add ali_lvs v3 beta,kernel-ipvs and tools-keepalived,ipvsadm,quagga

kernel/drivers/net/ixgbe/ixgbe.h

@@ -75,12 +75,12 @@
 		__func__ , ## args)))
 
 /* TX/RX descriptor defines */
-#define IXGBE_DEFAULT_TXD		512
+#define IXGBE_DEFAULT_TXD		4096
 #define IXGBE_DEFAULT_TX_WORK		256
 #define IXGBE_MAX_TXD			4096
 #define IXGBE_MIN_TXD			64
 
-#define IXGBE_DEFAULT_RXD		512
+#define IXGBE_DEFAULT_RXD		4096
 #define IXGBE_DEFAULT_RX_WORK		256
 #define IXGBE_MAX_RXD			4096
 #define IXGBE_MIN_RXD			64

kernel/drivers/net/ixgbe/ixgbe_main.c

@@ -3435,11 +3435,13 @@ static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
 			  0x6A3E67EA, 0x14364D17, 0x3BED200D};
 	u32 mrqc = 0, reta = 0;
 	u32 rxcsum;
-	int i, j;
+	int i, j, offset, rss_limit;
 	int maxq = adapter->ring_feature[RING_F_RSS].indices;
 	int mask;
 	u8 tcs = netdev_get_num_tc(adapter->netdev);
 
+	pr_info("maxq=%d,num_tx_queues=%d, tcs=%u\n",
+			maxq, adapter->num_tx_queues, tcs);
 #ifdef HAVE_MQPRIO
 	if (tcs)
 		maxq = min(maxq, adapter->num_tx_queues / tcs);
@@ -3449,15 +3451,19 @@ static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
 	for (i = 0; i < 10; i++)
 		IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), seed[i]);
 
+	rss_limit = min_t(int, adapter->num_tx_queues,
+				IXGBE_MAX_RSS_INDICES);
+	offset = (rss_limit > maxq) ? (rss_limit - maxq) : 0;
 	/* Fill out redirection table */
-	for (i = 0, j = 0; i < 128; i++, j++) {
-		if (j == maxq)
-			j = 0;
+	for (i = 0, j = offset; i < 128; i++, j++) {
+		if (j == (maxq + offset))
+			j = offset;
 		/* reta = 4-byte sliding window of
 		 * 0x00..(indices-1)(indices-1)00..etc. */
 		reta = (reta << 8) | (j * 0x11);
-		if ((i & 3) == 3)
+		if ((i & 3) == 3) {
 			IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
+		}
 	}
 
 	/* Disable indicating checksum in descriptor, enables RSS hash */
@@ -8725,7 +8731,8 @@ static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
 	}
 #endif /* IXGBE_FCOE */
 
-	if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
+	if (adapter->flags & (IXGBE_FLAG_FDIR_HASH_CAPABLE |
+				IXGBE_FLAG_FDIR_PERFECT_CAPABLE)) {
 		while (unlikely(txq >= dev->real_num_tx_queues))
 			txq -= dev->real_num_tx_queues;
 		return txq;

kernel/drivers/net/ixgbe/ixgbe_param.c

@@ -1022,9 +1022,15 @@ void __devinit ixgbe_check_options(struct ixgbe_adapter *adapter)
 
 		/* limit the number of queues for FDIR using RSS param */
 		if (feature[RING_F_RSS].indices && num_RSS > bd && RSS[bd])
+		{
 			feature[RING_F_FDIR].indices =
-						feature[RING_F_RSS].indices;
+				min_t(int, num_online_cpus(),
+					feature[RING_F_FDIR].indices);
 
+			DPRINTK(PROBE, INFO, "FDIR.indices:%d, RSS.indices:%d\n",
+					feature[RING_F_FDIR].indices,
+					feature[RING_F_RSS].indices);
+		}
 no_flow_director:
 		/* empty code line with semi-colon */ ;
 	}

kernel/include/linux/ip_vs.h

@@ -20,6 +20,7 @@
 #define IP_VS_SVC_F_PERSISTENT	0x0001	/* persistent port */
 #define IP_VS_SVC_F_HASHED	0x0002	/* hashed entry */
 #define IP_VS_SVC_F_ONEPACKET	0x0004	/* one-packet scheduling */
+#define IP_VS_SVC_F_SYNPROXY	0x8000	/* synproxy flag */
 
 /*
  *      Destination Server Flags
@@ -352,6 +353,8 @@ enum {
 	IPVS_SVC_ATTR_NETMASK,	/* persistent netmask */
 
 	IPVS_SVC_ATTR_STATS,	/* nested attribute for service stats */
+	IPVS_SVC_ATTR_EST_TIMEOUT,	/* establish timeout */
+
 	__IPVS_SVC_ATTR_MAX,
 };
 

kernel/include/net/ip_vs.h

@@ -287,7 +287,7 @@ struct ip_vs_protocol {
 			     struct ip_vs_protocol * pp,
 			     struct ip_vs_conn * cp);
 
