Skip to contentSkip to conversation

IPVS: checksum updates

RFC In review Simon Horman 4 participants 28 messages Sep 8, 2008 – Sep 11, 2008

Subsystems (4)
lore
Share & tools
lore

Simon Horman

author maintainer
#0 lore raw

cover IPVS: checksum updates

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

author maintainer
#1 lore raw

RFC 1/3 ipvs: handle PARTIAL_CHECKSUM

Now that LVS can load balance locally generated traffic, packets may come from the loopback device and thus may have a partial checksum.

The existing code allows for the case where there is no checksum at all for TCP, however Herbert Xu has confirmed that this is not legal.

---

This patch implements *_partial_csum_update() in the style of the existing *_fast_csum_update() code. A subsequent patch will reimplement these functions in terms of the more standard inet_proto_csum_replace*() functions.

 net/ipv4/ipvs/ip_vs_proto_tcp.c |   37 +++++++++++++++++++++++++++++++++++--
 net/ipv4/ipvs/ip_vs_proto_udp.c |   37 +++++++++++++++++++++++++++++++++++--
 2 files changed, 70 insertions(+), 4 deletions(-)
net/ipv4/ipvs/ip_vs_proto_tcp.c +35 -2
Index: lvs-2.6/net/ipv4/ipvs/ip_vs_proto_tcp.c
===================================================================
@@ -134,12 +134,34 @@ tcp_fast_csum_update(int af, struct tcph
 }
 
 
+static inline void
+tcp_partial_csum_update(int af, struct tcphdr *tcph,
+		     const union nf_inet_addr *oldip,
+		     const union nf_inet_addr *newip,
+		     __be16 oldlen, __be16 newlen)
+{
+#ifdef CONFIG_IP_VS_IPV6
+	if (af == AF_INET6)
+		tcph->check =
+			csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
+					 ip_vs_check_diff2(oldlen, newlen,
+						~csum_unfold(tcph->check))));
+	else
+#endif
+	tcph->check =
+		csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
+				ip_vs_check_diff2(oldlen, newlen,
+						~csum_unfold(tcph->check))));
+}
+
+
 static int
 tcp_snat_handler(struct sk_buff *skb,
 		 struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
 {
 	struct tcphdr *tcph;
 	unsigned int tcphoff;
+	int oldlen;
 
 #ifdef CONFIG_IP_VS_IPV6
 	if (cp->af == AF_INET6)
@@ -147,6 +169,7 @@ tcp_snat_handler(struct sk_buff *skb,
 	else
 #endif
 		tcphoff = ip_hdrlen(skb);
+	oldlen = skb->len - tcphoff;
 
 	/* csum_check requires unshared skb */
 	if (!skb_make_writable(skb, tcphoff+sizeof(*tcph)))
@@ -166,7 +189,11 @@ tcp_snat_handler(struct sk_buff *skb,
 	tcph->source = cp->vport;
 
 	/* Adjust TCP checksums */
-	if (!cp->app && (tcph->check != 0)) {
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		tcp_partial_csum_update(cp->af, tcph, &cp->daddr, &cp->vaddr,
+					htonl(oldlen),
+					htonl(skb->len - tcphoff));
+	} else if (!cp->app) {
 		/* Only port and addr are changed, do fast csum update */
 		tcp_fast_csum_update(cp->af, tcph, &cp->daddr, &cp->vaddr,
 				     cp->dport, cp->vport);
@@ -204,6 +231,7 @@ tcp_dnat_handler(struct sk_buff *skb,
 {
 	struct tcphdr *tcph;
 	unsigned int tcphoff;
+	int oldlen;
 
 #ifdef CONFIG_IP_VS_IPV6
 	if (cp->af == AF_INET6)
@@ -211,6 +239,7 @@ tcp_dnat_handler(struct sk_buff *skb,
 	else
 #endif
 		tcphoff = ip_hdrlen(skb);
+	oldlen = skb->len - tcphoff;
 
 	/* csum_check requires unshared skb */
 	if (!skb_make_writable(skb, tcphoff+sizeof(*tcph)))
@@ -235,7 +264,11 @@ tcp_dnat_handler(struct sk_buff *skb,
 	/*
 	 *	Adjust TCP checksums
 	 */
-	if (!cp->app && (tcph->check != 0)) {
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		tcp_partial_csum_update(cp->af, tcph, &cp->daddr, &cp->vaddr,
+					htonl(oldlen),
+					htonl(skb->len - tcphoff));
+	} else if (!cp->app) {
 		/* Only port and addr are changed, do fast csum update */
 		tcp_fast_csum_update(cp->af, tcph, &cp->vaddr, &cp->daddr,
 				     cp->vport, cp->dport);
net/ipv4/ipvs/ip_vs_proto_udp.c +35 -2
Index: lvs-2.6/net/ipv4/ipvs/ip_vs_proto_udp.c
===================================================================
@@ -141,12 +141,34 @@ udp_fast_csum_update(int af, struct udph
 		uhdr->check = CSUM_MANGLED_0;
 }
 
+static inline void
+udp_partial_csum_update(int af, struct udphdr *uhdr,
+		     const union nf_inet_addr *oldip,
+		     const union nf_inet_addr *newip,
+		     __be16 oldlen, __be16 newlen)
+{
+#ifdef CONFIG_IP_VS_IPV6
+	if (af == AF_INET6)
+		uhdr->check =
+			csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
+					 ip_vs_check_diff2(oldlen, newlen,
+						~csum_unfold(uhdr->check))));
+	else
+#endif
+	uhdr->check =
+		csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
+				ip_vs_check_diff2(oldlen, newlen,
+						~csum_unfold(uhdr->check))));
+}
+
+
 static int
 udp_snat_handler(struct sk_buff *skb,
 		 struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
 {
 	struct udphdr *udph;
 	unsigned int udphoff;
+	int oldlen;
 
