[PATCH v3] net: bonding: Add support for IPV6 ns/na

Since ipv6 neighbor solicitation and advertisement messages
isn't handled gracefully in bonding6 driver, we can see packet
drop due to inconsistency bewteen mac address in the option
message and source MAC .

Another examples is ipv6 neighbor solicitation and advertisement
messages from VM via tap attached to host brighe, the src mac
mighe be changed through balance-alb mode, but it is not synced
with Link-layer address in the option message.

The patch implements bond6's tx handle for ipv6 neighbor
solicitation and advertisement messages.

                        Border-Leaf
                        /        \
                       /          \
                    Tunnel1    Tunnel2
                     /              \
                    /                \
                  Leaf-1--Tunnel3--Leaf-2
                    \                /
                     \              /
                      \            /
                       \          /
                       NIC1    NIC2
                        \      /
                        server

We can see in our lab the Border-Leaf receives occasionally
a NA packet which is assigned to NIC1 mac in ND/NS option
message, but actaully send out via NIC2 mac due to tx-alb,
as a result, it will cause inconsistency between MAC table
and ND Table in Border-Leaf, i.e, NIC1 = Tunnel2 in ND table
and  NIC1 = Tunnel1 in mac table.

And then, Border-Leaf starts to forward packet destinated
to the Server, it will only check the ND table entry in some
switch to encapsulate the destination MAC of the message as
NIC1 MAC, and then send it out from Tunnel2 by ND table.
Then, Leaf-2 receives the packet, it notices the destination
MAC of message is NIC1 MAC and should forword it to Tunne1
by Tunnel3.

However, this traffic forward will be failure due to split
horizon of VxLAN tunnels.

Suggested-by: Hu Yadi <redacted>
Reviewed-by: Jay Vosburgh<redacted>
Reviewed-by: Eric Dumazet<redacted>
Reported-by: kernel test robot <redacted>
Signed-off-by: Sun Shouxin <redacted>
---
 drivers/net/bonding/bond_alb.c | 131 +++++++++++++++++++++++++++++++++
 1 file changed, 131 insertions(+)

diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c
index 533e476988f2..b14017364594 100644
--- a/drivers/net/bonding/bond_alb.c
+++ b/drivers/net/bonding/bond_alb.c

@@ -22,6 +22,7 @@
 #include <asm/byteorder.h>
 #include <net/bonding.h>
 #include <net/bond_alb.h>
+#include <net/ndisc.h>
 
 static const u8 mac_v6_allmcast[ETH_ALEN + 2] __long_aligned = {
 	0x33, 0x33, 0x00, 0x00, 0x00, 0x01

@@ -1269,6 +1270,119 @@ static int alb_set_mac_address(struct bonding *bond, void *addr)
 	return res;
 }
 
