Merge branch 'net-dst_entry-shrink'

David Miller says:

====================
net: Significantly shrink the size of routes.

Through a combination of several things, our route structures are
larger than they need to be.

Mostly this stems from having members in dst_entry which are only used
by one class of routes.  So the majority of the work in this series is
about "un-commoning" these members and pushing them into the type
specific structures.

Unfortunately, IPSEC needed the most surgery.  The majority of the
changes here had to do with bundle creation and management.

The other issue is the refcount alignment in dst_entry.  Once we get
rid of the not-so-common members, it really opens the door to removing
that alignment entirely.

I think the new layout looks really nice, so I'll reproduce it here:

	struct net_device       *dev;
	struct  dst_ops	        *ops;
	unsigned long		_metrics;
	unsigned long           expires;
	struct xfrm_state	*xfrm;
	int			(*input)(struct sk_buff *);
	int			(*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
	unsigned short		flags;
	short			obsolete;
	unsigned short		header_len;
	unsigned short		trailer_len;
	atomic_t		__refcnt;
	int			__use;
	unsigned long		lastuse;
	struct lwtunnel_state   *lwtstate;
	struct rcu_head		rcu_head;
	short			error;
	short			__pad;
	__u32			tclassid;

(This is for 64-bit, on 32-bit the __refcnt comes at the very end)

So, the good news:

1) struct dst_entry shrinks from 160 to 112 bytes.

2) struct rtable shrinks from 216 to 168 bytes.

3) struct rt6_info shrinks from 384 to 320 bytes.

Enjoy.

v2:
	Collapse some patches logically based upon feedback.
	Fix the strange patch #7.

v3:	xfrm_dst_path() needs inline keyword
	Properly align __refcnt on 32-bit.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/dn_route.h

@@ -69,6 +69,7 @@ int dn_route_rcv(struct sk_buff *skb, struct net_device *dev,
  */
 struct dn_route {
 	struct dst_entry dst;
+	struct dn_route __rcu *dn_next;
 
 	struct neighbour *n;
 

include/net/dst.h

@@ -34,13 +34,9 @@ struct sk_buff;
 
 struct dst_entry {
 	struct net_device       *dev;
-	struct rcu_head		rcu_head;
-	struct dst_entry	*child;
 	struct  dst_ops	        *ops;
 	unsigned long		_metrics;
 	unsigned long           expires;
-	struct dst_entry	*path;
-	struct dst_entry	*from;
 #ifdef CONFIG_XFRM
 	struct xfrm_state	*xfrm;
 #else
@@ -59,8 +55,6 @@ struct dst_entry {
 #define DST_XFRM_QUEUE		0x0040
 #define DST_METADATA		0x0080
 
-	short			error;
-
 	/* A non-zero value of dst->obsolete forces by-hand validation
 	 * of the route entry.  Positive values are set by the generic
 	 * dst layer to indicate that the entry has been forcefully
@@ -76,35 +70,24 @@ struct dst_entry {
 #define DST_OBSOLETE_KILL	-2
 	unsigned short		header_len;	/* more space at head required */
 	unsigned short		trailer_len;	/* space to reserve at tail */
-	unsigned short		__pad3;
 
-#ifdef CONFIG_IP_ROUTE_CLASSID
-	__u32			tclassid;
-#else
-	__u32			__pad2;
-#endif
-
-#ifdef CONFIG_64BIT
-	/*
-	 * Align __refcnt to a 64 bytes alignment
-	 * (L1_CACHE_SIZE would be too much)
-	 */
-	long			__pad_to_align_refcnt[2];
-#endif
 	/*
 	 * __refcnt wants to be on a different cache line from
 	 * input/output/ops or performance tanks badly
 	 */
-	atomic_t		__refcnt;	/* client references	*/
+#ifdef CONFIG_64BIT
+	atomic_t		__refcnt;	/* 64-bit offset 64 */
+#endif
 	int			__use;
 	unsigned long		lastuse;
 	struct lwtunnel_state   *lwtstate;
-	union {
-		struct dst_entry	*next;
-		struct rtable __rcu	*rt_next;
-		struct rt6_info __rcu	*rt6_next;
-		struct dn_route __rcu	*dn_next;
-	};
+	struct rcu_head		rcu_head;
+	short			error;
+	short			__pad;
+	__u32			tclassid;
+#ifndef CONFIG_64BIT
+	atomic_t		__refcnt;	/* 32-bit offset 64 */
+#endif
 };
 
 struct dst_metrics {
@@ -250,7 +233,7 @@ static inline void dst_hold(struct dst_entry *dst)
 {
 	/*
 	 * If your kernel compilation stops here, please check
-	 * __pad_to_align_refcnt declaration in struct dst_entry
+	 * the placement of __refcnt in struct dst_entry
 	 */
 	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 	WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);

include/net/ip6_fib.h

@@ -129,6 +129,8 @@ struct rt6_exception {
 
 struct rt6_info {
 	struct dst_entry		dst;
+	struct rt6_info __rcu		*rt6_next;
+	struct rt6_info			*from;
 