 #ifdef CONFIG_IP_VS_IPV6
 	if (cp->af == AF_INET6)
@@ -154,6 +176,7 @@ udp_snat_handler(struct sk_buff *skb,
 	else
 #endif
 		udphoff = ip_hdrlen(skb);
+	oldlen = skb->len - udphoff;
 
 	/* csum_check requires unshared skb */
 	if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
@@ -177,7 +200,11 @@ udp_snat_handler(struct sk_buff *skb,
 	/*
 	 *	Adjust UDP checksums
 	 */
-	if (!cp->app && (udph->check != 0)) {
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		udp_partial_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
+					htonl(oldlen),
+					htonl(skb->len - udphoff));
+	} else if (!cp->app && (udph->check != 0)) {
 		/* Only port and addr are changed, do fast csum update */
 		udp_fast_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
 				     cp->dport, cp->vport);
@@ -216,6 +243,7 @@ udp_dnat_handler(struct sk_buff *skb,
 {
 	struct udphdr *udph;
 	unsigned int udphoff;
+	int oldlen;
 
 #ifdef CONFIG_IP_VS_IPV6
 	if (cp->af == AF_INET6)
@@ -223,6 +251,7 @@ udp_dnat_handler(struct sk_buff *skb,
 	else
 #endif
 		udphoff = ip_hdrlen(skb);
+	oldlen = skb->len - udphoff;
 
 	/* csum_check requires unshared skb */
 	if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
@@ -247,7 +276,11 @@ udp_dnat_handler(struct sk_buff *skb,
 	/*
 	 *	Adjust UDP checksums
 	 */
-	if (!cp->app && (udph->check != 0)) {
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		udp_partial_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
+					htonl(oldlen),
+					htonl(skb->len - udphoff));
+	} else if (!cp->app && (udph->check != 0)) {
 		/* Only port and addr are changed, do fast csum update */
 		udp_fast_csum_update(cp->af, udph, &cp->vaddr, &cp->daddr,
 				     cp->vport, cp->dport);
author maintainer
#3 lore raw

Re: [rfc 1/3] ipvs: handle PARTIAL_CHECKSUM

Quoted 12 lines · Herbert Xu

net/ipv4/ipvs/ip_vs_proto_tcp.c tcp_snat_handler(struct sk_buff *skb, · line 195

Quoted 7 lines · Simon Horman
net/ipv4/ipvs/ip_vs_proto_tcp.c +4 -1
 	/* Adjust TCP checksums */
-	if (!cp->app && (tcph->check != 0)) {
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		tcp_partial_csum_update(cp->af, tcph, &cp->daddr, &cp->vaddr,
+					htonl(oldlen),
+					htonl(skb->len - tcphoff));

I don't know what cp->app is but should we be updating the checksum when it's set? The previous code seems to want to compute a full checksum instead.

Hi Herbert,

If cp->app is not present, then only the destiantion IP address and possibly port will have been altered. If it is present, then other parts of the packet may also have been altered.

In the case where (skb->ip_summed == CHECKSUM_PARTIAL) as we are only concerned with checkumming the pseudo header, the only changes that nat could make that we care about are the address or the length. The latter may change if cp->app is set, but I think that my code handles this in tcp_partial_csum_update().

So I think that is safe to use tcp_partial_csum_update() if (skb->ip_summed == CHECKSUM_PARTIAL), regarless of if cp->app is set or not.

author maintainer
#5 lore raw

RFC 2/3 ipvs: Use inet_proto_csum_replace*()

In reply to Simon Horman

Implement IPVS checksuming functions using the common inet_proto_csum_replace*() functions. This patch adds inet_proto_csum_replace16() to handle IPv6, which isn't provided by netfilter as it doesn't handle IPv6 NAT.

---

 include/net/ip_vs.h             |   31 +++---------------
 net/ipv4/ipvs/ip_vs_proto.c     |   23 +++++++++++++
 net/ipv4/ipvs/ip_vs_proto_tcp.c |   62 +++++++++++++++---------------------
 net/ipv4/ipvs/ip_vs_proto_udp.c |   67 +++++++++++++++++----------------------
 4 files changed, 86 insertions(+), 97 deletions(-)
include/net/ip_vs.h +6 -25
Index: lvs-2.6/include/net/ip_vs.h
===================================================================
@@ -932,34 +932,15 @@ extern void ip_vs_nat_icmp(struct sk_buf
 #ifdef CONFIG_IP_VS_IPV6
 extern void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
 			      struct ip_vs_conn *cp, int dir);
-#endif
-
-extern __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
-
-static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
-{
-	__be32 diff[2] = { ~old, new };
-
-	return csum_partial((char *) diff, sizeof(diff), oldsum);
-}
 
-#ifdef CONFIG_IP_VS_IPV6
-static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
-					__wsum oldsum)
-{
-	__be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
-			    new[3],  new[2],  new[1],  new[0] };
-
-	return csum_partial((char *) diff, sizeof(diff), oldsum);
-}
+#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
+extern void inet_proto_csum_replace16(__sum16 *sum, struct sk_buff *skb,
+				      __be32 from[4], __be32 to[4],
+				      int pseudohdr);
+#endif
 #endif
 
-static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
-{
-	__be16 diff[2] = { ~old, new };
-
-	return csum_partial((char *) diff, sizeof(diff), oldsum);
-}
+extern __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
 