+/*determine if the packet is NA or NS*/
+static bool alb_determine_nd(struct icmp6hdr *hdr)
+{
+	if (hdr->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT ||
+	    hdr->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
+		return true;
+	}
+
+	return false;
+}
+
+static void alb_change_nd_option(struct sk_buff *skb, void *data)
+{
+	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
+	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
+	struct net_device *dev = skb->dev;
+	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
+	struct ipv6hdr *ip6hdr = ipv6_hdr(skb);
+	u8 *lladdr = NULL;
+	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
+				offsetof(struct nd_msg, opt));
+
+	while (ndoptlen) {
+		int l;
+
+		switch (nd_opt->nd_opt_type) {
+		case ND_OPT_SOURCE_LL_ADDR:
+		case ND_OPT_TARGET_LL_ADDR:
+		lladdr = ndisc_opt_addr_data(nd_opt, dev);
+		break;
+
+		default:
+		lladdr = NULL;
+		break;
+		}
+
+		l = nd_opt->nd_opt_len << 3;
+
+		if (ndoptlen < l || l == 0)
+			return;
+
+		if (lladdr) {
+			memcpy(lladdr, data, dev->addr_len);
+			icmp6h->icmp6_cksum = 0;
+
+			icmp6h->icmp6_cksum = csum_ipv6_magic(&ip6hdr->saddr,
+							      &ip6hdr->daddr,
+						ntohs(ip6hdr->payload_len),
+						IPPROTO_ICMPV6,
+						csum_partial(icmp6h,
+							     ntohs(ip6hdr->payload_len), 0));
+		}
+		ndoptlen -= l;
+		nd_opt = ((void *)nd_opt) + l;
+	}
+}
+
+static u8 *alb_get_lladdr(struct sk_buff *skb)
+{
+	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
+	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
+	struct net_device *dev = skb->dev;
+	u8 *lladdr = NULL;
+	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
+				offsetof(struct nd_msg, opt));
+
+	while (ndoptlen) {
+		int l;
+
+		switch (nd_opt->nd_opt_type) {
+		case ND_OPT_SOURCE_LL_ADDR:
+		case ND_OPT_TARGET_LL_ADDR:
+			lladdr = ndisc_opt_addr_data(nd_opt, dev);
+			break;
+
+		default:
+			break;
+		}
+
+		l = nd_opt->nd_opt_len << 3;
+
+		if (ndoptlen < l || l == 0)
+			return NULL;
+
+		if (lladdr)
+			return lladdr;
+
+		ndoptlen -= l;
+		nd_opt = ((void *)nd_opt) + l;
+	}
+
+	return lladdr;
+}
+
+static void alb_set_nd_option(struct sk_buff *skb, struct bonding *bond,
+			      struct slave *tx_slave)
+{
+	struct ipv6hdr *ip6hdr;
+	struct icmp6hdr *hdr = NULL;
+
+	if (skb->protocol == htons(ETH_P_IPV6)) {
+		if (tx_slave && tx_slave !=
+		    rcu_access_pointer(bond->curr_active_slave)) {
+			ip6hdr = ipv6_hdr(skb);
+			if (ip6hdr->nexthdr == IPPROTO_ICMPV6) {
+				hdr = icmp6_hdr(skb);
+				if (alb_determine_nd(hdr))
+					alb_change_nd_option(skb, tx_slave->dev->dev_addr);
+			}
+		}
+	}
+}
+
 /************************ exported alb functions ************************/
 
 int bond_alb_initialize(struct bonding *bond, int rlb_enabled)

@@ -1415,6 +1529,7 @@ struct slave *bond_xmit_alb_slave_get(struct bonding *bond,
 	}
 	case ETH_P_IPV6: {
 		const struct ipv6hdr *ip6hdr;
+		struct icmp6hdr *hdr = NULL;
 
 		/* IPv6 doesn't really use broadcast mac address, but leave
 		 * that here just in case.

@@ -1446,6 +1561,21 @@ struct slave *bond_xmit_alb_slave_get(struct bonding *bond,
 			break;
 		}
 
+		if (ip6hdr->nexthdr == IPPROTO_ICMPV6) {
+			hdr = icmp6_hdr(skb);
+			if (alb_determine_nd(hdr)) {
+				u8 *lladdr = NULL;
+
+				lladdr = alb_get_lladdr(skb);
+				if (lladdr) {
+					if (!bond_slave_has_mac_rx(bond, lladdr)) {
+						do_tx_balance = false;
+						break;
+					}
+				}
+			}
+		}
+
 		hash_start = (char *)&ip6hdr->daddr;
 		hash_size = sizeof(ip6hdr->daddr);
 		break;

@@ -1489,6 +1619,7 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev)
 	struct slave *tx_slave = NULL;
 
 	tx_slave = bond_xmit_alb_slave_get(bond, skb);
+	alb_set_nd_option(skb, bond, tx_slave);
 	return bond_do_alb_xmit(skb, bond, tx_slave);
 }
 

base-commit: 6441998e2e37131b0a4c310af9156d79d3351c16
-- 
2.34.1

	For clarity, please add "to balance-alb mode" to the Subject.

Sun Shouxin [off-list ref] wrote:

Since ipv6 neighbor solicitation and advertisement messages
isn't handled gracefully in bonding6 driver, we can see packet
drop due to inconsistency bewteen mac address in the option
message and source MAC .