-	int (*fnat_in_handler) (struct sk_buff ** skb_p,
+	int (*fnat_in_handler) (struct sk_buff * skb,
 				struct ip_vs_protocol * pp,
 				struct ip_vs_conn * cp);
 
@@ -381,6 +381,8 @@ struct ip_vs_conn {
 					 * state transition triggerd
 					 * synchronization
 					 */
+	u16 cpuid;
+
 	/* Control members */
 	struct ip_vs_conn *control;	/* Master control connection */
 	atomic_t n_control;	/* Number of controlled ones */
@@ -424,8 +426,19 @@ struct ip_vs_conn {
 
 	/* L2 direct response xmit */
 	struct net_device	*indev;
-	unsigned char		src_hwaddr[MAX_ADDR_LEN];
-	unsigned char		dst_hwaddr[MAX_ADDR_LEN];
+	unsigned char		src_hwaddr[ETH_ALEN];
+	unsigned char		dst_hwaddr[ETH_ALEN];
+	struct net_device	*dev_inside;
+	unsigned char		src_hwaddr_inside[ETH_ALEN];
+	unsigned char		dst_hwaddr_inside[ETH_ALEN];
+
+	int est_timeout;	/* Now, we decide that every VS
+				 * should have its private
+				 * establish state timeout for user requirement.
+				 * Each conn inherit this value from VS and
+				 * set this value into conn timer
+				 * when state change to establishment
+				 */
 };
 
 /*
@@ -449,6 +462,7 @@ struct ip_vs_service_user_kern {
 	unsigned flags;		/* virtual service flags */
 	unsigned timeout;	/* persistent timeout in sec */
 	u32 netmask;		/* persistent netmask */
+	unsigned est_timeout;	/* vs private establish state timeout */
 };
 
 struct ip_vs_dest_user_kern {
@@ -477,7 +491,6 @@ struct ip_vs_service {
 	struct list_head s_list;	/* for normal service table */
 	struct list_head f_list;	/* for fwmark-based service table */
 	atomic_t refcnt;	/* reference counter */
-	atomic_t usecnt;	/* use counter */
 
 	u16 af;			/* address family */
 	__u16 protocol;		/* which protocol (TCP/UDP) */
@@ -491,20 +504,26 @@ struct ip_vs_service {
 	/* for realservers list */
 	struct list_head destinations;	/* real server d-linked list */
 	__u32 num_dests;	/* number of servers */
+	long weight;           /* sum of servers weight */
 
 	/* for local ip address list, now only used in FULL NAT model */
 	struct list_head laddr_list;	/* local ip address list */
 	rwlock_t laddr_lock;	/* lock for protect curr_laddr */
 	__u32 num_laddrs;	/* number of local ip address */
 	struct list_head *curr_laddr;	/* laddr data list head */
 
-	struct ip_vs_stats *stats;	/* Use per-cpu statistics for the service */
+	struct ip_vs_stats stats;	/* statistics for the service */
 	struct ip_vs_app *inc;	/* bind conns to this app inc */
 
 	/* for scheduling */
 	struct ip_vs_scheduler *scheduler;	/* bound scheduler object */
 	rwlock_t sched_lock;	/* lock sched_data */
 	void *sched_data;	/* scheduler application data */
+
+	/* for VS private establish state timeout, it should be inherited by every connection data structure */
+	unsigned est_timeout;
+
+	struct ip_vs_service *svc0;	/* the svc of cpu0 */
 };
 
 /*
@@ -523,7 +542,7 @@ struct ip_vs_dest {
 	atomic_t weight;	/* server weight */
 
 	atomic_t refcnt;	/* reference counter */
-	struct ip_vs_stats *stats;	/* Use per-cpu statistics for destination server */
+	struct ip_vs_stats stats;	/* statistics for destination server */
 
 	/* connection counters and thresholds */
 	atomic_t activeconns;	/* active connections */
@@ -551,6 +570,7 @@ struct ip_vs_dest {
 struct ip_vs_laddr {
 	struct list_head n_list;	/* for the local address in the service */
 	u16 af;			/* address family */
+	u16 cpuid;		/* record the cpu laddr has been assigned */
 	union nf_inet_addr addr;	/* ip address */
 	atomic64_t port;	/* port counts */
 	atomic_t refcnt;	/* reference count */
@@ -710,6 +730,9 @@ enum {
 	FAST_XMIT_NO_MAC,
 	FAST_XMIT_SYNPROXY_SAVE,
 	FAST_XMIT_DEV_LOST,
+	FAST_XMIT_REJECT_INSIDE,
+	FAST_XMIT_PASS_INSIDE,
+	FAST_XMIT_SYNPROXY_SAVE_INSIDE,
 	RST_IN_SYN_SENT,
 	RST_OUT_SYN_SENT,
 	RST_IN_ESTABLISHED,
@@ -718,6 +741,8 @@ enum {
 	LRO_REJECT,
 	XMIT_UNEXPECTED_MTU,
 	CONN_SCHED_UNREACH,
+	SYNPROXY_NO_DEST,
+	CONN_EXCEEDED,
 	IP_VS_EXT_STAT_LAST
 };
 