 #endif /* __KERNEL__ */
 
net/ipv4/ipvs/ip_vs_proto.c +23 -0
Index: lvs-2.6/net/ipv4/ipvs/ip_vs_proto.c
===================================================================
@@ -286,3 +286,26 @@ void ip_vs_protocol_cleanup(void)
 			unregister_ip_vs_protocol(pp);
 	}
 }
+
+
+#if defined(CONFIG_IP_VS_IPV6) &&					\
+	(defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP))
+void
+inet_proto_csum_replace16(__sum16 *sum, struct sk_buff *skb,
+			  __be32 from[4], __be32 to[4], int pseudohdr)
+{
+	__be32 diff[8] = { ~from[3], ~from[2], ~from[1], ~from[0],
+			   to[3],  to[2],  to[1], to[0] };
+
+	if (skb->ip_summed != CHECKSUM_PARTIAL) {
+		*sum = csum_fold(csum_partial(diff, sizeof(diff),
+				~csum_unfold(*sum)));
+		if (skb->ip_summed == CHECKSUM_COMPLETE && pseudohdr)
+			skb->csum = ~csum_partial(diff, sizeof(diff),
+						~skb->csum);
+	} else if (pseudohdr)
+		*sum = ~csum_fold(csum_partial(diff, sizeof(diff),
+				csum_unfold(*sum)));
+}
+#endif
+
net/ipv4/ipvs/ip_vs_proto_tcp.c +27 -35

97 lines not on this page · raw message

net/ipv4/ipvs/ip_vs_proto_udp.c +30 -37

101 lines not on this page · raw message

author maintainer
#6 lore raw

RFC 3/3 ipvs: Consolidate checksuming code

In reply to Simon Horman

I'm not sure if this is a good idea, but a lot
of code is dublicated both within and between
ip_vs_proto_udp.c and ip_vs_proto_tcp.c.

This removes a small ammount of duplication.

---

 include/net/ip_vs.h             |   10 ++++++----
 net/ipv4/ipvs/ip_vs_proto.c     |   23 ++++++++++++++++++++---
 net/ipv4/ipvs/ip_vs_proto_tcp.c |   25 +++----------------------
 net/ipv4/ipvs/ip_vs_proto_udp.c |   14 ++------------
 4 files changed, 31 insertions(+), 41 deletions(-)
include/net/ip_vs.h +6 -4
Index: lvs-2.6/include/net/ip_vs.h
===================================================================
@@ -933,11 +933,13 @@ extern void ip_vs_nat_icmp(struct sk_buf
 extern void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
 			      struct ip_vs_conn *cp, int dir);
 
-#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
-extern void inet_proto_csum_replace16(__sum16 *sum, struct sk_buff *skb,
-				      __be32 from[4], __be32 to[4],
-				      int pseudohdr);
 #endif
+
+#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
+extern void ip_vs_fast_csum_update(int af, __sum16 *sum, struct sk_buff *skb,
+				   union nf_inet_addr *oldip,
+				   union nf_inet_addr *newip,
+				   __be16 old2, __be16 new2, int pseudo2);
 #endif
 
 extern __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
net/ipv4/ipvs/ip_vs_proto.c +20 -3
Index: lvs-2.6/net/ipv4/ipvs/ip_vs_proto.c
===================================================================
@@ -288,9 +288,9 @@ void ip_vs_protocol_cleanup(void)
 }
 
 
-#if defined(CONFIG_IP_VS_IPV6) &&					\
-	(defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP))
-void
+#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
+#ifdef CONFIG_IP_VS_IPV6
+static void
 inet_proto_csum_replace16(__sum16 *sum, struct sk_buff *skb,
 			  __be32 from[4], __be32 to[4], int pseudohdr)
 {
@@ -309,3 +309,20 @@ inet_proto_csum_replace16(__sum16 *sum, 
 }
 #endif
 
+void
+ip_vs_fast_csum_update(int af, __sum16 *sum, struct sk_buff *skb,
+		       union nf_inet_addr *oldip, union nf_inet_addr *newip,
+		       __be16 old2, __be16 new2, int pseudo2)
+{
+#ifdef CONFIG_IP_VS_IPV6
+	if (af == AF_INET6)
+		inet_proto_csum_replace16(sum, skb, oldip->ip6, newip->ip6, 1);
+	else
+#endif
+		inet_proto_csum_replace4(sum, skb, oldip->ip, newip->ip, 1);
+	if (new2 != old2)
+		inet_proto_csum_replace2(sum, skb, old2, new2, pseudo2);
+	if (skb->ip_summed == CHECKSUM_COMPLETE)
+		skb->ip_summed = CHECKSUM_NONE;
+}
+#endif
net/ipv4/ipvs/ip_vs_proto_udp.c +2 -12
Index: lvs-2.6/net/ipv4/ipvs/ip_vs_proto_udp.c
===================================================================
@@ -124,14 +124,8 @@ __udp_csum_update(int af, __sum16 *sum, 
 		       union nf_inet_addr *oldip, union nf_inet_addr *newip,
 		       __be16 old2, __be16 new2, int pseudo2)
 {
-#ifdef CONFIG_IP_VS_IPV6
-	if (af == AF_INET6)
-		inet_proto_csum_replace16(sum, skb, oldip->ip6, newip->ip6, 1);
-	else
-#endif
-		inet_proto_csum_replace4(sum, skb, oldip->ip, newip->ip, 1);
-	if (old2 != new2)
-		inet_proto_csum_replace2(sum, skb, old2, new2, pseudo2);
+	ip_vs_fast_csum_update(af, sum, skb, oldip, newip,
+			       old2, new2, pseudo2);
 	if (!*sum)
 		*sum = CSUM_MANGLED_0;
 }
@@ -201,8 +195,6 @@ udp_snat_handler(struct sk_buff *skb,
 		/* Only port and addr are changed, do fast csum update */
 		udp_fast_csum_update(cp->af, &udph->check, skb, &cp->daddr,
 				     &cp->vaddr, cp->dport, cp->vport);
-		if (skb->ip_summed == CHECKSUM_COMPLETE)
-			skb->ip_summed = CHECKSUM_NONE;
 	} else {
 		/* full checksum calculation */
 		udph->check = 0;
@@ -277,8 +269,6 @@ udp_dnat_handler(struct sk_buff *skb,
 		/* Only port and addr are changed, do fast csum update */
 		udp_fast_csum_update(cp->af, &udph->check, skb, &cp->vaddr,
 				     &cp->daddr, cp->vport, cp->dport);
-		if (skb->ip_summed == CHECKSUM_COMPLETE)
-			skb->ip_summed = CHECKSUM_NONE;
 	} else {
 		/* full checksum calculation */
 		udph->check = 0;
net/ipv4/ipvs/ip_vs_proto_tcp.c +3 -22
Index: lvs-2.6/net/ipv4/ipvs/ip_vs_proto_tcp.c
===================================================================
@@ -114,27 +114,12 @@ tcp_conn_schedule(int af, struct sk_buff
 
