Merge davem@nuts.ninka.net:/home/davem/src/BK/net-2.5

into kernel.bkbits.net:/home/davem/net-2.5

drivers/net/sungem.c

@@ -2850,6 +2850,7 @@ static int __devinit gem_get_device_address(struct gem *gp)
 		printk(KERN_ERR "%s: can't get mac-address\n", dev->name);
 		return -1;
 	}
+#warning MAX_ADDR_LEN is now 32 bytes instead of 8, please fix this as appropriate
 	memcpy(dev->dev_addr, addr, MAX_ADDR_LEN);
 #else
 	get_gem_mac_nonobp(gp->pdev, gp->dev->dev_addr);

include/linux/if_arp.h

@@ -40,6 +40,7 @@
 #define ARPHRD_METRICOM	23		/* Metricom STRIP (new IANA id)	*/
 #define	ARPHRD_IEEE1394	24		/* IEEE 1394 IPv4 - RFC 2734	*/
 #define ARPHRD_EUI64	27		/* EUI-64                       */
+#define ARPHRD_INFINIBAND 32		/* InfiniBand			*/
 
 /* Dummy types for non ARP hardware */
 #define ARPHRD_SLIP	256

include/linux/netdevice.h

@@ -66,7 +66,7 @@ struct vlan_group;
 
 #endif
 
-#define MAX_ADDR_LEN	8		/* Largest hardware address length */
+#define MAX_ADDR_LEN	32		/* Largest hardware address length */
 
 /*
  *	Compute the worst case header length according to the protocols

include/linux/rtnetlink.h

@@ -17,6 +17,7 @@
 #define	RTM_NEWLINK	(RTM_BASE+0)
 #define	RTM_DELLINK	(RTM_BASE+1)
 #define	RTM_GETLINK	(RTM_BASE+2)
+#define	RTM_SETLINK	(RTM_BASE+3)
 
 #define	RTM_NEWADDR	(RTM_BASE+4)
 #define	RTM_DELADDR	(RTM_BASE+5)

include/net/icmp.h

@@ -23,6 +23,7 @@
 
 #include <net/sock.h>
 #include <net/protocol.h>
+#include <net/snmp.h>
 #include <linux/ip.h>
 
 struct icmp_err {
@@ -31,10 +32,22 @@ struct icmp_err {
 };
 
 extern struct icmp_err icmp_err_convert[];
-extern struct icmp_mib icmp_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct icmp_mib, icmp_statistics);
 #define ICMP_INC_STATS(field)		SNMP_INC_STATS(icmp_statistics, field)
 #define ICMP_INC_STATS_BH(field)	SNMP_INC_STATS_BH(icmp_statistics, field)
 #define ICMP_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(icmp_statistics, field)
+#define ICMP_INC_STATS_FIELD(offt)					\
+	(*((unsigned long *) ((void *)					\
+			     per_cpu_ptr(icmp_statistics[!in_softirq()],\
+					 smp_processor_id())) + offt))++;
+#define ICMP_INC_STATS_BH_FIELD(offt)					\
+	(*((unsigned long *) ((void *)					\
+			     per_cpu_ptr(icmp_statistics[0],		\
+					 smp_processor_id())) + offt))++;
+#define ICMP_INC_STATS_USER_FIELD(offt)					\
+	(*((unsigned long *) ((void *)					\
+			     per_cpu_ptr(icmp_statistics[1],		\
+					 smp_processor_id())) + offt))++;
 
 extern void	icmp_send(struct sk_buff *skb_in,  int type, int code, u32 info);
 extern int	icmp_rcv(struct sk_buff *skb);

include/net/ip.h

@@ -149,14 +149,16 @@ struct ipv4_config
 };
 
 extern struct ipv4_config ipv4_config;
-extern struct ip_mib	ip_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct ip_mib, ip_statistics);
 #define IP_INC_STATS(field)		SNMP_INC_STATS(ip_statistics, field)
 #define IP_INC_STATS_BH(field)		SNMP_INC_STATS_BH(ip_statistics, field)
 #define IP_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(ip_statistics, field)
-extern struct linux_mib	net_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct linux_mib, net_statistics);
 #define NET_INC_STATS(field)		SNMP_INC_STATS(net_statistics, field)
 #define NET_INC_STATS_BH(field)		SNMP_INC_STATS_BH(net_statistics, field)
 #define NET_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(net_statistics, field)
+#define NET_ADD_STATS_BH(field, adnd)	SNMP_ADD_STATS_BH(net_statistics, field, adnd)
+#define NET_ADD_STATS_USER(field, adnd)	SNMP_ADD_STATS_USER(net_statistics, field, adnd)
 
 extern int sysctl_local_port_range[2];
 extern int sysctl_ip_default_ttl;

include/net/ipv6.h

@@ -19,6 +19,7 @@
 #include <asm/hardirq.h>
 #include <net/ndisc.h>
 #include <net/flow.h>
+#include <net/snmp.h>
 
 #define SIN6_LEN_RFC2133	24
 
@@ -105,15 +106,19 @@ struct frag_hdr {
 /* sysctls */
 extern int sysctl_ipv6_bindv6only;
 
-extern struct ipv6_mib		ipv6_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct ipv6_mib, ipv6_statistics);
 #define IP6_INC_STATS(field)		SNMP_INC_STATS(ipv6_statistics, field)
 #define IP6_INC_STATS_BH(field)		SNMP_INC_STATS_BH(ipv6_statistics, field)
 #define IP6_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(ipv6_statistics, field)
-extern struct icmpv6_mib	icmpv6_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics);
 #define ICMP6_INC_STATS(field)		SNMP_INC_STATS(icmpv6_statistics, field)
 #define ICMP6_INC_STATS_BH(field)	SNMP_INC_STATS_BH(icmpv6_statistics, field)
 #define ICMP6_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(icmpv6_statistics, field)
-extern struct udp_mib		udp_stats_in6[NR_CPUS*2];
+#define ICMP6_STATS_PTR_BH(field) 					\
+	(&								\
+	 ((per_cpu_ptr(icmpv6_statistics[0], smp_processor_id()))->	\
+	  field))
+DECLARE_SNMP_STAT(struct udp_mib, udp_stats_in6);
 #define UDP6_INC_STATS(field)		SNMP_INC_STATS(udp_stats_in6, field)
 #define UDP6_INC_STATS_BH(field)	SNMP_INC_STATS_BH(udp_stats_in6, field)
 #define UDP6_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(udp_stats_in6, field)

include/net/route.h

@@ -102,7 +102,11 @@ struct rt_cache_stat
         unsigned int gc_ignored;
         unsigned int gc_goal_miss;
         unsigned int gc_dst_overflow;
-} ____cacheline_aligned_in_smp;
+};
+
+extern struct rt_cache_stat *rt_cache_stat;
+#define RT_CACHE_STAT_INC(field)					  \
+		(per_cpu_ptr(rt_cache_stat, smp_processor_id())->field++)
 
 extern struct ip_rt_acct *ip_rt_acct;
 

include/net/sctp/sctp.h

@@ -203,7 +203,7 @@ extern void sctp_hash_digest(const char *secret, const int secret_len,
 #define SCTP_SOCK_SLEEP_POST(sk) SOCK_SLEEP_POST(sk)
 
