Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934399Ab3FSJyg (ORCPT ); Wed, 19 Jun 2013 05:54:36 -0400 Received: from mx0.eldamar.org.uk ([78.47.157.133]:46738 "EHLO mx0.eldamar.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934149Ab3FSJyd (ORCPT ); Wed, 19 Jun 2013 05:54:33 -0400 Date: Wed, 19 Jun 2013 10:54:25 +0100 From: Alexander Frolkin To: Julian Anastasov , lvs-devel@vger.kernel.org Cc: Wensong Zhang , Simon Horman , netdev@vger.kernel.org, linux-kernel@vger.kernel.org Subject: [PATCH] ipvs: SH fallback and L4 hashing Message-ID: <20130619095425.GA20145@eldamar.org.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-GPG-Key-Fingerprint: 7820 960F C361 C9CE 401F D07D 993A 2951 D970 4FA4 X-Operating-System: Linux 3.2.0-41-generic X-Editor: Vi X-Uptime: 10:34:40 up 37 days, 19:58, 9 users, load average: 0.35, 0.28, 0.26 User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6335 Lines: 197 By default the SH scheduler rejects connections that are hashed onto a realserver of weight 0. This patch adds a flag to make SH choose a different realserver in this case, instead of rejecting the connection. The patch also adds a flag to make SH include the source port (TCP, UDP, SCTP) in the hash as well as the source address. This basically allows for deterministic round-robin load balancing (i.e., where any director in a cluster of directors with identical config will send the same packet the same way). The flags are service flags (IP_VS_SVC_F_SCHED*) so that these options can be set per service. They are set using a new option to ipvsadm. Signed-off-by: Alexander Frolkin --- The patch is against the ipvs-next tree. diff --git a/include/uapi/linux/ip_vs.h b/include/uapi/linux/ip_vs.h index a245377..2945822 100644 --- a/include/uapi/linux/ip_vs.h +++ b/include/uapi/linux/ip_vs.h @@ -20,6 +20,12 @@ #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_SCHED1 0x0008 /* scheduler flag 1 */ +#define IP_VS_SVC_F_SCHED2 0x0010 /* scheduler flag 2 */ +#define IP_VS_SVC_F_SCHED3 0x0020 /* scheduler flag 3 */ + +#define IP_VS_SVC_F_SCHED_SH_FALLBACK IP_VS_SVC_F_SCHED1 /* SH fallback */ +#define IP_VS_SVC_F_SCHED_SH_PORT IP_VS_SVC_F_SCHED2 /* SH use port */ /* * Destination Server Flags diff --git a/net/netfilter/ipvs/ip_vs_sh.c b/net/netfilter/ipvs/ip_vs_sh.c index e0130f8..b7e2c5a 100644 --- a/net/netfilter/ipvs/ip_vs_sh.c +++ b/net/netfilter/ipvs/ip_vs_sh.c @@ -48,6 +48,10 @@ #include +#include +#include +#include + /* * IPVS SH bucket @@ -71,10 +75,19 @@ struct ip_vs_sh_state { struct rcu_head rcu_head; }; +/* Helper function to determine if server is unavailable */ +static inline bool is_unavailable(struct ip_vs_dest *dest) +{ + return atomic_read(&dest->weight) <= 0 || + dest->flags & IP_VS_DEST_F_OVERLOAD; +} + /* * Returns hash value for IPVS SH entry */ -static inline unsigned int ip_vs_sh_hashkey(int af, const union nf_inet_addr *addr) +static inline unsigned int +ip_vs_sh_hashkey(int af, const union nf_inet_addr *addr, + __be16 port, unsigned int offset) { __be32 addr_fold = addr->ip; @@ -83,7 +96,8 @@ static inline unsigned int ip_vs_sh_hashkey(int af, const union nf_inet_addr *ad addr_fold = addr->ip6[0]^addr->ip6[1]^ addr->ip6[2]^addr->ip6[3]; #endif - return (ntohl(addr_fold)*2654435761UL) & IP_VS_SH_TAB_MASK; + return (offset + (ntohs(port) + ntohl(addr_fold))*2654435761UL) & + IP_VS_SH_TAB_MASK; } @@ -91,12 +105,42 @@ static inline unsigned int ip_vs_sh_hashkey(int af, const union nf_inet_addr *ad * Get ip_vs_dest associated with supplied parameters. */ static inline struct ip_vs_dest * -ip_vs_sh_get(int af, struct ip_vs_sh_state *s, const union nf_inet_addr *addr) +ip_vs_sh_get(struct ip_vs_service *svc, struct ip_vs_sh_state *s, + const union nf_inet_addr *addr, __be16 port) { - return rcu_dereference(s->buckets[ip_vs_sh_hashkey(af, addr)].dest); + unsigned int hash = ip_vs_sh_hashkey(svc->af, addr, port, 0); + struct ip_vs_dest *dest = rcu_dereference(s->buckets[hash].dest); + + return (!dest || is_unavailable(dest)) ? NULL : dest; } +/* As ip_vs_sh_get, but with fallback if selected server is unavailable */ +static inline struct ip_vs_dest * +ip_vs_sh_get_fallback(struct ip_vs_service *svc, struct ip_vs_sh_state *s, + const union nf_inet_addr *addr, __be16 port) +{ + unsigned int offset; + unsigned int hash; + struct ip_vs_dest *dest; + + for (offset = 0; offset < IP_VS_SH_TAB_SIZE; offset++) { + hash = ip_vs_sh_hashkey(svc->af, addr, port, offset); + dest = rcu_dereference(s->buckets[hash].dest); + if (!dest) + break; + if (is_unavailable(dest)) + IP_VS_DBG_BUF(6, "SH: selected unavailable server " + "%s:%d (offset %d)", + IP_VS_DBG_ADDR(svc->af, &dest->addr), + ntohs(dest->port), offset); + else + return dest; + } + + return NULL; +} + /* * Assign all the hash buckets of the specified table with the service. */ @@ -213,13 +257,33 @@ static int ip_vs_sh_dest_changed(struct ip_vs_service *svc, } -/* - * If the dest flags is set with IP_VS_DEST_F_OVERLOAD, - * consider that the server is overloaded here. - */ -static inline int is_overloaded(struct ip_vs_dest *dest) +/* Helper function to get port number */ +static inline __be16 +ip_vs_sh_get_port(const struct sk_buff *skb, struct ip_vs_iphdr *iph) { - return dest->flags & IP_VS_DEST_F_OVERLOAD; + __be16 port; + struct tcphdr _tcph, *th; + struct udphdr _udph, *uh; + sctp_sctphdr_t _sctph, *sh; + + switch (iph->protocol) { + case IPPROTO_TCP: + th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph); + port = th->source; + break; + case IPPROTO_UDP: + uh = skb_header_pointer(skb, iph->len, sizeof(_udph), &_udph); + port = uh->source; + break; + case IPPROTO_SCTP: + sh = skb_header_pointer(skb, iph->len, sizeof(_sctph), &_sctph); + port = sh->source; + break; + default: + port = 0; + } + + return port; } @@ -232,15 +296,21 @@ ip_vs_sh_schedule(struct ip_vs_service *svc, const struct sk_buff *skb, { struct ip_vs_dest *dest; struct ip_vs_sh_state *s; + __be16 port = 0; IP_VS_DBG(6, "ip_vs_sh_schedule(): Scheduling...\n"); + if (svc->flags & IP_VS_SVC_F_SCHED_SH_PORT) + port = ip_vs_sh_get_port(skb, iph); + s = (struct ip_vs_sh_state *) svc->sched_data; - dest = ip_vs_sh_get(svc->af, s, &iph->saddr); - if (!dest - || !(dest->flags & IP_VS_DEST_F_AVAILABLE) - || atomic_read(&dest->weight) <= 0 - || is_overloaded(dest)) { + + if (svc->flags & IP_VS_SVC_F_SCHED_SH_FALLBACK) + dest = ip_vs_sh_get_fallback(svc, s, &iph->saddr, port); + else + dest = ip_vs_sh_get(svc, s, &iph->saddr, port); + + if (!dest) { ip_vs_scheduler_err(svc, "no destination available"); return NULL; } -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/