 
 static inline void
-__tcp_csum_update(int af, __sum16 *sum, struct sk_buff *skb,
-		     union nf_inet_addr *oldip, union nf_inet_addr *newip,
-		     __be16 old2, __be16 new2, int pseudo2)
-{
-#ifdef CONFIG_IP_VS_IPV6
-	if (af == AF_INET6)
-		inet_proto_csum_replace16(sum, skb, oldip->ip6, newip->ip6, 1);
-	else
-#endif
-		inet_proto_csum_replace4(sum, skb, oldip->ip, newip->ip, 1);
-	if (old2 != new2)
-		inet_proto_csum_replace2(sum, skb, old2, new2, pseudo2);
-}
-
-
-static inline void
 tcp_fast_csum_update(int af, __sum16 *sum, struct sk_buff *skb,
 		     union nf_inet_addr *oldip, union nf_inet_addr *newip,
 		     __be16 oldport, __be16 newport)
 {
-	__tcp_csum_update(af, sum, skb, oldip, newip, oldport, newport, 0);
+	ip_vs_fast_csum_update(af, sum, skb, oldip, newip,
+			       oldport, newport, 0);
 }
 
 
@@ -143,7 +128,7 @@ tcp_partial_csum_update(int af, __sum16 
 		     union nf_inet_addr *oldip, union nf_inet_addr *newip,
 		     __be16 oldlen, __be16 newlen)
 {
-	__tcp_csum_update(af, sum, skb, oldip, newip, oldlen, newlen, 1);
+	ip_vs_fast_csum_update(af, sum, skb, oldip, newip, oldlen, newlen, 1);
 }
 
 
@@ -189,8 +174,6 @@ tcp_snat_handler(struct sk_buff *skb,
 		/* Only port and addr are changed, do fast csum update */
 		tcp_fast_csum_update(cp->af, &tcph->check, skb, &cp->daddr,
 				     &cp->vaddr, cp->dport, cp->vport);
-		if (skb->ip_summed == CHECKSUM_COMPLETE)
-			skb->ip_summed = CHECKSUM_NONE;
 	} else {
 		/* full checksum calculation */
 		tcph->check = 0;
@@ -264,8 +247,6 @@ tcp_dnat_handler(struct sk_buff *skb,
 		/* Only port and addr are changed, do fast csum update */
 		tcp_fast_csum_update(cp->af, &tcph->check, skb, &cp->vaddr,
 				     &cp->daddr, cp->vport, cp->dport);
-		if (skb->ip_summed == CHECKSUM_COMPLETE)
-			skb->ip_summed = CHECKSUM_NONE;
 	} else {
 		/* full checksum calculation */
 		tcph->check = 0;
#7 lore raw
In reply to Simon Horman
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#8 lore raw
Quoted 16 lines · Julius Volz
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

Quoted 8 lines

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Ok, is this for both IPv4 & IPv6? Does it still occur with just the first patch in this series applied?

#9 lore raw
Quoted 22 lines · Simon Horman
Quoted 21 lines · Julius Volz
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Quoted 6 lines
Quoted 3 lines

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Ok, is this for both IPv4 & IPv6? Does it still occur with just the first patch in this series applied?

It's for both, although I only tested IPv4 at first. Here is a complete test matrix of what works when:

CR = connection refused
T = connection timeout
C = connection established, but not working afterwards OK = working

remote client | local client
COMMIT v4 v6 | v4 v6
======================================|=================

CSUM 3/3		OK	T     |	C	T
CSUM 2/3		OK	T     |	C	T
CSUM 1/3		OK	T     |	OK	T
W/O CSUM		OK	T     |	C	T

... |
f2428ed5 OK T | CR CR
4856c84c OK CR | CR CR
f94fd041 (my last one) OK OK | CR CR

So the last time that IPv6 was working _at all_ was at my last commit of the big v6 series...

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#10 lore raw
Quoted 45 lines · Julius Volz
Quoted 22 lines · Simon Horman
Quoted 21 lines · Julius Volz
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Quoted 6 lines
Quoted 3 lines

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Ok, is this for both IPv4 & IPv6? Does it still occur with just the first patch in this series applied?