 /* SCTP SNMP MIB stats handlers */
-extern struct sctp_mib sctp_statistics[NR_CPUS * 2];
+DECLARE_SNMP_STAT(struct sctp_mib, sctp_statistics);
 #define SCTP_INC_STATS(field)		SNMP_INC_STATS(sctp_statistics, field)
 #define SCTP_INC_STATS_BH(field)	SNMP_INC_STATS_BH(sctp_statistics, field)
 #define SCTP_INC_STATS_USER(field)	SNMP_INC_STATS_USER(sctp_statistics, field)

include/net/snmp.h

@@ -62,7 +62,7 @@ struct ip_mib
  	unsigned long	IpFragFails;
  	unsigned long	IpFragCreates;
 	unsigned long   __pad[0]; 
-} ____cacheline_aligned;
+};
  
 struct ipv6_mib
 {
@@ -89,7 +89,7 @@ struct ipv6_mib
  	unsigned long	Ip6InMcastPkts;
  	unsigned long	Ip6OutMcastPkts;
 	unsigned long   __pad[0]; 
-} ____cacheline_aligned;
+};
  
 struct icmp_mib
 {
@@ -121,7 +121,7 @@ struct icmp_mib
  	unsigned long	IcmpOutAddrMaskReps;
 	unsigned long	dummy;
 	unsigned long   __pad[0]; 
-} ____cacheline_aligned;
+};
 
 struct icmpv6_mib
 {
@@ -159,7 +159,7 @@ struct icmpv6_mib
 	unsigned long	Icmp6OutGroupMembResponses;
 	unsigned long	Icmp6OutGroupMembReductions;
 	unsigned long   __pad[0]; 
-} ____cacheline_aligned;
+};
  
 struct tcp_mib
 {
@@ -178,7 +178,7 @@ struct tcp_mib
  	unsigned long	TcpInErrs;
  	unsigned long	TcpOutRsts;
 	unsigned long   __pad[0]; 
-} ____cacheline_aligned;
+};
  
 struct udp_mib
 {
@@ -187,7 +187,7 @@ struct udp_mib
  	unsigned long	UdpInErrors;
  	unsigned long	UdpOutDatagrams;
 	unsigned long   __pad[0];
-} ____cacheline_aligned; 
+}; 
 
 /* draft-ietf-sigtran-sctp-mib-07.txt */
 struct sctp_mib
@@ -216,7 +216,7 @@ struct sctp_mib
 	unsigned long   SctpValCookieLife;
 	unsigned long   SctpMaxInitRetr;
 	unsigned long   __pad[0];
-} ____cacheline_aligned;
+};
 
 struct linux_mib 
 {
@@ -286,7 +286,7 @@ struct linux_mib
 	unsigned long	TCPAbortFailed;
 	unsigned long	TCPMemoryPressures;
 	unsigned long   __pad[0];
-} ____cacheline_aligned;
+};
 
 
 /* 
@@ -294,8 +294,25 @@ struct linux_mib
  * addl $1,memory is atomic against interrupts (but atomic_inc would be overkill because of the lock 
  * cycles). Wants new nonlocked_atomic_inc() primitives -AK
  */ 
-#define SNMP_INC_STATS(mib, field) ((mib)[2*smp_processor_id()+!in_softirq()].field++)
-#define SNMP_INC_STATS_BH(mib, field) ((mib)[2*smp_processor_id()].field++)
-#define SNMP_INC_STATS_USER(mib, field) ((mib)[2*smp_processor_id()+1].field++)
+#define DEFINE_SNMP_STAT(type, name)	\
+	__typeof__(type) *name[2]
+#define DECLARE_SNMP_STAT(type, name)	\
+	extern __typeof__(type) *name[2]
+
+#define SNMP_STAT_USRPTR(name)	(name[0])
+#define SNMP_STAT_BHPTR(name)	(name[1])
+
+#define SNMP_INC_STATS_BH(mib, field) 	\
+	(per_cpu_ptr(mib[0], smp_processor_id())->field++)
+#define SNMP_INC_STATS_USER(mib, field) \
+	(per_cpu_ptr(mib[1], smp_processor_id())->field++)
+#define SNMP_INC_STATS(mib, field) 	\
+	(per_cpu_ptr(mib[!in_softirq()], smp_processor_id())->field++)
+#define SNMP_DEC_STATS(mib, field) 	\
+	(per_cpu_ptr(mib[!in_softirq()], smp_processor_id())->field--)
+#define SNMP_ADD_STATS_BH(mib, field, addend) 	\
+	(per_cpu_ptr(mib[0], smp_processor_id())->field += addend)
+#define SNMP_ADD_STATS_USER(mib, field, addend) 	\
+	(per_cpu_ptr(mib[1], smp_processor_id())->field += addend)
  	
 #endif

include/net/tcp.h

@@ -28,6 +28,7 @@
 #include <linux/tcp.h>
 #include <linux/slab.h>
 #include <linux/cache.h>
+#include <linux/percpu.h>
 #include <net/checksum.h>
 #include <net/sock.h>
 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
@@ -630,10 +631,11 @@ extern __inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
 
 extern struct proto tcp_prot;
 
-extern struct tcp_mib tcp_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
 #define TCP_INC_STATS(field)		SNMP_INC_STATS(tcp_statistics, field)
 #define TCP_INC_STATS_BH(field)		SNMP_INC_STATS_BH(tcp_statistics, field)
 #define TCP_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(tcp_statistics, field)
+#define TCP_DEC_STATS(field)		SNMP_DEC_STATS(tcp_statistics, field)
 
 extern void			tcp_put_port(struct sock *sk);
 extern void			__tcp_put_port(struct sock *sk);
@@ -1399,7 +1401,7 @@ static __inline__ void tcp_set_state(struct sock *sk, int state)
 		/* fall through */
 	default:
 		if (oldstate==TCP_ESTABLISHED)
-			tcp_statistics[smp_processor_id()*2+!in_softirq()].TcpCurrEstab--;
+			TCP_DEC_STATS(TcpCurrEstab);
 	}
 
 	/* Change state AFTER socket is unhashed to avoid closed

include/net/udp.h

@@ -25,6 +25,7 @@
 #include <linux/udp.h>
 #include <linux/ip.h>
 #include <net/sock.h>
+#include <net/snmp.h>
 
 #define UDP_HTABLE_SIZE		128
 
@@ -71,7 +72,7 @@ extern int	udp_rcv(struct sk_buff *skb);
 extern int	udp_ioctl(struct sock *sk, int cmd, unsigned long arg);
 extern int	udp_disconnect(struct sock *sk, int flags);
 
-extern struct udp_mib udp_statistics[NR_CPUS*2];
+DECLARE_SNMP_STAT(struct udp_mib, udp_statistics);
 #define UDP_INC_STATS(field)		SNMP_INC_STATS(udp_statistics, field)
 #define UDP_INC_STATS_BH(field)		SNMP_INC_STATS_BH(udp_statistics, field)
 #define UDP_INC_STATS_USER(field) 	SNMP_INC_STATS_USER(udp_statistics, field)

net/core/dev.c

@@ -2133,7 +2133,7 @@ static int dev_ifsioc(struct ifreq *ifr, unsigned int cmd)
 
 		case SIOCGIFHWADDR:
 			memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr,
-			       MAX_ADDR_LEN);
+			       min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len));
 			ifr->ifr_hwaddr.sa_family = dev->type;
 			return 0;
 
@@ -2154,7 +2154,7 @@ static int dev_ifsioc(struct ifreq *ifr, unsigned int cmd)
 			if (ifr->ifr_hwaddr.sa_family != dev->type)
 				return -EINVAL;
 			memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data,
-			       MAX_ADDR_LEN);
+			       min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len));
 			notifier_call_chain(&netdev_chain,
 					    NETDEV_CHANGEADDR, dev);
 			return 0;

net/core/rtnetlink.c

@@ -220,6 +220,40 @@ int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 	return skb->len;
 }
 
+static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
+{
+	struct ifinfomsg  *ifm = NLMSG_DATA(nlh);
+	struct rtattr    **ida = arg;
+	struct net_device *dev;
+	int err;
+
+	dev = dev_get_by_index(ifm->ifi_index);
+	if (!dev)
+		return -ENODEV;
+
+	err = -EINVAL;
+
+	if (ida[IFLA_ADDRESS - 1]) {
+		if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
+			goto out;
+		memcpy(dev->dev_addr, RTA_DATA(ida[IFLA_ADDRESS - 1]),
+		       dev->addr_len);
+	}
+
+	if (ida[IFLA_BROADCAST - 1]) {
+		if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
+			goto out;
+		memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
+		       dev->addr_len);
+	}
+
+	err = 0;
+
+out:
+	dev_put(dev);
+	return err;
+}
+
 int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
 {
 	int idx;
@@ -457,33 +491,15 @@ static void rtnetlink_rcv(struct sock *sk, int len)
 