Another examples is ipv6 neighbor solicitation and advertisement
messages from VM via tap attached to host brighe, the src mac
mighe be changed through balance-alb mode, but it is not synced
with Link-layer address in the option message.

The patch implements bond6's tx handle for ipv6 neighbor
solicitation and advertisement messages.

                       Border-Leaf
                       /        \
                      /          \
                   Tunnel1    Tunnel2
                    /              \
                   /                \
                 Leaf-1--Tunnel3--Leaf-2
                   \                /
                    \              /
                     \            /
                      \          /
                      NIC1    NIC2
                       \      /
                       server

We can see in our lab the Border-Leaf receives occasionally
a NA packet which is assigned to NIC1 mac in ND/NS option
message, but actaully send out via NIC2 mac due to tx-alb,
as a result, it will cause inconsistency between MAC table
and ND Table in Border-Leaf, i.e, NIC1 = Tunnel2 in ND table
and  NIC1 = Tunnel1 in mac table.

And then, Border-Leaf starts to forward packet destinated
to the Server, it will only check the ND table entry in some
switch to encapsulate the destination MAC of the message as
NIC1 MAC, and then send it out from Tunnel2 by ND table.
Then, Leaf-2 receives the packet, it notices the destination
MAC of message is NIC1 MAC and should forword it to Tunne1
by Tunnel3.

However, this traffic forward will be failure due to split
horizon of VxLAN tunnels.

	I believe I understand what problem you're trying to solve here,
but the solution seems to be incomplete, as (from our prior discussion)
a rebalance event for balance-alb will apparently induce the same
problem.  Granted, those do not occur frequently (only when interfaces
are added to the bond, or an interface link state changes), but have you
tested what happens if NIC1 or NIC2 (or in a situation with more than
two interfaces) undergoes a link state change?

Suggested-by: Hu Yadi <redacted>
Reviewed-by: Jay Vosburgh<redacted>

	I did not include this signoff tag in my prior message.  Please
do not include such tags unless explicitly provided by the relevant
person.  Discussion on the mailing list is not equivalent to providing
the tag; please review Documentation/process/submitting-patches.rst.

	The above block is indented incorrectly (the "lladdr" and
"break" lines should be further in).

	If I'm reading RFC 4861 section 4.4 correctly, a Neighbor
Advertisement will only have ND_OPT_TARGET_LL_ADDR, and a Neighbor
Solicitation will only have ND_OPT_SOURCE_LL_ADDR.  Assuming that's a
correct reading, can the above break out of the loop after processing
the first TARGET or SOURCE option seen?

	hdr does not need to be initialized, as it's always assigned to
before being inspected.

	As above, hdr does not need to be initialized.

	-J

---
	-Jay Vosburgh, jay.vosburgh@canonical.com

在 2021/12/18 7:09, Jay Vosburgh 写道:

	For clarity, please add "to balance-alb mode" to the Subject.

Thanks your comment, I'll adjust it and send out V4 soon.

Sun Shouxin [off-list ref] wrote:

quoted

Since ipv6 neighbor solicitation and advertisement messages
isn't handled gracefully in bonding6 driver, we can see packet
drop due to inconsistency bewteen mac address in the option
message and source MAC .

Another examples is ipv6 neighbor solicitation and advertisement
messages from VM via tap attached to host brighe, the src mac
mighe be changed through balance-alb mode, but it is not synced
with Link-layer address in the option message.

The patch implements bond6's tx handle for ipv6 neighbor
solicitation and advertisement messages.

                        Border-Leaf
                        /        \
                       /          \
                    Tunnel1    Tunnel2
                     /              \
                    /                \
                  Leaf-1--Tunnel3--Leaf-2
                    \                /
                     \              /
                      \            /
                       \          /
                       NIC1    NIC2
                        \      /
                        server

We can see in our lab the Border-Leaf receives occasionally
a NA packet which is assigned to NIC1 mac in ND/NS option
message, but actaully send out via NIC2 mac due to tx-alb,
as a result, it will cause inconsistency between MAC table
and ND Table in Border-Leaf, i.e, NIC1 = Tunnel2 in ND table
and  NIC1 = Tunnel1 in mac table.