It's for both, although I only tested IPv4 at first. Here is a complete test matrix of what works when:

CR = connection refused
T = connection timeout
C = connection established, but not working afterwards OK = working

remote client | local client
COMMIT v4 v6 | v4 v6
======================================|=================

CSUM 3/3		OK	T     |	C	T
CSUM 2/3		OK	T     |	C	T
CSUM 1/3		OK	T     |	OK	T
W/O CSUM		OK	T     |	C	T
Quoted 7 lines

... |
f2428ed5 OK T | CR CR
4856c84c OK CR | CR CR
f94fd041 (my last one) OK OK | CR CR

So the last time that IPv6 was working _at all_ was at my last commit of the big v6 series...

Ok, I'm really sorry about that :-(

Do you want me to revert f2428ed5 & 4856c84c until this has been tracked down?

author maintainer
#11 lore raw
Quoted 51 lines · Simon Horman
Quoted 50 lines · Julius Volz
Quoted 22 lines · Simon Horman
Quoted 21 lines · Julius Volz
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Quoted 6 lines
Quoted 3 lines

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Ok, is this for both IPv4 & IPv6? Does it still occur with just the first patch in this series applied?

It's for both, although I only tested IPv4 at first. Here is a complete test matrix of what works when:

CR = connection refused
T = connection timeout
C = connection established, but not working afterwards OK = working

remote client | local client
COMMIT v4 v6 | v4 v6
======================================|=================

CSUM 3/3		OK	T     |	C	T
CSUM 2/3		OK	T     |	C	T
CSUM 1/3		OK	T     |	OK	T
W/O CSUM		OK	T     |	C	T

... |
f2428ed5 OK T | CR CR
4856c84c OK CR | CR CR
f94fd041 (my last one) OK OK | CR CR

Quoted 7 lines
Quoted 2 lines

So the last time that IPv6 was working _at all_ was at my last commit of the big v6 series...

Ok, I'm really sorry about that :-(

Do you want me to revert f2428ed5 & 4856c84c until this has been tracked down?

Hi,

Does 4856c84c + the following change (which you pointed out over the weekend) work for remote IPv6 ?

net/ipv4/ipvs/ip_vs_core.c +1 -1
index 26e3d99..c413444 100644
@@ -1282,7 +1282,7 @@ ip_vs_in(unsigned int hooknum, struct sk_buff *skb,
 	 *	Don't handle local packets on IPv6 for now
 	 */
 	if (unlikely(skb->pkt_type != PACKET_HOST ||
-		     (af == AF_INET6 || (skb->dev->flags & IFF_LOOPBACK ||
+		     (af == AF_INET6 && (skb->dev->flags & IFF_LOOPBACK ||
 					 skb->sk)))) {
 		IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n",
 			      skb->pkt_type,
#12 lore raw
Quoted 71 lines · Simon Horman
Quoted 51 lines · Simon Horman
Quoted 50 lines · Julius Volz
Quoted 22 lines · Simon Horman
Quoted 21 lines · Julius Volz
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Quoted 6 lines
Quoted 3 lines

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Ok, is this for both IPv4 & IPv6? Does it still occur with just the first patch in this series applied?

It's for both, although I only tested IPv4 at first. Here is a complete test matrix of what works when:

CR = connection refused
T = connection timeout
C = connection established, but not working afterwards OK = working

remote client | local client
COMMIT v4 v6 | v4 v6
======================================|=================

CSUM 3/3            OK      T     | C       T
CSUM 2/3            OK      T     | C       T
CSUM 1/3            OK      T     | OK      T
W/O CSUM            OK      T     | C       T
...                               |
f2428ed5            OK      T     | CR      CR
4856c84c            OK      CR    | CR      CR

f94fd041 (my last one) OK OK | CR CR

Quoted 7 lines
Quoted 2 lines

So the last time that IPv6 was working _at all_ was at my last commit of the big v6 series...

Ok, I'm really sorry about that :-(

Do you want me to revert f2428ed5 & 4856c84c until this has been tracked down?

Hi,

Does 4856c84c + the following change (which you pointed out over the weekend) work for remote IPv6 ?

net/ipv4/ipvs/ip_vs_core.c +0 -0
index 26e3d99..c413444 100644

net/ipv4/ipvs/ip_vs_core.c ip_vs_in(unsigned int hooknum, struct sk_buff *skb, · line 1288

Quoted 8 lines
net/ipv4/ipvs/ip_vs_core.c +1 -1
@@ -1282,7 +1282,7 @@ ip_vs_in(unsigned int hooknum, struct sk_buff *skb,
        *      Don't handle local packets on IPv6 for now
        */
       if (unlikely(skb->pkt_type != PACKET_HOST ||
-                    (af == AF_INET6 || (skb->dev->flags & IFF_LOOPBACK ||
+                    (af == AF_INET6 && (skb->dev->flags & IFF_LOOPBACK ||
                                        skb->sk)))) {
               IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n",
                             skb->pkt_type,

No, that just changes the behavior from "connection refused" to timeout...

I'm actually looking at that case now (4856c84c1358b, but with the fix above). It seems that the NAT isn't working (DR works, by the way!). At least the first packet arriving at the real server still has the client's IP as the source (in the v6 case)...

Let's wait with reverting the local client patches until tomorrow... maybe I can find the problem until then.

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#13 lore raw
Quoted 83 lines · Julius Volz
Quoted 71 lines · Simon Horman
Quoted 51 lines · Simon Horman
Quoted 50 lines · Julius Volz
Quoted 22 lines · Simon Horman
Quoted 21 lines · Julius Volz
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After some first tests, remote connections are still working, but not local ones from the director. The TCP handshake works and the connection is established, but all following packets arriving at the real server have an incorrect TCP checksum.