 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
 {
-	{ NULL,			NULL,			},
-	{ NULL,			NULL,			},
-	{ NULL,			rtnetlink_dump_ifinfo,	},
-	{ NULL,			NULL,			},
-
-	{ NULL,			NULL,			},
-	{ NULL,			NULL,			},
-	{ NULL,			rtnetlink_dump_all,	},
-	{ NULL,			NULL,			},
-
-	{ NULL,			NULL,			},
-	{ NULL,			NULL,			},
-	{ NULL,			rtnetlink_dump_all,	},
-	{ NULL,			NULL,			},
-
-	{ neigh_add,		NULL,			},
-	{ neigh_delete,		NULL,			},
-	{ NULL,			neigh_dump_info,	},
-	{ NULL,			NULL,			},
-
-	{ NULL,			NULL,			},
-	{ NULL,			NULL,			},
-	{ NULL,			NULL,			},
-	{ NULL,			NULL,			},
+	[RTM_GETLINK  - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
+	[RTM_SETLINK  - RTM_BASE] = { .doit   = do_setlink	      },
+	[RTM_GETADDR  - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
+	[RTM_GETROUTE - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
+	[RTM_NEWNEIGH - RTM_BASE] = { .doit   = neigh_add	      },
+	[RTM_DELNEIGH - RTM_BASE] = { .doit   = neigh_delete	      },
+	[RTM_GETNEIGH - RTM_BASE] = { .dumpit = neigh_dump_info	      }
 };
 
-
 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
 {
 	struct net_device *dev = ptr;

net/core/skbuff.c

@@ -324,7 +324,9 @@ void __kfree_skb(struct sk_buff *skb)
 	}
 
 	dst_release(skb->dst);
+#ifdef CONFIG_INET
 	secpath_put(skb->sp);
+#endif
 	if(skb->destructor) {
 		if (in_irq())
 			printk(KERN_WARNING "Warning: kfree_skb on "
@@ -378,7 +380,9 @@ struct sk_buff *skb_clone(struct sk_buff *skb, int gfp_mask)
 	C(dst);
 	dst_clone(n->dst);
 	C(sp);
+#ifdef CONFIG_INET
 	secpath_get(n->sp);
+#endif
 	memcpy(n->cb, skb->cb, sizeof(skb->cb));
 	C(len);
 	C(data_len);
@@ -438,7 +442,9 @@ static void copy_skb_header(struct sk_buff *new, const struct sk_buff *old)
 	new->priority	= old->priority;
 	new->protocol	= old->protocol;
 	new->dst	= dst_clone(old->dst);
+#ifdef CONFIG_INET
 	new->sp		= secpath_get(old->sp);
+#endif
 	new->h.raw	= old->h.raw + offset;
 	new->nh.raw	= old->nh.raw + offset;
 	new->mac.raw	= old->mac.raw + offset;

net/ipv4/af_inet.c

@@ -113,7 +113,7 @@
 #include <linux/mroute.h>
 #endif
 
-struct linux_mib net_statistics[NR_CPUS * 2];
+DEFINE_SNMP_STAT(struct linux_mib, net_statistics);
 
 #ifdef INET_REFCNT_DEBUG
 atomic_t inet_sock_nr;
@@ -1053,6 +1053,59 @@ static struct inet_protocol icmp_protocol = {
 	.handler =	icmp_rcv,
 };
 
+static int __init init_ipv4_mibs(void)
+{
+	int i;
+
+	net_statistics[0] =
+	    kmalloc_percpu(sizeof (struct linux_mib), GFP_KERNEL);
+	net_statistics[1] =
+	    kmalloc_percpu(sizeof (struct linux_mib), GFP_KERNEL);
+	ip_statistics[0] = kmalloc_percpu(sizeof (struct ip_mib), GFP_KERNEL);
+	ip_statistics[1] = kmalloc_percpu(sizeof (struct ip_mib), GFP_KERNEL);
+	icmp_statistics[0] =
+	    kmalloc_percpu(sizeof (struct icmp_mib), GFP_KERNEL);
+	icmp_statistics[1] =
+	    kmalloc_percpu(sizeof (struct icmp_mib), GFP_KERNEL);
+	tcp_statistics[0] = kmalloc_percpu(sizeof (struct tcp_mib), GFP_KERNEL);
+	tcp_statistics[1] = kmalloc_percpu(sizeof (struct tcp_mib), GFP_KERNEL);
+	udp_statistics[0] = kmalloc_percpu(sizeof (struct udp_mib), GFP_KERNEL);
+	udp_statistics[1] = kmalloc_percpu(sizeof (struct udp_mib), GFP_KERNEL);
+	if (!
+	    (net_statistics[0] && net_statistics[1] && ip_statistics[0]
+	     && ip_statistics[1] && tcp_statistics[0] && tcp_statistics[1]
+	     && udp_statistics[0] && udp_statistics[1]))
+		return -ENOMEM;
+
+	/* Set all the per cpu copies of the mibs to zero */
+	for (i = 0; i < NR_CPUS; i++) {
+		if (cpu_possible(i)) {
+			memset(per_cpu_ptr(net_statistics[0], i), 0,
+			       sizeof (struct linux_mib));
+			memset(per_cpu_ptr(net_statistics[1], i), 0,
+			       sizeof (struct linux_mib));
+			memset(per_cpu_ptr(ip_statistics[0], i), 0,
+			       sizeof (struct ip_mib));
+			memset(per_cpu_ptr(ip_statistics[1], i), 0,
+			       sizeof (struct ip_mib));
+			memset(per_cpu_ptr(icmp_statistics[0], i), 0,
+			       sizeof (struct icmp_mib));
+			memset(per_cpu_ptr(icmp_statistics[1], i), 0,
+			       sizeof (struct icmp_mib));
+			memset(per_cpu_ptr(tcp_statistics[0], i), 0,
+			       sizeof (struct tcp_mib));
+			memset(per_cpu_ptr(tcp_statistics[1], i), 0,
+			       sizeof (struct tcp_mib));
+			memset(per_cpu_ptr(udp_statistics[0], i), 0,
+			       sizeof (struct udp_mib));
+			memset(per_cpu_ptr(udp_statistics[1], i), 0,
+			       sizeof (struct udp_mib));
+		}
+	}
+
+	return 0;
+}
+
 int ipv4_proc_init(void);
 
 static int __init inet_init(void)
@@ -1148,6 +1201,12 @@ static int __init inet_init(void)
 #if defined(CONFIG_IP_MROUTE)
 	ip_mr_init();
 #endif
+	/*
+	 *	Initialise per-cpu ipv4 mibs
+	 */ 
+
+	if(init_ipv4_mibs())
+		printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
 	
 	ipv4_proc_init();
 	return 0;

net/ipv4/icmp.c

@@ -113,7 +113,7 @@ struct icmp_bxm {
 /*
  *	Statistics
  */
-struct icmp_mib icmp_statistics[NR_CPUS * 2];
+DEFINE_SNMP_STAT(struct icmp_mib, icmp_statistics);
 
 /* An array of errno for error messages from dest unreach. */
 /* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOS_UNREACH and SR_FAIELD MUST be considered 'transient errs'. */
@@ -213,10 +213,9 @@ int sysctl_icmp_ratemask = 0x1818;
  *	ICMP control array. This specifies what to do with each ICMP.
  */
 
-struct icmp_control
-{
-	unsigned long *output;		/* Address to increment on output */
-	unsigned long *input;		/* Address to increment on input */
+struct icmp_control {
+	int output_off;		/* Field offset for increment on output */
+	int input_off;		/* Field offset for increment on input */
 	void (*handler)(struct sk_buff *skb);
 	short   error;		/* This ICMP is classed as an error message */
 };
@@ -343,10 +342,7 @@ static inline int icmpv4_xrlim_allow(struct rtable *rt, int type, int code)
 static void icmp_out_count(int type)
 {
 	if (type <= NR_ICMP_TYPES) {
-		(icmp_pointers[type].output)[(smp_processor_id() * 2 +
-					      !in_softirq()) *
-					     sizeof(struct icmp_mib) /
-					     sizeof(unsigned long)]++;
+		ICMP_INC_STATS_FIELD(icmp_pointers[type].output_off);
 		ICMP_INC_STATS(IcmpOutMsgs);
 	}
 }
@@ -1005,9 +1001,7 @@ int icmp_rcv(struct sk_buff *skb)
   		}
 	}
 