@@ -767,7 +792,8 @@ extern void ip_vs_init_hash_table(struct list_head *table, int rows);
 #define CONFIG_IP_VS_TAB_BITS   22
 #endif
 
-#define IP_VS_CONN_TAB_BITS	CONFIG_IP_VS_TAB_BITS
+//#define IP_VS_CONN_TAB_BITS	CONFIG_IP_VS_TAB_BITS
+#define IP_VS_CONN_TAB_BITS	20
 #define IP_VS_CONN_TAB_SIZE     (1 << IP_VS_CONN_TAB_BITS)
 #define IP_VS_CONN_TAB_MASK     (IP_VS_CONN_TAB_SIZE - 1)
 
@@ -954,6 +980,12 @@ extern int sysctl_ip_vs_tcp_drop_entry;
 extern int sysctl_ip_vs_udp_drop_entry;
 extern int sysctl_ip_vs_conn_expire_tcp_rst;
 extern int sysctl_ip_vs_fast_xmit;
+extern int sysctl_ip_vs_fast_xmit_inside;
+extern int sysctl_ip_vs_csum_offload;
+extern int sysctl_ip_vs_reserve_core;
+extern int sysctl_ip_vs_conn_max_num;
+
+DECLARE_PER_CPU(spinlock_t, ip_vs_svc_lock);
 
 extern struct ip_vs_service *ip_vs_service_get(int af, __u32 fwmark,
 					       __u16 protocol,
@@ -964,7 +996,8 @@ extern struct ip_vs_service *ip_vs_lookup_vip(int af, __u16 protocol,
 
 static inline void ip_vs_service_put(struct ip_vs_service *svc)
 {
-	atomic_dec(&svc->usecnt);
+	if (likely(svc != NULL))
+		spin_unlock(&__get_cpu_var(ip_vs_svc_lock));
 }
 
 extern struct ip_vs_dest *ip_vs_lookup_real_service(int af, __u16 protocol,

kernel/include/net/ip_vs_synproxy.h

@@ -15,8 +15,10 @@
  * [19-16] snd_wscale
  * [15-0]  MSSIND
  */
-#define IP_VS_SYNPROXY_MSS_BITS 16
-#define IP_VS_SYNPROXY_MSS_MASK (((__u32)1 << IP_VS_SYNPROXY_MSS_BITS) - 1)
+#define IP_VS_SYNPROXY_OTHER_BITS 12
+#define IP_VS_SYNPROXY_OTHER_MASK (((__u32)1 << IP_VS_SYNPROXY_OTHER_BITS)-1)
+#define IP_VS_SYNPROXY_MSS_BITS 12
+#define IP_VS_SYNPROXY_MSS_MASK ((__u32)0xf << IP_VS_SYNPROXY_MSS_BITS)
 
 #define IP_VS_SYNPROXY_SACKOK_BIT 21
 #define IP_VS_SYNPROXY_SACKOK_MASK ((__u32)1 << IP_VS_SYNPROXY_SACKOK_BIT)
@@ -38,20 +40,8 @@ struct ip_vs_synproxy_opt {
 	u16 mss_clamp;		/* Maximal mss, negotiated at connection setup  */
 };
 
-/* 
- * For syncookie compute and check 
- */
-extern __u32 ip_vs_synproxy_cookie_v4_init_sequence(struct sk_buff *skb,
-						    struct ip_vs_synproxy_opt
-						    *opts);
-extern int ip_vs_synproxy_v4_cookie_check(struct sk_buff *skb, __u32 cookie,
-					  struct ip_vs_synproxy_opt *opt);
-
-extern __u32 ip_vs_synproxy_cookie_v6_init_sequence(struct sk_buff *skb,
-						    struct ip_vs_synproxy_opt
-						    *opts);
-extern int ip_vs_synproxy_v6_cookie_check(struct sk_buff *skb, __u32 cookie,
-					  struct ip_vs_synproxy_opt *opt);
+/* syncookie init */
+extern int ip_vs_net_secret_init(void);
 
 /*
  * Syn-proxy step 1 logic: receive client's Syn.