Quoted 6 lines
Quoted 3 lines

Btw., this happens both with and without this last series of patches, so I can't get the local client feature working at all. Looking at it further...

Ok, is this for both IPv4 & IPv6? Does it still occur with just the first patch in this series applied?

It's for both, although I only tested IPv4 at first. Here is a complete test matrix of what works when:

CR = connection refused
T = connection timeout
C = connection established, but not working afterwards OK = working

remote client | local client
COMMIT v4 v6 | v4 v6
======================================|=================

CSUM 3/3            OK      T     | C       T
CSUM 2/3            OK      T     | C       T
CSUM 1/3            OK      T     | OK      T
W/O CSUM            OK      T     | C       T
...                               |
f2428ed5            OK      T     | CR      CR
4856c84c            OK      CR    | CR      CR

f94fd041 (my last one) OK OK | CR CR

Quoted 7 lines
Quoted 2 lines

So the last time that IPv6 was working _at all_ was at my last commit of the big v6 series...

Ok, I'm really sorry about that :-(

Do you want me to revert f2428ed5 & 4856c84c until this has been tracked down?

Hi,

Does 4856c84c + the following change (which you pointed out over the weekend) work for remote IPv6 ?

net/ipv4/ipvs/ip_vs_core.c +0 -0
index 26e3d99..c413444 100644

net/ipv4/ipvs/ip_vs_core.c ip_vs_in(unsigned int hooknum, struct sk_buff *skb, · line 1288

Quoted 8 lines
net/ipv4/ipvs/ip_vs_core.c +1 -1
@@ -1282,7 +1282,7 @@ ip_vs_in(unsigned int hooknum, struct sk_buff *skb,
        *      Don't handle local packets on IPv6 for now
        */
       if (unlikely(skb->pkt_type != PACKET_HOST ||
-                    (af == AF_INET6 || (skb->dev->flags & IFF_LOOPBACK ||
+                    (af == AF_INET6 && (skb->dev->flags & IFF_LOOPBACK ||
                                        skb->sk)))) {
               IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n",
                             skb->pkt_type,

No, that just changes the behavior from "connection refused" to timeout...

Quoted 4 lines

I'm actually looking at that case now (4856c84c1358b, but with the fix above). It seems that the NAT isn't working (DR works, by the way!). At least the first packet arriving at the real server still has the client's IP as the source (in the v6 case)...

Ok, I'm looking at NAT with 4856c84c1358b + that fix too too, but on v4 :-)

Quoted 2 lines

Let's wait with reverting the local client patches until tomorrow... maybe I can find the problem until then.

Ok, I was just concerned that this might hold up merging your code into Dave's tree for too long, thats all.

#14 lore raw
Quoted 6 lines · Simon Horman
Quoted 4 lines · Julius Volz

I'm actually looking at that case now (4856c84c1358b, but with the fix above). It seems that the NAT isn't working (DR works, by the way!). At least the first packet arriving at the real server still has the client's IP as the source (in the v6 case)...

Quoted 8 lines

Ok, I'm looking at NAT with 4856c84c1358b + that fix too too, but on v4 :-)

Quoted 2 lines

Let's wait with reverting the local client patches until tomorrow... maybe I can find the problem until then.

Ok, I was just concerned that this might hold up merging your code into Dave's tree for too long, thats all.

Ok, with my last patch (and your two/three last ones reverted), I get everything working except IPv6 with a local client...

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#15 lore raw
Quoted 10 lines · Julius Volz
Quoted 9 lines · Simon Horman
Quoted 4 lines · Julius Volz

I'm actually looking at that case now (4856c84c1358b, but with the fix above). It seems that the NAT isn't working (DR works, by the way!). At least the first packet arriving at the real server still has the client's IP as the source (in the v6 case)...

Ok, I'm looking at NAT with 4856c84c1358b + that fix too too, but on v4 :-)

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Let's wait with reverting the local client patches until tomorrow... maybe I can find the problem until then.

Ok, I was just concerned that this might hold up merging your code into Dave's tree for too long, thats all.

Ok, with my last patch (and your two/three last ones reverted), I get everything working except IPv6 with a local client...

Ok, that is good news :-)

Sorry for the stuff up. That bugus ip_route_me_harder() call was caused by me trying to merge the IPv6 patches with the local client patches.

I'm not so concerned about IPv6 local not working for now - its a combination of two new features. And hopefully we can iron it out in the not to distant future anyway.

I'll drop CSUM2/3 and 3/3 for now, they aren't essential and it really ought to be a matter of (me) sitting down and fixing some correctness issues.

Its the end of my day now. I'll try and update lvs-next-2.6 into a working state in the morning.

#16 lore raw
Quoted 17 lines · Simon Horman
Quoted 16 lines · Julius Volz
Quoted 14 lines · Simon Horman
Quoted 4 lines · Julius Volz

I'm actually looking at that case now (4856c84c1358b, but with the fix above). It seems that the NAT isn't working (DR works, by the way!). At least the first packet arriving at the real server still has the client's IP as the source (in the v6 case)...

Ok, I'm looking at NAT with 4856c84c1358b + that fix too too, but on v4 :-)

Quoted 2 lines

Let's wait with reverting the local client patches until tomorrow... maybe I can find the problem until then.

Ok, I was just concerned that this might hold up merging your code into Dave's tree for too long, thats all.

Quoted 8 lines
Quoted 2 lines

Ok, with my last patch (and your two/three last ones reverted), I get everything working except IPv6 with a local client...