-	icmp_pointers[icmph->type].input[smp_processor_id() * 2 *
-					 sizeof(struct icmp_mib) /
-					 sizeof(unsigned long)]++;
+	ICMP_INC_STATS_BH_FIELD(icmp_pointers[icmph->type].input_off);
 	(icmp_pointers[icmph->type].handler)(skb);
 
 drop:
@@ -1024,122 +1018,122 @@ int icmp_rcv(struct sk_buff *skb)
 static struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
 	/* ECHO REPLY (0) */
 	[0] = {
-		.output =	 &icmp_statistics[0].IcmpOutEchoReps,
-		.input = &icmp_statistics[0].IcmpInEchoReps,
+		.output_off = offsetof(struct icmp_mib, IcmpOutEchoReps),
+		.input_off = offsetof(struct icmp_mib, IcmpInEchoReps),
 		.handler = icmp_discard,
 	},
 	[1] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].IcmpInErrors,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib,IcmpInErrors),
 		.handler = icmp_discard,
 		.error = 1,
 	},
 	[2] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].IcmpInErrors,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib,IcmpInErrors),
 		.handler = icmp_discard,
 		.error = 1,
 	},
 	/* DEST UNREACH (3) */
 	[3] = {
-		.output =	 &icmp_statistics[0].IcmpOutDestUnreachs,
-		.input = &icmp_statistics[0].IcmpInDestUnreachs,
+		.output_off = offsetof(struct icmp_mib, IcmpOutDestUnreachs),
+		.input_off = offsetof(struct icmp_mib, IcmpInDestUnreachs),
 		.handler = icmp_unreach,
 		.error = 1,
 	},
 	/* SOURCE QUENCH (4) */
 	[4] = {
-		.output =	 &icmp_statistics[0].IcmpOutSrcQuenchs,
-		.input = &icmp_statistics[0].IcmpInSrcQuenchs,
+		.output_off = offsetof(struct icmp_mib, IcmpOutSrcQuenchs),
+		.input_off = offsetof(struct icmp_mib, IcmpInSrcQuenchs),
 		icmp_unreach,
 		.error = 1,
 	},
 	/* REDIRECT (5) */
 	[5] = {
-		.output =	 &icmp_statistics[0].IcmpOutRedirects,
-		.input = &icmp_statistics[0].IcmpInRedirects,
+		.output_off = offsetof(struct icmp_mib, IcmpOutRedirects),
+		.input_off = offsetof(struct icmp_mib, IcmpInRedirects),
 		.handler = icmp_redirect,
 		.error = 1,
 	},
 	[6] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].IcmpInErrors,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib, IcmpInErrors),
 		.handler = icmp_discard,
 		.error = 1,
 	},
 	[7] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].IcmpInErrors,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib, IcmpInErrors),
 		.handler = icmp_discard,
 		.error = 1,
 	},
 	/* ECHO (8) */
 	[8] = {
-		.output =	 &icmp_statistics[0].IcmpOutEchos,
-		.input = &icmp_statistics[0].IcmpInEchos,
+		.output_off = offsetof(struct icmp_mib, IcmpOutEchos),
+		.input_off = offsetof(struct icmp_mib, IcmpInEchos),
 		.handler = icmp_echo,
 		.error = 0,
 	},
 	[9] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].IcmpInErrors,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib, IcmpInErrors),
 		.handler = icmp_discard,
 		.error = 1,
 	},
 	[10] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].IcmpInErrors,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib, IcmpInErrors),
 		.handler = icmp_discard,
 		.error = 1,
 	},
 	/* TIME EXCEEDED (11) */
 	[11] = {
-		.output =	 &icmp_statistics[0].IcmpOutTimeExcds,
-		.input = &icmp_statistics[0].IcmpInTimeExcds,
+		.output_off = offsetof(struct icmp_mib, IcmpOutTimeExcds),
+		.input_off = offsetof(struct icmp_mib,IcmpInTimeExcds),
 		.handler = icmp_unreach,
 		.error = 1,
 	},
 	/* PARAMETER PROBLEM (12) */
 	[12] = {
-		.output =	 &icmp_statistics[0].IcmpOutParmProbs,
-		.input = &icmp_statistics[0].IcmpInParmProbs,
+		.output_off = offsetof(struct icmp_mib, IcmpOutParmProbs),
+		.input_off = offsetof(struct icmp_mib, IcmpInParmProbs),
 		.handler = icmp_unreach,
 		.error = 1,
 	},
 	/* TIMESTAMP (13) */
 	[13] = {
-		.output =	 &icmp_statistics[0].IcmpOutTimestamps,
-		.input = &icmp_statistics[0].IcmpInTimestamps,
+		.output_off = offsetof(struct icmp_mib, IcmpOutTimestamps),
+		.input_off = offsetof(struct icmp_mib, IcmpInTimestamps),
 		.handler = icmp_timestamp,
 	},
 	/* TIMESTAMP REPLY (14) */
 	[14] = {
-		.output =	 &icmp_statistics[0].IcmpOutTimestampReps,
-		.input = &icmp_statistics[0].IcmpInTimestampReps,
+		.output_off = offsetof(struct icmp_mib, IcmpOutTimestampReps),
+		.input_off = offsetof(struct icmp_mib, IcmpInTimestampReps),
 		.handler = icmp_discard,
 	},
 	/* INFO (15) */
 	[15] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].dummy,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib, dummy),
 		.handler = icmp_discard,
 	},
 	/* INFO REPLY (16) */
  	[16] = {
-		.output =	 &icmp_statistics[0].dummy,
-		.input = &icmp_statistics[0].dummy,
+		.output_off = offsetof(struct icmp_mib, dummy),
+		.input_off = offsetof(struct icmp_mib, dummy),
 		.handler = icmp_discard,
 	},
 	/* ADDR MASK (17) */
 	[17] = {
-		.output =	 &icmp_statistics[0].IcmpOutAddrMasks,
-		.input = &icmp_statistics[0].IcmpInAddrMasks,
+		.output_off = offsetof(struct icmp_mib, IcmpOutAddrMasks),
+		.input_off = offsetof(struct icmp_mib, IcmpInAddrMasks),
 		.handler = icmp_address,
 	},
 	/* ADDR MASK REPLY (18) */
 	[18] = {
-		.output =	 &icmp_statistics[0].IcmpOutAddrMaskReps,
-		.input = &icmp_statistics[0].IcmpInAddrMaskReps,
+		.output_off = offsetof(struct icmp_mib, IcmpOutAddrMaskReps),
+		.input_off = offsetof(struct icmp_mib, IcmpInAddrMaskReps),
 		.handler = icmp_address_reply,
 	}
 };

net/ipv4/ip_input.c

@@ -149,7 +149,7 @@
  *	SNMP management statistics
  */
 
-struct ip_mib ip_statistics[NR_CPUS*2];
+DEFINE_SNMP_STAT(struct ip_mib, ip_statistics);
 
 /*
  *	Process Router Attention IP option

net/ipv4/proc.c

@@ -86,16 +86,21 @@ static struct file_operations sockstat_seq_fops = {
 	.release = single_release,
 };
 
-static unsigned long fold_field(unsigned long *begin, int sz, int nr)
+static unsigned long
+fold_field(void *mib[], int nr)
 {
 	unsigned long res = 0;
 	int i;
 