And then, Border-Leaf starts to forward packet destinated
to the Server, it will only check the ND table entry in some
switch to encapsulate the destination MAC of the message as
NIC1 MAC, and then send it out from Tunnel2 by ND table.
Then, Leaf-2 receives the packet, it notices the destination
MAC of message is NIC1 MAC and should forword it to Tunne1
by Tunnel3.

However, this traffic forward will be failure due to split
horizon of VxLAN tunnels.

	I believe I understand what problem you're trying to solve here,
but the solution seems to be incomplete, as (from our prior discussion)
a rebalance event for balance-alb will apparently induce the same
problem.  Granted, those do not occur frequently (only when interfaces
are added to the bond, or an interface link state changes), but have you
tested what happens if NIC1 or NIC2 (or in a situation with more than
two interfaces) undergoes a link state change?

The code in the bond_xmit_alb_slave_get should act for ns/na in the rebalance.
what's more, with NIC1/NIC2 link state change, I don't observe abnormal scene.

quoted

Suggested-by: Hu Yadi <redacted>
Reviewed-by: Jay Vosburgh<redacted>

	I did not include this signoff tag in my prior message.  Please
do not include such tags unless explicitly provided by the relevant
person.  Discussion on the mailing list is not equivalent to providing
the tag; please review Documentation/process/submitting-patches.rst.

Thanks your comment, I'll adjust it.

quoted

Reviewed-by: Eric Dumazet<redacted>
Reported-by: kernel test robot <redacted>
Signed-off-by: Sun Shouxin <redacted>
---
drivers/net/bonding/bond_alb.c | 131 +++++++++++++++++++++++++++++++++
1 file changed, 131 insertions(+)

diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c
index 533e476988f2..b14017364594 100644
--- a/drivers/net/bonding/bond_alb.c
+++ b/drivers/net/bonding/bond_alb.c

@@ -22,6 +22,7 @@

#include <asm/byteorder.h>
#include <net/bonding.h>
#include <net/bond_alb.h>
+#include <net/ndisc.h>

static const u8 mac_v6_allmcast[ETH_ALEN + 2] __long_aligned = {
	0x33, 0x33, 0x00, 0x00, 0x00, 0x01

@@ -1269,6 +1270,119 @@ static int alb_set_mac_address(struct bonding *bond, void *addr)

	return res;
}

+/*determine if the packet is NA or NS*/
+static bool alb_determine_nd(struct icmp6hdr *hdr)
+{
+	if (hdr->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT ||
+	    hdr->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
+		return true;
+	}
+
+	return false;
+}
+
+static void alb_change_nd_option(struct sk_buff *skb, void *data)
+{
+	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
+	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
+	struct net_device *dev = skb->dev;
+	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
+	struct ipv6hdr *ip6hdr = ipv6_hdr(skb);
+	u8 *lladdr = NULL;
+	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
+				offsetof(struct nd_msg, opt));
+
+	while (ndoptlen) {
+		int l;
+
+		switch (nd_opt->nd_opt_type) {
+		case ND_OPT_SOURCE_LL_ADDR:
+		case ND_OPT_TARGET_LL_ADDR:
+		lladdr = ndisc_opt_addr_data(nd_opt, dev);
+		break;
+
+		default:
+		lladdr = NULL;
+		break;
+		}

	The above block is indented incorrectly (the "lladdr" and
"break" lines should be further in).

Thanks your comment, I'll adjust it.

quoted

+
+		l = nd_opt->nd_opt_len << 3;
+
+		if (ndoptlen < l || l == 0)
+			return;
+
+		if (lladdr) {
+			memcpy(lladdr, data, dev->addr_len);
+			icmp6h->icmp6_cksum = 0;
+
+			icmp6h->icmp6_cksum = csum_ipv6_magic(&ip6hdr->saddr,
+							      &ip6hdr->daddr,
+						ntohs(ip6hdr->payload_len),
+						IPPROTO_ICMPV6,
+						csum_partial(icmp6h,
+							     ntohs(ip6hdr->payload_len), 0));
+		}
+		ndoptlen -= l;
+		nd_opt = ((void *)nd_opt) + l;

	If I'm reading RFC 4861 section 4.4 correctly, a Neighbor
Advertisement will only have ND_OPT_TARGET_LL_ADDR, and a Neighbor
Solicitation will only have ND_OPT_SOURCE_LL_ADDR.  Assuming that's a
correct reading, can the above break out of the loop after processing
the first TARGET or SOURCE option seen?