 	/*
 	 * Tail elements of dst_entry (__refcnt etc.)
@@ -176,11 +178,11 @@ struct rt6_info {
 
 #define for_each_fib6_node_rt_rcu(fn)					\
 	for (rt = rcu_dereference((fn)->leaf); rt;			\
-	     rt = rcu_dereference(rt->dst.rt6_next))
+	     rt = rcu_dereference(rt->rt6_next))
 
 #define for_each_fib6_walker_rt(w)					\
 	for (rt = (w)->leaf; rt;					\
-	     rt = rcu_dereference_protected(rt->dst.rt6_next, 1))
+	     rt = rcu_dereference_protected(rt->rt6_next, 1))
 
 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst)
 {
@@ -203,11 +205,9 @@ static inline void rt6_update_expires(struct rt6_info *rt0, int timeout)
 {
 	struct rt6_info *rt;
 
-	for (rt = rt0; rt && !(rt->rt6i_flags & RTF_EXPIRES);
-	     rt = (struct rt6_info *)rt->dst.from);
+	for (rt = rt0; rt && !(rt->rt6i_flags & RTF_EXPIRES); rt = rt->from);
 	if (rt && rt != rt0)
 		rt0->dst.expires = rt->dst.expires;
-
 	dst_set_expires(&rt0->dst, timeout);
 	rt0->rt6i_flags |= RTF_EXPIRES;
 }
@@ -242,8 +242,8 @@ static inline u32 rt6_get_cookie(const struct rt6_info *rt)
 	u32 cookie = 0;
 
 	if (rt->rt6i_flags & RTF_PCPU ||
-	    (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->dst.from))
-		rt = (struct rt6_info *)(rt->dst.from);
+	    (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->from))
+		rt = rt->from;
 
 	rt6_get_cookie_safe(rt, &cookie);
 

include/net/xfrm.h

@@ -968,7 +968,7 @@ static inline bool xfrm_sec_ctx_match(struct xfrm_sec_ctx *s1, struct xfrm_sec_c
 
 /* A struct encoding bundle of transformations to apply to some set of flow.
  *
- * dst->child points to the next element of bundle.
+ * xdst->child points to the next element of bundle.
  * dst->xfrm  points to an instanse of transformer.
  *
  * Due to unfortunate limitations of current routing cache, which we
@@ -984,6 +984,8 @@ struct xfrm_dst {
 		struct rt6_info		rt6;
 	} u;
 	struct dst_entry *route;
+	struct dst_entry *child;
+	struct dst_entry *path;
 	struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
 	int num_pols, num_xfrms;
 	u32 xfrm_genid;
@@ -994,7 +996,35 @@ struct xfrm_dst {
 	u32 path_cookie;
 };
 
+static inline struct dst_entry *xfrm_dst_path(const struct dst_entry *dst)
+{
 #ifdef CONFIG_XFRM
+	if (dst->xfrm) {
+		const struct xfrm_dst *xdst = (const struct xfrm_dst *) dst;
+
+		return xdst->path;
+	}
+#endif
+	return (struct dst_entry *) dst;
+}
+
+static inline struct dst_entry *xfrm_dst_child(const struct dst_entry *dst)
+{
+#ifdef CONFIG_XFRM
+	if (dst->xfrm) {
+		struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
+		return xdst->child;
+	}
+#endif
+	return NULL;
+}
+
+#ifdef CONFIG_XFRM
+static inline void xfrm_dst_set_child(struct xfrm_dst *xdst, struct dst_entry *child)
+{
+	xdst->child = child;
+}
+
 static inline void xfrm_dst_destroy(struct xfrm_dst *xdst)
 {
 	xfrm_pols_put(xdst->pols, xdst->num_pols);
@@ -1866,12 +1896,14 @@ bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x);
 static inline bool xfrm_dst_offload_ok(struct dst_entry *dst)
 {
 	struct xfrm_state *x = dst->xfrm;
+	struct xfrm_dst *xdst;
 
 	if (!x || !x->type_offload)
 		return false;
 