-	sz /= sizeof(unsigned long);
-
 	for (i = 0; i < NR_CPUS; i++) {
-		res += begin[2 * i * sz + nr];
-		res += begin[(2 * i + 1) * sz + nr];
+		if (!cpu_possible(i))
+			continue;
+		res +=
+		    *((unsigned long *) (((void *) per_cpu_ptr(mib[0], i)) +
+					 sizeof (unsigned long) * nr));
+		res +=
+		    *((unsigned long *) (((void *) per_cpu_ptr(mib[0], i)) +
+					 sizeof (unsigned long) * nr));
 	}
 	return res;
 }
@@ -118,8 +123,7 @@ static int snmp_seq_show(struct seq_file *seq, void *v)
 	for (i = 0;
 	     i < offsetof(struct ip_mib, __pad) / sizeof(unsigned long); i++)
 		seq_printf(seq, " %lu",
-			   fold_field((unsigned long *)ip_statistics,
-				      sizeof(struct ip_mib), i));
+			   fold_field((void **) ip_statistics, i));
 
 	seq_printf(seq, "\nIcmp: InMsgs InErrors InDestUnreachs InTimeExcds "
 			"InParmProbs InSrcQuenchs InRedirects InEchos "
@@ -132,8 +136,7 @@ static int snmp_seq_show(struct seq_file *seq, void *v)
 	for (i = 0;
 	     i < offsetof(struct icmp_mib, dummy) / sizeof(unsigned long); i++)
 		seq_printf(seq, " %lu",
-			   fold_field((unsigned long *)icmp_statistics,
-				      sizeof(struct icmp_mib), i));
+			   fold_field((void **) icmp_statistics, i)); 
 
 	seq_printf(seq, "\nTcp: RtoAlgorithm RtoMin RtoMax MaxConn ActiveOpens "
 			"PassiveOpens AttemptFails EstabResets CurrEstab "
@@ -142,8 +145,7 @@ static int snmp_seq_show(struct seq_file *seq, void *v)
 	for (i = 0;
 	     i < offsetof(struct tcp_mib, __pad) / sizeof(unsigned long); i++)
 		seq_printf(seq, " %lu",
-			   fold_field((unsigned long *)tcp_statistics,
-				      sizeof(struct tcp_mib), i));
+			   fold_field((void **) tcp_statistics, i));
 
 	seq_printf(seq, "\nUdp: InDatagrams NoPorts InErrors OutDatagrams\n"
 			"Udp:");
@@ -151,8 +153,7 @@ static int snmp_seq_show(struct seq_file *seq, void *v)
 	for (i = 0;
 	     i < offsetof(struct udp_mib, __pad) / sizeof(unsigned long); i++)
 		seq_printf(seq, " %lu", 
-			   fold_field((unsigned long *)udp_statistics,
-				      sizeof(struct udp_mib), i));
+				fold_field((void **) udp_statistics, i));
 
 	seq_putc(seq, '\n');
 	return 0;
@@ -206,10 +207,10 @@ static int netstat_seq_show(struct seq_file *seq, void *v)
 		      " TCPAbortFailed TCPMemoryPressures\n"
 		      "TcpExt:");
 	for (i = 0;
-	     i < offsetof(struct linux_mib, __pad) / sizeof(unsigned long); i++)
+	     i < offsetof(struct linux_mib, __pad) / sizeof(unsigned long); 
+	     i++)
 		seq_printf(seq, " %lu", 
-			   fold_field((unsigned long *)net_statistics,
-				      sizeof(struct linux_mib), i));
+		 	   fold_field((void **) net_statistics, i)); 
 	seq_putc(seq, '\n');
 	return 0;
 }

net/ipv4/route.c

@@ -196,7 +196,7 @@ static struct rt_hash_bucket 	*rt_hash_table;
 static unsigned			rt_hash_mask;
 static int			rt_hash_log;
 
-struct rt_cache_stat rt_cache_stat[NR_CPUS];
+struct rt_cache_stat *rt_cache_stat;
 
 static int rt_intern_hash(unsigned hash, struct rtable *rth,
 				struct rtable **res);
@@ -318,24 +318,26 @@ static int rt_cache_stat_get_info(char *buffer, char **start, off_t offset, int
 	int len = 0;
 
 	for (i = 0; i < NR_CPUS; i++) {
+		if (!cpu_possible(i))
+			continue;
 		len += sprintf(buffer+len, "%08x  %08x %08x %08x %08x %08x %08x %08x  %08x %08x %08x %08x %08x %08x %08x \n",
 			       dst_entries,		       
-			       rt_cache_stat[i].in_hit,
-			       rt_cache_stat[i].in_slow_tot,
-			       rt_cache_stat[i].in_slow_mc,
-			       rt_cache_stat[i].in_no_route,
-			       rt_cache_stat[i].in_brd,
-			       rt_cache_stat[i].in_martian_dst,
-			       rt_cache_stat[i].in_martian_src,
-
-			       rt_cache_stat[i].out_hit,
-			       rt_cache_stat[i].out_slow_tot,
-			       rt_cache_stat[i].out_slow_mc, 
-
-			       rt_cache_stat[i].gc_total,
-			       rt_cache_stat[i].gc_ignored,
-			       rt_cache_stat[i].gc_goal_miss,
-			       rt_cache_stat[i].gc_dst_overflow
+			       per_cpu_ptr(rt_cache_stat, i)->in_hit,
+			       per_cpu_ptr(rt_cache_stat, i)->in_slow_tot,
+			       per_cpu_ptr(rt_cache_stat, i)->in_slow_mc,
+			       per_cpu_ptr(rt_cache_stat, i)->in_no_route,
+			       per_cpu_ptr(rt_cache_stat, i)->in_brd,
+			       per_cpu_ptr(rt_cache_stat, i)->in_martian_dst,
+			       per_cpu_ptr(rt_cache_stat, i)->in_martian_src,
+
+			       per_cpu_ptr(rt_cache_stat, i)->out_hit,
+			       per_cpu_ptr(rt_cache_stat, i)->out_slow_tot,
+			       per_cpu_ptr(rt_cache_stat, i)->out_slow_mc, 
+
+			       per_cpu_ptr(rt_cache_stat, i)->gc_total,
+			       per_cpu_ptr(rt_cache_stat, i)->gc_ignored,
+			       per_cpu_ptr(rt_cache_stat, i)->gc_goal_miss,
+			       per_cpu_ptr(rt_cache_stat, i)->gc_dst_overflow
 
 			);
 	}
@@ -591,11 +593,11 @@ static int rt_garbage_collect(void)
 	 * do not make it too frequently.
 	 */
 
-	rt_cache_stat[smp_processor_id()].gc_total++;
+	RT_CACHE_STAT_INC(gc_total);
 
 	if (now - last_gc < ip_rt_gc_min_interval &&
 	    atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) {
-		rt_cache_stat[smp_processor_id()].gc_ignored++;
+		RT_CACHE_STAT_INC(gc_ignored);
 		goto out;
 	}
 
@@ -663,7 +665,7 @@ static int rt_garbage_collect(void)
 		     We will not spin here for long time in any case.
 		 */
 
-		rt_cache_stat[smp_processor_id()].gc_goal_miss++;
+		RT_CACHE_STAT_INC(gc_goal_miss);
 
 		if (expire == 0)
 			break;
@@ -682,7 +684,7 @@ static int rt_garbage_collect(void)
 		goto out;
 	if (net_ratelimit())
 		printk(KERN_WARNING "dst cache overflow\n");
-	rt_cache_stat[smp_processor_id()].gc_dst_overflow++;
+	RT_CACHE_STAT_INC(gc_dst_overflow);
 	return 1;
 
 work_done:
@@ -1400,7 +1402,7 @@ static int ip_route_input_mc(struct sk_buff *skb, u32 daddr, u32 saddr,
 	if (!LOCAL_MCAST(daddr) && IN_DEV_MFORWARD(in_dev))
 		rth->u.dst.input = ip_mr_input;
 #endif
-	rt_cache_stat[smp_processor_id()].in_slow_mc++;
+	RT_CACHE_STAT_INC(in_slow_mc);
 
 	in_dev_put(in_dev);
 	hash = rt_hash_code(daddr, saddr ^ (dev->ifindex << 5), tos);
@@ -1485,7 +1487,7 @@ int ip_route_input_slow(struct sk_buff *skb, u32 daddr, u32 saddr,
 	}
 	free_res = 1;
 
-	rt_cache_stat[smp_processor_id()].in_slow_tot++;
+	RT_CACHE_STAT_INC(in_slow_tot);
 