Ok, that is good news :-)

Sorry for the stuff up. That bugus ip_route_me_harder() call was caused by me trying to merge the IPv6 patches with the local client patches.

No problem!

Quoted 3 lines

I'm not so concerned about IPv6 local not working for now - its a combination of two new features. And hopefully we can iron it out in the not to distant future anyway.

Yes, probably just some minor bug...

Quoted 3 lines

I'll drop CSUM2/3 and 3/3 for now, they aren't essential and it really ought to be a matter of (me) sitting down and fixing some correctness issues.

Ok!

Quoted 2 lines

Its the end of my day now. I'll try and update lvs-next-2.6 into a working state in the morning.

Yes, let's stabilize it a bit... have a nice evening!

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
#17 lore raw
Quoted 6 lines · Julius Volz
Quoted 3 lines · Simon Horman

I'm not so concerned about IPv6 local not working for now - its a combination of two new features. And hopefully we can iron it out in the not to distant future anyway.

Yes, probably just some minor bug...

Actually, it wasn't even a kernel bug, but a configuration error - I was using LVS/NAT, but still had the dummy0 interface with the IPv6 VIP configured on the real server (from the DR case). And in the local client case, the source address is the VIP, so the real server just delivered the response to itself.

Now I can confirm that everything is working (without your last two patches and with the one removing the wrong ip_route_me_harder)!

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#18 lore raw
Quoted 8 lines · Julius Volz
Quoted 6 lines · Julius Volz
Quoted 3 lines · Simon Horman

I'm not so concerned about IPv6 local not working for now - its a combination of two new features. And hopefully we can iron it out in the not to distant future anyway.

Yes, probably just some minor bug...

Quoted 8 lines

Actually, it wasn't even a kernel bug, but a configuration error - I was using LVS/NAT, but still had the dummy0 interface with the IPv6 VIP configured on the real server (from the DR case). And in the local client case, the source address is the VIP, so the real server just delivered the response to itself.

Now I can confirm that everything is working (without your last two patches and with the one removing the wrong ip_route_me_harder)!

Great!

author maintainer
#19 lore raw
In reply to Simon Horman
Quoted 8 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

Quoted 6 lines

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After extensive testing by Julius Volz and limited testing by myself, I have applied the first patch, which does indeed allow packets with PARTIAL_CHECKSUM to work, to lvs-next-2.6. I have dropped the second two patches which produce bogus checksums.

I will revisit the second to patches at some point and try and produce versions that work.

#20 lore raw
Quoted 12 lines · Simon Horman
Quoted 11 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

Quoted 8 lines
Quoted 3 lines

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

After extensive testing by Julius Volz and limited testing by myself, I have applied the first patch, which does indeed allow packets with PARTIAL_CHECKSUM to work, to lvs-next-2.6. I have dropped the second two patches which produce bogus checksums.

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#21 lore raw
Quoted 17 lines · Julius Volz
Quoted 16 lines · Simon Horman
Quoted 14 lines · Simon Horman

Hi,

The impetus for this series of patches is Julian Anastasov noting that "load balance IPv4 connections from a local process" checks for 0 TCP checksums. Herbert Xu confirmed that this is not legal, even on loopback traffic, but that rather partial checksums are possible.

The first patch in this series is a proposed solution to handle partial checksums for both TCP and UDP.

The other two patches clean things up a bit.

I have not tested this code beyond compilation yet.

Quoted 8 lines
Quoted 4 lines

After extensive testing by Julius Volz and limited testing by myself, I have applied the first patch, which does indeed allow packets with PARTIAL_CHECKSUM to work, to lvs-next-2.6. I have dropped the second two patches which produce bogus checksums.

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

#22 lore raw
Quoted 7 lines · Simon Horman
Quoted 3 lines · Julius Volz

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

Thanks very much!

(CC: Wensong)
I guess it would make sense to start hosting the new ipvsadm version somewhere on http://www.linuxvirtualserver.org/ soon. It could still be marked as unstable and saying, "If you intend to use the experimental IPv6 support for IPVS introduced in kernel 2.6.XX, you will need this new version of ipvsadm: ...". Also, we can then point to it in the Kconfig help text in CONFIG_IP_VS_IPV6. What do you think?

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
author maintainer
#23 lore raw
Quoted 11 lines · Julius Volz
Quoted 7 lines · Simon Horman
Quoted 3 lines · Julius Volz

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

Thanks very much!

Quoted 8 lines

(CC: Wensong)
I guess it would make sense to start hosting the new ipvsadm version somewhere on http://www.linuxvirtualserver.org/ soon. It could still be marked as unstable and saying, "If you intend to use the experimental IPv6 support for IPVS introduced in kernel 2.6.XX, you will need this new version of ipvsadm: ...". Also, we can then point to it in the Kconfig help text in CONFIG_IP_VS_IPV6. What do you think?

Yes I agree. It would be good to start hosting tarballs of the new/unstable/experimental version of ipvsadm on www.linuxvirtualserver.org.

Wensong, can we also host some sort of repository for ipvsadm on linuxvirtualserver.org? If not, I guess the next most logical choice would be to use kernel.org.

#24 lore raw
Quoted 25 lines · Simon Horman
Quoted 13 lines · Julius Volz
Quoted 9 lines · Simon Horman
Quoted 4 lines · Julius Volz

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

Thanks very much!

Quoted 9 lines

(CC: Wensong)
I guess it would make sense to start hosting the new ipvsadm version somewhere on http://www.linuxvirtualserver.org/ soon. It could still be marked as unstable and saying, "If you intend to use the experimental IPv6 support for IPVS introduced in kernel 2.6.XX, you will need this new version of ipvsadm: ...". Also, we can then point to it in the Kconfig help text in CONFIG_IP_VS_IPV6. What do you think?