Thanks your comment, I'll adjust it.

quoted

+	}
+}
+
+static u8 *alb_get_lladdr(struct sk_buff *skb)
+{
+	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
+	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
+	struct net_device *dev = skb->dev;
+	u8 *lladdr = NULL;
+	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
+				offsetof(struct nd_msg, opt));
+
+	while (ndoptlen) {
+		int l;
+
+		switch (nd_opt->nd_opt_type) {
+		case ND_OPT_SOURCE_LL_ADDR:
+		case ND_OPT_TARGET_LL_ADDR:
+			lladdr = ndisc_opt_addr_data(nd_opt, dev);
+			break;
+
+		default:
+			break;
+		}
+
+		l = nd_opt->nd_opt_len << 3;
+
+		if (ndoptlen < l || l == 0)
+			return NULL;
+
+		if (lladdr)
+			return lladdr;
+
+		ndoptlen -= l;
+		nd_opt = ((void *)nd_opt) + l;
+	}
+
+	return lladdr;
+}
+
+static void alb_set_nd_option(struct sk_buff *skb, struct bonding *bond,
+			      struct slave *tx_slave)
+{
+	struct ipv6hdr *ip6hdr;
+	struct icmp6hdr *hdr = NULL;

	hdr does not need to be initialized, as it's always assigned to
before being inspected.

Thanks your comment, I'll adjust it.

quoted

+
+	if (skb->protocol == htons(ETH_P_IPV6)) {
+		if (tx_slave && tx_slave !=
+		    rcu_access_pointer(bond->curr_active_slave)) {
+			ip6hdr = ipv6_hdr(skb);
+			if (ip6hdr->nexthdr == IPPROTO_ICMPV6) {
+				hdr = icmp6_hdr(skb);
+				if (alb_determine_nd(hdr))
+					alb_change_nd_option(skb, tx_slave->dev->dev_addr);
+			}
+		}
+	}
+}
+
/************************ exported alb functions ************************/

int bond_alb_initialize(struct bonding *bond, int rlb_enabled)

@@ -1415,6 +1529,7 @@ struct slave *bond_xmit_alb_slave_get(struct bonding *bond,

	}
	case ETH_P_IPV6: {
		const struct ipv6hdr *ip6hdr;
+		struct icmp6hdr *hdr = NULL;

	As above, hdr does not need to be initialized.

	-J

Thanks your comment, I'll adjust it.

quoted

		/* IPv6 doesn't really use broadcast mac address, but leave
		 * that here just in case.

@@ -1446,6 +1561,21 @@ struct slave *bond_xmit_alb_slave_get(struct bonding *bond,

			break;
		}

+		if (ip6hdr->nexthdr == IPPROTO_ICMPV6) {
+			hdr = icmp6_hdr(skb);
+			if (alb_determine_nd(hdr)) {
+				u8 *lladdr = NULL;
+
+				lladdr = alb_get_lladdr(skb);
+				if (lladdr) {
+					if (!bond_slave_has_mac_rx(bond, lladdr)) {
+						do_tx_balance = false;
+						break;
+					}
+				}
+			}
+		}
+
		hash_start = (char *)&ip6hdr->daddr;
		hash_size = sizeof(ip6hdr->daddr);
		break;

@@ -1489,6 +1619,7 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev)

	struct slave *tx_slave = NULL;

	tx_slave = bond_xmit_alb_slave_get(bond, skb);
+	alb_set_nd_option(skb, bond, tx_slave);
	return bond_do_alb_xmit(skb, bond, tx_slave);
}


base-commit: 6441998e2e37131b0a4c310af9156d79d3351c16
-- 
2.34.1

---
	-Jay Vosburgh, jay.vosburgh@canonical.com

在 2021/12/20 23:03, Eric Dumazet 写道:

On 12/17/21 8:48 AM, Sun Shouxin wrote:

quoted

Since ipv6 neighbor solicitation and advertisement messages
isn't handled gracefully in bonding6 driver, we can see packet
drop due to inconsistency bewteen mac address in the option
message and source MAC .