 #ifdef CONFIG_IP_ROUTE_NAT
 	/* Policy is applied before mapping destination,
@@ -1642,7 +1644,7 @@ out:	return err;
 	}
 	flags |= RTCF_BROADCAST;
 	res.type = RTN_BROADCAST;
-	rt_cache_stat[smp_processor_id()].in_brd++;
+	RT_CACHE_STAT_INC(in_brd);
 
 local_input:
 	rth = dst_alloc(&ipv4_dst_ops);
@@ -1687,7 +1689,7 @@ out:	return err;
 	goto intern;
 
 no_route:
-	rt_cache_stat[smp_processor_id()].in_no_route++;
+	RT_CACHE_STAT_INC(in_no_route);
 	spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
 	res.type = RTN_UNREACHABLE;
 	goto local_input;
@@ -1696,7 +1698,7 @@ out:	return err;
 	 *	Do not cache martian addresses: they should be logged (RFC1812)
 	 */
 martian_destination:
-	rt_cache_stat[smp_processor_id()].in_martian_dst++;
+	RT_CACHE_STAT_INC(in_martian_dst);
 #ifdef CONFIG_IP_ROUTE_VERBOSE
 	if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit())
 		printk(KERN_WARNING "martian destination %u.%u.%u.%u from "
@@ -1713,7 +1715,7 @@ out:	return err;
 
 martian_source:
 
-	rt_cache_stat[smp_processor_id()].in_martian_src++;
+	RT_CACHE_STAT_INC(in_martian_src);
 #ifdef CONFIG_IP_ROUTE_VERBOSE
 	if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
 		/*
@@ -1763,7 +1765,7 @@ int ip_route_input(struct sk_buff *skb, u32 daddr, u32 saddr,
 			rth->u.dst.lastuse = jiffies;
 			dst_hold(&rth->u.dst);
 			rth->u.dst.__use++;
-			rt_cache_stat[smp_processor_id()].in_hit++;
+			RT_CACHE_STAT_INC(in_hit);
 			rcu_read_unlock();
 			skb->dst = (struct dst_entry*)rth;
 			return 0;
@@ -2060,7 +2062,7 @@ int ip_route_output_slow(struct rtable **rp, const struct flowi *oldflp)
 
 	rth->u.dst.output=ip_output;
 
-	rt_cache_stat[smp_processor_id()].out_slow_tot++;
+	RT_CACHE_STAT_INC(out_slow_tot);
 
 	if (flags & RTCF_LOCAL) {
 		rth->u.dst.input = ip_local_deliver;
@@ -2070,7 +2072,7 @@ int ip_route_output_slow(struct rtable **rp, const struct flowi *oldflp)
 		rth->rt_spec_dst = fl.fl4_src;
 		if (flags & RTCF_LOCAL && !(dev_out->flags & IFF_LOOPBACK)) {
 			rth->u.dst.output = ip_mc_output;
-			rt_cache_stat[smp_processor_id()].out_slow_mc++;
+			RT_CACHE_STAT_INC(out_slow_mc);
 		}
 #ifdef CONFIG_IP_MROUTE
 		if (res.type == RTN_MULTICAST) {
@@ -2129,7 +2131,7 @@ int __ip_route_output_key(struct rtable **rp, const struct flowi *flp)
 			rth->u.dst.lastuse = jiffies;
 			dst_hold(&rth->u.dst);
 			rth->u.dst.__use++;
-			rt_cache_stat[smp_processor_id()].out_hit++;
+			RT_CACHE_STAT_INC(out_hit);
 			rcu_read_unlock();
 			*rp = rth;
 			return 0;
@@ -2650,6 +2652,11 @@ int __init ip_rt_init(void)
 	ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1);
 	ip_rt_max_size = (rt_hash_mask + 1) * 16;
 
+	rt_cache_stat = kmalloc_percpu(sizeof (struct rt_cache_stat),
+					GFP_KERNEL);
+	if (!rt_cache_stat) 
+		goto out_enomem1;
+
 	devinet_init();
 	ip_fib_init();
 
@@ -2675,6 +2682,8 @@ int __init ip_rt_init(void)
 out:
 	return rc;
 out_enomem:
+	kfree_percpu(rt_cache_stat);
+out_enomem1:
 	rc = -ENOMEM;
 	goto out;
 }

net/ipv4/tcp.c

@@ -258,13 +258,15 @@
 #include <net/icmp.h>
 #include <net/tcp.h>
 #include <net/xfrm.h>
+#include <net/ip.h>
+
 
 #include <asm/uaccess.h>
 #include <asm/ioctls.h>
 
 int sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT;
 
-struct tcp_mib	tcp_statistics[NR_CPUS * 2];
+DEFINE_SNMP_STAT(struct tcp_mib, tcp_statistics);
 
 kmem_cache_t *tcp_openreq_cachep;
 kmem_cache_t *tcp_bucket_cachep;
@@ -1395,8 +1397,7 @@ static void tcp_prequeue_process(struct sock *sk)
 	struct sk_buff *skb;
 	struct tcp_opt *tp = tcp_sk(sk);
 
-	net_statistics[smp_processor_id() * 2 + 1].TCPPrequeued +=
-					    skb_queue_len(&tp->ucopy.prequeue);
+	NET_ADD_STATS_USER(TCPPrequeued, skb_queue_len(&tp->ucopy.prequeue));
 
 	/* RX process wants to run with disabled BHs, though it is not
 	 * necessary */
@@ -1676,7 +1677,7 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 			/* __ Restore normal policy in scheduler __ */
 
 			if ((chunk = len - tp->ucopy.len) != 0) {
-				net_statistics[smp_processor_id() * 2 + 1].TCPDirectCopyFromBacklog += chunk;
+				NET_ADD_STATS_USER(TCPDirectCopyFromBacklog, chunk);
 				len -= chunk;
 				copied += chunk;
 			}
@@ -1687,7 +1688,7 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 				tcp_prequeue_process(sk);
 
 				if ((chunk = len - tp->ucopy.len) != 0) {
-					net_statistics[smp_processor_id() * 2 + 1].TCPDirectCopyFromPrequeue += chunk;
+					NET_ADD_STATS_USER(TCPDirectCopyFromPrequeue, chunk);
 					len -= chunk;
 					copied += chunk;
 				}
@@ -1770,7 +1771,7 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 			tcp_prequeue_process(sk);
 
 			if (copied > 0 && (chunk = len - tp->ucopy.len) != 0) {
-				net_statistics[smp_processor_id() * 2 + 1].TCPDirectCopyFromPrequeue += chunk;
+				NET_ADD_STATS_USER(TCPDirectCopyFromPrequeue, chunk);
 				len -= chunk;
 				copied += chunk;
 			}

net/ipv4/tcp_input.c

@@ -3043,8 +3043,8 @@ static int tcp_prune_queue(struct sock *sk)
 
 	/* First, purge the out_of_order queue. */
 	if (skb_queue_len(&tp->out_of_order_queue)) {
-		net_statistics[smp_processor_id() * 2].OfoPruned +=
-					skb_queue_len(&tp->out_of_order_queue);
+		NET_ADD_STATS_BH(OfoPruned,
+				 skb_queue_len(&tp->out_of_order_queue));
 		__skb_queue_purge(&tp->out_of_order_queue);
 
 		/* Reset SACK state.  A conforming SACK implementation will

net/ipv4/tcp_minisocks.c

@@ -464,7 +464,7 @@ static void SMP_TIMER_NAME(tcp_twkill)(unsigned long dummy)
 
 	if ((tcp_tw_count -= killed) != 0)
 		mod_timer(&tcp_tw_timer, jiffies+TCP_TWKILL_PERIOD);
-	net_statistics[smp_processor_id()*2].TimeWaited += killed;
+	NET_ADD_STATS_BH(TimeWaited, killed);
 out:
 	spin_unlock(&tw_death_lock);
 }
@@ -628,7 +628,7 @@ void SMP_TIMER_NAME(tcp_twcal_tick)(unsigned long dummy)
 out:
 	if ((tcp_tw_count -= killed) == 0)
 		del_timer(&tcp_tw_timer);
-	net_statistics[smp_processor_id()*2].TimeWaitKilled += killed;
+	NET_ADD_STATS_BH(TimeWaitKilled, killed);
 	spin_unlock(&tw_death_lock);
 }
 

net/ipv4/tcp_timer.c

@@ -237,7 +237,8 @@ static void tcp_delack_timer(unsigned long data)
 	if (skb_queue_len(&tp->ucopy.prequeue)) {
 		struct sk_buff *skb;
 
