 	struct skb_ext *ext = skb->extensions;
@@ -5687,6 +5709,14 @@ void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
 	if (skb->active_extensions == 0) {
 		skb->extensions = NULL;
 		__skb_ext_put(ext);
+#ifdef CONFIG_XFRM
+	} else if (id == SKB_EXT_SEC_PATH &&
+		   refcount_read(&ext->refcnt) == 1) {
+		struct sec_path *sp = skb_ext_get_ptr(ext, SKB_EXT_SEC_PATH);
+
+		skb_ext_put_sp(sp);
+		sp->len = 0;
+#endif
 	}
 }
 EXPORT_SYMBOL(__skb_ext_del);
@@ -5702,6 +5732,11 @@ void __skb_ext_put(struct skb_ext *ext)
 	if (!refcount_dec_and_test(&ext->refcnt))
 		return;
 free_now:
+#ifdef CONFIG_XFRM
+	if (__skb_ext_exist(ext, SKB_EXT_SEC_PATH))
+		skb_ext_put_sp(skb_ext_get_ptr(ext, SKB_EXT_SEC_PATH));
+#endif
+
 	kmem_cache_free(skbuff_ext_cache, ext);
 }
 EXPORT_SYMBOL(__skb_ext_put);
diff --git a/net/xfrm/xfrm_input.c b/net/xfrm/xfrm_input.c
index b4db25b244fa..6bc817359b58 100644
--- a/net/xfrm/xfrm_input.c
+++ b/net/xfrm/xfrm_input.c
@@ -38,8 +38,6 @@ struct xfrm_trans_cb {
 
 #define XFRM_TRANS_SKB_CB(__skb) ((struct xfrm_trans_cb *)&((__skb)->cb[0]))
 
-static struct kmem_cache *secpath_cachep __ro_after_init;
-
 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
 static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[AF_INET6 + 1];
 
@@ -111,54 +109,21 @@ static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
 	return ret;
 }
 
-void __secpath_destroy(struct sec_path *sp)
-{
-	int i;
-	for (i = 0; i < sp->len; i++)
-		xfrm_state_put(sp->xvec[i]);
-	kmem_cache_free(secpath_cachep, sp);
-}
-EXPORT_SYMBOL(__secpath_destroy);
-
-struct sec_path *secpath_dup(struct sec_path *src)
+struct sec_path *secpath_set(struct sk_buff *skb)
 {
-	struct sec_path *sp;
+	struct sec_path *sp, *tmp = skb_ext_find(skb, SKB_EXT_SEC_PATH);
 
-	sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
+	sp = skb_ext_add(skb, SKB_EXT_SEC_PATH);
 	if (!sp)
 		return NULL;
 
-	sp->len = 0;
-	sp->olen = 0;
+	if (tmp) /* reused existing one (was COW'd if needed) */
+		return sp;
 
+	/* allocated new secpath */
 	memset(sp->ovec, 0, sizeof(sp->ovec));
-
-	if (src) {
-		int i;
-
-		memcpy(sp, src, sizeof(*sp));
-		for (i = 0; i < sp->len; i++)
-			xfrm_state_hold(sp->xvec[i]);
-	}
-	refcount_set(&sp->refcnt, 1);
-	return sp;
-}
-EXPORT_SYMBOL(secpath_dup);
-
-struct sec_path *secpath_set(struct sk_buff *skb)
-{
-	struct sec_path *sp = skb->sp;
-
-	/* Allocate new secpath or COW existing one. */
-	if (!sp || refcount_read(&sp->refcnt) != 1) {
-		sp = secpath_dup(skb->sp);
-		if (!sp)
-			return NULL;
-
-		if (skb->sp)
-			secpath_put(skb->sp);
-		skb->sp = sp;
-	}
+	sp->olen = 0;
+	sp->len = 0;
 
 	return sp;
 }
@@ -552,11 +517,6 @@ void __init xfrm_input_init(void)
 	if (err)
 		gro_cells.cells = NULL;
 
-	secpath_cachep = kmem_cache_create("secpath_cache",
-					   sizeof(struct sec_path),
-					   0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
-					   NULL);
-
 	for_each_possible_cpu(i) {
 		struct xfrm_trans_tasklet *trans;