-	if (x->xso.offload_handle && (x->xso.dev == dst->path->dev) &&
-	    !dst->child->xfrm)
+	xdst = (struct xfrm_dst *) dst;
+	if (x->xso.offload_handle && (x->xso.dev == xfrm_dst_path(dst)->dev) &&
+	    !xdst->child->xfrm)
 		return true;
 
 	return false;

net/bridge/br_nf_core.c

@@ -78,7 +78,6 @@ void br_netfilter_rtable_init(struct net_bridge *br)
 
 	atomic_set(&rt->dst.__refcnt, 1);
 	rt->dst.dev = br->dev;
-	rt->dst.path = &rt->dst;
 	dst_init_metrics(&rt->dst, br_dst_default_metrics, true);
 	rt->dst.flags	= DST_NOXFRM | DST_FAKE_RTABLE;
 	rt->dst.ops = &fake_dst_ops;

net/core/dst.c

@@ -21,6 +21,7 @@
 #include <linux/sched.h>
 #include <linux/prefetch.h>
 #include <net/lwtunnel.h>
+#include <net/xfrm.h>
 
 #include <net/dst.h>
 #include <net/dst_metadata.h>
@@ -62,15 +63,12 @@ void dst_init(struct dst_entry *dst, struct dst_ops *ops,
 	      struct net_device *dev, int initial_ref, int initial_obsolete,
 	      unsigned short flags)
 {
-	dst->child = NULL;
 	dst->dev = dev;
 	if (dev)
 		dev_hold(dev);
 	dst->ops = ops;
 	dst_init_metrics(dst, dst_default_metrics.metrics, true);
 	dst->expires = 0UL;
-	dst->path = dst;
-	dst->from = NULL;
 #ifdef CONFIG_XFRM
 	dst->xfrm = NULL;
 #endif
@@ -88,7 +86,6 @@ void dst_init(struct dst_entry *dst, struct dst_ops *ops,
 	dst->__use = 0;
 	dst->lastuse = jiffies;
 	dst->flags = flags;
-	dst->next = NULL;
 	if (!(flags & DST_NOCOUNT))
 		dst_entries_add(ops, 1);
 }
@@ -116,12 +113,17 @@ EXPORT_SYMBOL(dst_alloc);
 
 struct dst_entry *dst_destroy(struct dst_entry * dst)
 {
-	struct dst_entry *child;
+	struct dst_entry *child = NULL;
 
 	smp_rmb();
 
-	child = dst->child;
+#ifdef CONFIG_XFRM
+	if (dst->xfrm) {
+		struct xfrm_dst *xdst = (struct xfrm_dst *) dst;
 
+		child = xdst->child;
+	}
+#endif
 	if (!(dst->flags & DST_NOCOUNT))
 		dst_entries_add(dst->ops, -1);
 

net/core/pktgen.c

@@ -399,7 +399,7 @@ struct pktgen_dev {
 	__u8	ipsmode;		/* IPSEC mode (config) */
 	__u8	ipsproto;		/* IPSEC type (config) */
 	__u32	spi;
-	struct dst_entry dst;
+	struct xfrm_dst xdst;
 	struct dst_ops dstops;
 #endif
 	char result[512];
@@ -2609,7 +2609,7 @@ static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
 	 * supports both transport/tunnel mode + ESP/AH type.
 	 */
 	if ((x->props.mode == XFRM_MODE_TUNNEL) && (pkt_dev->spi != 0))
-		skb->_skb_refdst = (unsigned long)&pkt_dev->dst | SKB_DST_NOREF;
+		skb->_skb_refdst = (unsigned long)&pkt_dev->xdst.u.dst | SKB_DST_NOREF;
 
 	rcu_read_lock_bh();
 	err = x->outer_mode->output(x, skb);
@@ -3742,10 +3742,10 @@ static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
 	 * performance under such circumstance.
 	 */
 	pkt_dev->dstops.family = AF_INET;
-	pkt_dev->dst.dev = pkt_dev->odev;
-	dst_init_metrics(&pkt_dev->dst, pktgen_dst_metrics, false);
-	pkt_dev->dst.child = &pkt_dev->dst;
-	pkt_dev->dst.ops = &pkt_dev->dstops;
+	pkt_dev->xdst.u.dst.dev = pkt_dev->odev;
+	dst_init_metrics(&pkt_dev->xdst.u.dst, pktgen_dst_metrics, false);
+	pkt_dev->xdst.child = &pkt_dev->xdst.u.dst;
+	pkt_dev->xdst.u.dst.ops = &pkt_dev->dstops;
 #endif
 
 	return add_dev_to_thread(t, pkt_dev);