-		net_statistics[smp_processor_id()*2].TCPSchedulerFailed += skb_queue_len(&tp->ucopy.prequeue);
+		NET_ADD_STATS_BH(TCPSchedulerFailed,
+				  skb_queue_len(&tp->ucopy.prequeue));
 
 		while ((skb = __skb_dequeue(&tp->ucopy.prequeue)) != NULL)
 			sk->backlog_rcv(sk, skb);

net/ipv4/udp.c

@@ -110,7 +110,7 @@
  *	Snmp MIB for the UDP layer
  */
 
-struct udp_mib		udp_statistics[NR_CPUS*2];
+DEFINE_SNMP_STAT(struct udp_mib, udp_statistics);
 
 struct sock *udp_hash[UDP_HTABLE_SIZE];
 rwlock_t udp_hash_lock = RW_LOCK_UNLOCKED;

net/ipv6/af_inet6.c

@@ -619,6 +619,81 @@ inet6_unregister_protosw(struct inet_protosw *p)
 	inet_unregister_protosw(p);
 }
 
+static int __init init_ipv6_mibs(void)
+{
+	int i;
+ 
+	ipv6_statistics[0] = kmalloc_percpu(sizeof (struct ipv6_mib),
+						GFP_KERNEL);
+	if (!ipv6_statistics[0])
+		goto err_ip_mib0;
+	ipv6_statistics[1] = kmalloc_percpu(sizeof (struct ipv6_mib),
+						GFP_KERNEL);
+	if (!ipv6_statistics[1])
+		goto err_ip_mib1;
+	
+	icmpv6_statistics[0] = kmalloc_percpu(sizeof (struct icmpv6_mib),
+						GFP_KERNEL);
+	if (!icmpv6_statistics[0])
+		goto err_icmp_mib0;
+	icmpv6_statistics[1] = kmalloc_percpu(sizeof (struct icmpv6_mib),
+						GFP_KERNEL);
+	if (!icmpv6_statistics[1])
+		goto err_icmp_mib1;
+	
+	udp_stats_in6[0] = kmalloc_percpu(sizeof (struct udp_mib),
+						GFP_KERNEL);
+	if (!udp_stats_in6[0])
+		goto err_udp_mib0;
+	udp_stats_in6[1] = kmalloc_percpu(sizeof (struct udp_mib),
+						GFP_KERNEL);
+	if (!udp_stats_in6[1])
+		goto err_udp_mib1;
+
+	/* Zero all percpu versions of the mibs */
+	for (i = 0; i < NR_CPUS; i++) {
+		if (cpu_possible(i)) {
+		memset(per_cpu_ptr(ipv6_statistics[0], i), 0,
+				sizeof (struct ipv6_mib));
+		memset(per_cpu_ptr(ipv6_statistics[1], i), 0,
+				sizeof (struct ipv6_mib));
+		memset(per_cpu_ptr(icmpv6_statistics[0], i), 0,
+				sizeof (struct icmpv6_mib));
+		memset(per_cpu_ptr(icmpv6_statistics[1], i), 0,
+				sizeof (struct icmpv6_mib));
+		memset(per_cpu_ptr(udp_stats_in6[0], i), 0,
+				sizeof (struct udp_mib));
+		memset(per_cpu_ptr(udp_stats_in6[1], i), 0,
+				sizeof (struct udp_mib));
+		}
+	}
+	return 0;
+
+err_udp_mib1:
+	kfree_percpu(udp_stats_in6[0]);
+err_udp_mib0:
+	kfree_percpu(icmpv6_statistics[1]);
+err_icmp_mib1:
+	kfree_percpu(icmpv6_statistics[0]);
+err_icmp_mib0:
+	kfree_percpu(ipv6_statistics[1]);
+err_ip_mib1:
+	kfree_percpu(ipv6_statistics[0]);
+err_ip_mib0:
+	return -ENOMEM;
+	
+}
+
+static void __exit cleanup_ipv6_mibs(void)
+{
+	kfree_percpu(ipv6_statistics[0]);
+	kfree_percpu(ipv6_statistics[1]);
+	kfree_percpu(icmpv6_statistics[0]);
+	kfree_percpu(icmpv6_statistics[1]);
+	kfree_percpu(udp_stats_in6[0]);
+	kfree_percpu(udp_stats_in6[1]);
+}
+	
 static int __init inet6_init(void)
 {
 	struct sk_buff *dummy_skb;
@@ -669,6 +744,11 @@ static int __init inet6_init(void)
 	 */
 	(void) sock_register(&inet6_family_ops);
 
+	/* Initialise ipv6 mibs */
+	err = init_ipv6_mibs();
+	if (err)
+		goto init_mib_fail;
+	
 	/*
 	 *	ipngwg API draft makes clear that the correct semantics
 	 *	for TCP and UDP is to consider one TCP and UDP instance
@@ -735,6 +815,8 @@ static int __init inet6_init(void)
 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 	ipv6_sysctl_unregister();
 #endif
+	cleanup_ipv6_mibs();
+init_mib_fail:
 	return err;
 }
 module_init(inet6_init);
@@ -765,6 +847,7 @@ static void inet6_exit(void)
 #ifdef CONFIG_SYSCTL
 	ipv6_sysctl_unregister();	
 #endif
+	cleanup_ipv6_mibs();
 }
 module_exit(inet6_exit);
 #endif /* MODULE */

net/ipv6/icmp.c

@@ -65,7 +65,7 @@
 #include <asm/uaccess.h>
 #include <asm/system.h>
 
-struct icmpv6_mib icmpv6_statistics[NR_CPUS*2];
+DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics);
 
 /*
  *	ICMP socket for flow control.
@@ -377,7 +377,7 @@ void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info,
 	ip6_build_xmit(sk, icmpv6_getfrag, &msg, &fl, len, NULL, -1,
 		       MSG_DONTWAIT);
 	if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB)
-		(&(icmpv6_statistics[smp_processor_id()*2].Icmp6OutDestUnreachs))[type-1]++;
+		ICMP6_STATS_PTR_BH(Icmp6OutDestUnreachs) [type-1]++;
 	ICMP6_INC_STATS_BH(Icmp6OutMsgs);
 out:
 	icmpv6_xmit_unlock();
@@ -539,9 +539,9 @@ static int icmpv6_rcv(struct sk_buff *skb)
 	type = hdr->icmp6_type;
 
 	if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB)
-		(&icmpv6_statistics[smp_processor_id()*2].Icmp6InDestUnreachs)[type-ICMPV6_DEST_UNREACH]++;
+		ICMP6_STATS_PTR_BH(Icmp6InDestUnreachs)[type-ICMPV6_DEST_UNREACH]++;
 	else if (type >= ICMPV6_ECHO_REQUEST && type <= NDISC_REDIRECT)
-		(&icmpv6_statistics[smp_processor_id()*2].Icmp6InEchos)[type-ICMPV6_ECHO_REQUEST]++;
+		ICMP6_STATS_PTR_BH(Icmp6InEchos)[type-ICMPV6_ECHO_REQUEST]++;
 
 	switch (type) {
 	case ICMPV6_ECHO_REQUEST:

net/ipv6/ipv6_sockglue.c

@@ -51,7 +51,7 @@
 
 #include <asm/uaccess.h>
 
-struct ipv6_mib ipv6_statistics[NR_CPUS*2];
+DEFINE_SNMP_STAT(struct ipv6_mib, ipv6_statistics);
 