kernel/net/ipv4/syncookies.c

@@ -149,7 +149,9 @@ static __u16 const msstab[] = {
 	512 - 1,
 	536 - 1,
 	1024 - 1,
+	1280 - 1,
 	1440 - 1,
+	1452 - 1,
 	1460 - 1,
 	4312 - 1,
 	(__u16)-1
@@ -374,7 +376,7 @@ out:	return ret;
  * [21] SACKOK
  * [20] TimeStampOK
  * [19-16] snd_wscale
- * [15-0] MSSIND 
+ * [15-12] MSSIND 
  */
 __u32 ip_vs_synproxy_cookie_v4_init_sequence(struct sk_buff *skb, 
 					     struct ip_vs_synproxy_opt *opts) 
@@ -390,7 +392,7 @@ __u32 ip_vs_synproxy_cookie_v4_init_sequence(struct sk_buff *skb,
 		;
 	opts->mss_clamp = msstab[mssind] + 1;
 
-	data = mssind & IP_VS_SYNPROXY_MSS_MASK;
+	data = ((mssind & 0x0f) << IP_VS_SYNPROXY_MSS_BITS);
 	data |= opts->sack_ok << IP_VS_SYNPROXY_SACKOK_BIT;
 	data |= opts->tstamp_ok << IP_VS_SYNPROXY_TSOK_BIT;
 	data |= ((opts->snd_wscale & 0x0f) << IP_VS_SYNPROXY_SND_WSCALE_BITS);
@@ -424,11 +426,10 @@ int ip_vs_synproxy_v4_cookie_check(struct sk_buff * skb, __u32 cookie,
 	if(res == (__u32)-1) /* count is invalid, jiffies' >> jiffies */
 		goto out;
 
-	mssind = res & IP_VS_SYNPROXY_MSS_MASK;
+	mssind = (res & IP_VS_SYNPROXY_MSS_MASK) >> IP_VS_SYNPROXY_MSS_BITS;
 
 	memset(opt, 0, sizeof(struct ip_vs_synproxy_opt));
-
-	if (mssind < NUM_MSS) {
+	if ((mssind < NUM_MSS) && ((res & IP_VS_SYNPROXY_OTHER_MASK) == 0)) {
 		opt->mss_clamp = msstab[mssind] + 1;
 		opt->sack_ok = (res & IP_VS_SYNPROXY_SACKOK_MASK) >> 
 					IP_VS_SYNPROXY_SACKOK_BIT;

kernel/net/ipv6/syncookies.c

@@ -43,7 +43,9 @@ static __u16 const msstab[] = {
 	512 - 1,
 	536 - 1,
 	1024 - 1,
+	1280 - 1,
 	1440 - 1,
+	1452 - 1,
 	1460 - 1,
 	4312 - 1,
 	(__u16)-1
@@ -289,7 +291,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
  * [21] SACKOK
  * [20] TimeStampOK
  * [19-16] snd_wscale
- * [15-0] MSSIND 
+ * [15-12] MSSIND 
  */
 __u32 ip_vs_synproxy_cookie_v6_init_sequence(struct sk_buff *skb, 
 					     struct ip_vs_synproxy_opt *opts) 
@@ -305,7 +307,7 @@ __u32 ip_vs_synproxy_cookie_v6_init_sequence(struct sk_buff *skb,
 		;
 	opts->mss_clamp = msstab[mssind] + 1;
 
-	data = mssind & IP_VS_SYNPROXY_MSS_MASK;
+	data = ((mssind & 0x0f) << IP_VS_SYNPROXY_MSS_BITS);
 	data |= opts->sack_ok << IP_VS_SYNPROXY_SACKOK_BIT;
 	data |= opts->tstamp_ok << IP_VS_SYNPROXY_TSOK_BIT;
 	data |= ((opts->snd_wscale & 0x0f) << IP_VS_SYNPROXY_SND_WSCALE_BITS);
@@ -338,11 +340,11 @@ int ip_vs_synproxy_v6_cookie_check(struct sk_buff * skb, __u32 cookie,
 	if(res == (__u32)-1) /* count is invalid, jiffies' >> jiffies */
 		goto out;
 
-	mssind = res & IP_VS_SYNPROXY_MSS_MASK;
+	mssind = (res & IP_VS_SYNPROXY_MSS_MASK) >> IP_VS_SYNPROXY_MSS_BITS;
 
 	memset(opt, 0, sizeof(struct ip_vs_synproxy_opt));
 
-	if (mssind < NUM_MSS) {
+	if ((mssind < NUM_MSS) && ((res & IP_VS_SYNPROXY_OTHER_MASK) == 0)) {
 		opt->mss_clamp = msstab[mssind] + 1;
 		opt->sack_ok = (res & IP_VS_SYNPROXY_SACKOK_MASK) >> 
 					IP_VS_SYNPROXY_SACKOK_BIT;