Quoted 8 lines

Yes I agree. It would be good to start hosting tarballs of the new/unstable/experimental version of ipvsadm on www.linuxvirtualserver.org.

Wensong, can we also host some sort of repository for ipvsadm on linuxvirtualserver.org? If not, I guess the next most logical choice would be to use kernel.org.

Hi Horms and Julius,

I have created a svn repository for ipvsadm at http://svn.linuxvirtualserver.org/repos/ipvsadm/ before, will create the check-in accounts for you right now.

Thanks,

Wensong

author maintainer
#25 lore raw
Quoted 34 lines · Wensong Zhang
Quoted 25 lines · Simon Horman
Quoted 13 lines · Julius Volz
Quoted 9 lines · Simon Horman
Quoted 4 lines · Julius Volz

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

Thanks very much!

Quoted 9 lines

(CC: Wensong)
I guess it would make sense to start hosting the new ipvsadm version somewhere on http://www.linuxvirtualserver.org/ soon. It could still be marked as unstable and saying, "If you intend to use the experimental IPv6 support for IPVS introduced in kernel 2.6.XX, you will need this new version of ipvsadm: ...". Also, we can then point to it in the Kconfig help text in CONFIG_IP_VS_IPV6. What do you think?

Quoted 8 lines

Yes I agree. It would be good to start hosting tarballs of the new/unstable/experimental version of ipvsadm on www.linuxvirtualserver.org.

Wensong, can we also host some sort of repository for ipvsadm on linuxvirtualserver.org? If not, I guess the next most logical choice would be to use kernel.org.

Quoted 5 lines

Hi Horms and Julius,

I have created a svn repository for ipvsadm at
http://svn.linuxvirtualserver.org/repos/ipvsadm/ before, will create the check-in accounts for you right now.

Great. I think that Vince needs an account too, he has done most (all?) of the relevant work so far.

Signature · 3 lines
Simon Horman
  VA Linux Systems Japan K.K., Sydney, Australia Satellite Office
  H: www.vergenet.net/~horms/             W: www.valinux.co.jp/en
#26 lore raw
Quoted 35 lines · Simon Horman
Quoted 34 lines · Wensong Zhang
Quoted 33 lines · Simon Horman
Quoted 13 lines · Julius Volz
Quoted 9 lines · Simon Horman
Quoted 4 lines · Julius Volz

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

Thanks very much!

Quoted 9 lines

(CC: Wensong)
I guess it would make sense to start hosting the new ipvsadm version somewhere on http://www.linuxvirtualserver.org/ soon. It could still be marked as unstable and saying, "If you intend to use the experimental IPv6 support for IPVS introduced in kernel 2.6.XX, you will need this new version of ipvsadm: ...". Also, we can then point to it in the Kconfig help text in CONFIG_IP_VS_IPV6. What do you think?

Yes I agree. It would be good to start hosting tarballs of the new/unstable/experimental version of ipvsadm on www.linuxvirtualserver.org.

Wensong, can we also host some sort of repository for ipvsadm on linuxvirtualserver.org? If not, I guess the next most logical choice would be to use kernel.org.

Quoted 5 lines
Quoted 5 lines

Hi Horms and Julius,

I have created a svn repository for ipvsadm at
http://svn.linuxvirtualserver.org/repos/ipvsadm/ before, will create the check-in accounts for you right now.

Thanks, cool!

Quoted 2 lines

Great. I think that Vince needs an account too, he has done most (all?) of the relevant work so far.

Yes, Vince did all of the ipvsadm porting!

Julius

Signature · 3 lines
Julius Volz - Corporate Operations - SysOps

Google Switzerland GmbH - Identification No.: CH-020.4.028.116-1
#27 lore raw
Quoted 55 lines · Julius Volz
Quoted 43 lines · Simon Horman
Quoted 41 lines · Wensong Zhang
Quoted 39 lines · Simon Horman
Quoted 17 lines · Julius Volz
Quoted 12 lines · Simon Horman
Quoted 5 lines · Julius Volz

Great, thanks! I have tested TCP+UDP, local+remote clients, v4+v6, NAT+DSR+TUN in all combinations that are expected to be working and found no problems.

Thanks once again for your testing. I'll send a pull request to Dave in the morning.

Thanks very much!

Quoted 10 lines

(CC: Wensong)
I guess it would make sense to start hosting the new ipvsadm version somewhere on http://www.linuxvirtualserver.org/ soon. It could still be marked as unstable and saying, "If you intend to use the experimental IPv6 support for IPVS introduced in kernel 2.6.XX, you will need this new version of ipvsadm: ...". Also, we can then point to it in the Kconfig help text in CONFIG_IP_VS_IPV6. What do you think?

Yes I agree. It would be good to start hosting tarballs of the new/unstable/experimental version of ipvsadm on www.linuxvirtualserver.org.

Wensong, can we also host some sort of repository for ipvsadm on linuxvirtualserver.org? If not, I guess the next most logical choice would be to use kernel.org.

Quoted 6 lines
Quoted 6 lines

Hi Horms and Julius,

I have created a svn repository for ipvsadm at
http://svn.linuxvirtualserver.org/repos/ipvsadm/ before, will create the check-in accounts for you right now.

Thanks, cool!

Quoted 9 lines
Quoted 3 lines

Great. I think that Vince needs an account too, he has done most (all?) of the relevant work so far.

Yes, Vince did all of the ipvsadm porting!

Julius

I'll create an account for Vince right now.

Thank you very much,

Wensong