 static struct packet_type ipv6_packet_type =
 {

net/ipv6/proc.c

@@ -59,11 +59,11 @@ int afinet6_get_info(char *buffer, char **start, off_t offset, int length)
 static struct snmp6_item
 {
 	char *name;
-	unsigned long *ptr;
-	int   mibsize;
+	void **mib;
+	int   offset;
 } snmp6_list[] = {
 /* ipv6 mib according to draft-ietf-ipngwg-ipv6-mib-04 */
-#define SNMP6_GEN(x) { #x , &ipv6_statistics[0].x, sizeof(struct ipv6_mib)/sizeof(unsigned long) }
+#define SNMP6_GEN(x) { #x , (void **)ipv6_statistics, offsetof(struct ipv6_mib, x) }
 	SNMP6_GEN(Ip6InReceives),
 	SNMP6_GEN(Ip6InHdrErrors),
 	SNMP6_GEN(Ip6InTooBigErrors),
@@ -97,7 +97,7 @@ static struct snmp6_item
 		OutRouterAdvertisements too.
 		OutGroupMembQueries too.
  */
-#define SNMP6_GEN(x) { #x , &icmpv6_statistics[0].x, sizeof(struct icmpv6_mib)/sizeof(unsigned long) }
+#define SNMP6_GEN(x) { #x , (void **)icmpv6_statistics, offsetof(struct icmpv6_mib, x) }
 	SNMP6_GEN(Icmp6InMsgs),
 	SNMP6_GEN(Icmp6InErrors),
 	SNMP6_GEN(Icmp6InDestUnreachs),
@@ -127,7 +127,7 @@ static struct snmp6_item
 	SNMP6_GEN(Icmp6OutGroupMembResponses),
 	SNMP6_GEN(Icmp6OutGroupMembReductions),
 #undef SNMP6_GEN
-#define SNMP6_GEN(x) { "Udp6" #x , &udp_stats_in6[0].Udp##x, sizeof(struct udp_mib)/sizeof(unsigned long) }
+#define SNMP6_GEN(x) { "Udp6" #x , (void **)udp_stats_in6, offsetof(struct udp_mib, Udp##x) }
 	SNMP6_GEN(InDatagrams),
 	SNMP6_GEN(NoPorts),
 	SNMP6_GEN(InErrors),
@@ -135,16 +135,22 @@ static struct snmp6_item
 #undef SNMP6_GEN
 };
 
-static unsigned long fold_field(unsigned long *ptr, int size)
+static unsigned long
+fold_field(void *mib[], int offt)
 {
         unsigned long res = 0;
         int i;
  
-	for (i=0; i<NR_CPUS; i++) {
-		res += ptr[2*i*size];
-		res += ptr[(2*i+1)*size];
+        for (i = 0; i < NR_CPUS; i++) {
+                if (!cpu_possible(i))
+                        continue;
+                res +=
+                    *((unsigned long *) (((void *)per_cpu_ptr(mib[0], i)) +
+                                         offt));
+                res +=
+                    *((unsigned long *) (((void *)per_cpu_ptr(mib[1], i)) +
+                                         offt));
         }
-
         return res;
 }
 
@@ -155,7 +161,7 @@ int afinet6_get_snmp(char *buffer, char **start, off_t offset, int length)
 
 	for (i=0; i<sizeof(snmp6_list)/sizeof(snmp6_list[0]); i++)
 		len += sprintf(buffer+len, "%-32s\t%ld\n", snmp6_list[i].name,
-			       fold_field(snmp6_list[i].ptr, snmp6_list[i].mibsize));
+			       fold_field(snmp6_list[i].mib, snmp6_list[i].offset));
 
 	len -= offset;
 

net/ipv6/udp.c

@@ -51,7 +51,7 @@
 
 #include <net/checksum.h>
 
-struct udp_mib udp_stats_in6[NR_CPUS*2];
+DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6);
 
 /* XXX This is identical to tcp_ipv6.c:ipv6_rcv_saddr_equal, put
  * XXX it somewhere common. -DaveM

net/sctp/protocol.c

@@ -60,7 +60,7 @@
 /* Global data structures. */
 sctp_protocol_t sctp_proto;
 struct proc_dir_entry	*proc_net_sctp;
-struct sctp_mib sctp_statistics[NR_CPUS * 2];
+DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics);
 
 /* This is the global socket data structure used for responding to
  * the Out-of-the-blue (OOTB) packets.  A control sock will be created
@@ -653,6 +653,40 @@ int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
 	return 1;
 }
 
+static int __init init_sctp_mibs(void)
+{
+	int i;
+	
+	sctp_statistics[0] = kmalloc_percpu(sizeof (struct sctp_mib),
+					    GFP_KERNEL);
+	if (!sctp_statistics[0])
+		return -ENOMEM;
+	sctp_statistics[1] = kmalloc_percpu(sizeof (struct sctp_mib),
+					    GFP_KERNEL);
+	if (!sctp_statistics[1]) {
+		kfree_percpu(sctp_statistics[0]);
+		return -ENOMEM;
+	}
+
+	/* Zero all percpu versions of the mibs */
+	for (i = 0; i < NR_CPUS; i++) {
+		if (cpu_possible(i)) {
+			memset(per_cpu_ptr(sctp_statistics[0], i), 0,
+					sizeof (struct sctp_mib));
+			memset(per_cpu_ptr(sctp_statistics[1], i), 0,
+					sizeof (struct sctp_mib));
+		}
+	}
+	return 0;
+	
+}
+
+static void cleanup_sctp_mibs(void)
+{
+	kfree_percpu(sctp_statistics[0]);
+	kfree_percpu(sctp_statistics[1]);
+}
+
 /* Initialize the universe into something sensible.  */
 int sctp_init(void)
 {
@@ -666,6 +700,11 @@ int sctp_init(void)
 	/* Add SCTP to inetsw linked list.  */
 	inet_register_protosw(&sctp_protosw);
 
+	/* Allocate and initialise sctp mibs.  */
+	status = init_sctp_mibs();
+	if (status) 
+		goto err_init_mibs;
+		
 	/* Initialize proc fs directory.  */
 	sctp_proc_init();
 
@@ -805,6 +844,8 @@ int sctp_init(void)
 err_ahash_alloc:
 	sctp_dbg_objcnt_exit();
 	sctp_proc_exit();
+	cleanup_sctp_mibs();
+err_init_mibs:	
 	inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
 	inet_unregister_protosw(&sctp_protosw);
 	return status;
@@ -836,6 +877,7 @@ void sctp_exit(void)
 
 	sctp_dbg_objcnt_exit();
 	sctp_proc_exit();
+	cleanup_sctp_mibs();
 
 	inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
 	inet_unregister_protosw(&sctp_protosw);

net/socket.c

@@ -75,7 +75,6 @@
 #include <linux/cache.h>
 #include <linux/module.h>
 #include <linux/highmem.h>
-#include <linux/wireless.h>
 #include <linux/divert.h>
 #include <linux/mount.h>
 
@@ -189,10 +188,7 @@ static __inline__ void net_family_read_unlock(void)
  *	Statistics counters of the socket lists
  */
 
-static union {
-	int	counter;
-	char	__pad[SMP_CACHE_BYTES];
-} sockets_in_use[NR_CPUS] __cacheline_aligned = {{0}};
+static DEFINE_PER_CPU(int, sockets_in_use) = 0;
 
 /*
  *	Support routines. Move socket addresses back and forth across the kernel/user
@@ -475,7 +471,8 @@ struct socket *sock_alloc(void)
 	inode->i_uid = current->fsuid;
 	inode->i_gid = current->fsgid;
 
-	sockets_in_use[smp_processor_id()].counter++;
+	get_cpu_var(sockets_in_use)++;
+	put_cpu_var(sockets_in_use);
 	return sock;
 }
 
@@ -511,7 +508,8 @@ void sock_release(struct socket *sock)
 	if (sock->fasync_list)
 		printk(KERN_ERR "sock_release: fasync list not empty!\n");
 
-	sockets_in_use[smp_processor_id()].counter--;
+	get_cpu_var(sockets_in_use)--;
+	put_cpu_var(sockets_in_use);
 	if (!sock->file) {
 		iput(SOCK_INODE(sock));
 		return;
@@ -1851,7 +1849,7 @@ void socket_seq_show(struct seq_file *seq)
 	int counter = 0;
 
 	for (cpu = 0; cpu < NR_CPUS; cpu++)
-		counter += sockets_in_use[cpu].counter;
+		counter += per_cpu(sockets_in_use, cpu);
 
 	/* It can be negative, by the way. 8) */
 	if (counter < 0)