Another examples is ipv6 neighbor solicitation and advertisement
messages from VM via tap attached to host brighe, the src mac
mighe be changed through balance-alb mode, but it is not synced
with Link-layer address in the option message.

The patch implements bond6's tx handle for ipv6 neighbor
solicitation and advertisement messages.

                         Border-Leaf
                         /        \
                        /          \
                     Tunnel1    Tunnel2
                      /              \
                     /                \
                   Leaf-1--Tunnel3--Leaf-2
                     \                /
                      \              /
                       \            /
                        \          /
                        NIC1    NIC2
                         \      /
                         server

We can see in our lab the Border-Leaf receives occasionally
a NA packet which is assigned to NIC1 mac in ND/NS option
message, but actaully send out via NIC2 mac due to tx-alb,
as a result, it will cause inconsistency between MAC table
and ND Table in Border-Leaf, i.e, NIC1 = Tunnel2 in ND table
and  NIC1 = Tunnel1 in mac table.

And then, Border-Leaf starts to forward packet destinated
to the Server, it will only check the ND table entry in some
switch to encapsulate the destination MAC of the message as
NIC1 MAC, and then send it out from Tunnel2 by ND table.
Then, Leaf-2 receives the packet, it notices the destination
MAC of message is NIC1 MAC and should forword it to Tunne1
by Tunnel3.

However, this traffic forward will be failure due to split
horizon of VxLAN tunnels.

Suggested-by: Hu Yadi <redacted>
Reviewed-by: Jay Vosburgh<redacted>
Reviewed-by: Eric Dumazet<redacted>

??? I do not recall giving my Reviewed-by:

Thanks your comment, I'll adjust it and send out V4 soon.

quoted

Reported-by: kernel test robot <redacted>
Signed-off-by: Sun Shouxin <redacted>
---
  drivers/net/bonding/bond_alb.c | 131 +++++++++++++++++++++++++++++++++
  1 file changed, 131 insertions(+)

diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c
index 533e476988f2..b14017364594 100644
--- a/drivers/net/bonding/bond_alb.c
+++ b/drivers/net/bonding/bond_alb.c

@@ -22,6 +22,7 @@

  #include <asm/byteorder.h>
  #include <net/bonding.h>
  #include <net/bond_alb.h>
+#include <net/ndisc.h>
    static const u8 mac_v6_allmcast[ETH_ALEN + 2] __long_aligned = {
      0x33, 0x33, 0x00, 0x00, 0x00, 0x01

@@ -1269,6 +1270,119 @@ static int alb_set_mac_address(struct bonding *bond, void *addr)

      return res;
  }
  +/*determine if the packet is NA or NS*/
+static bool alb_determine_nd(struct icmp6hdr *hdr)
+{
+    if (hdr->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT ||
+        hdr->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
+        return true;
+    }
+
+    return false;
+}
+
+static void alb_change_nd_option(struct sk_buff *skb, void *data)
+{
+    struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
+    struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
+    struct net_device *dev = skb->dev;
+    struct icmp6hdr *icmp6h = icmp6_hdr(skb);
+    struct ipv6hdr *ip6hdr = ipv6_hdr(skb);
+    u8 *lladdr = NULL;
+    u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
+                offsetof(struct nd_msg, opt));
+
+    while (ndoptlen) {
+        int l;
+
+        switch (nd_opt->nd_opt_type) {
+        case ND_OPT_SOURCE_LL_ADDR:
+        case ND_OPT_TARGET_LL_ADDR:
+        lladdr = ndisc_opt_addr_data(nd_opt, dev);
+        break;
+
+        default:
+        lladdr = NULL;
+        break;
+        }
+
+        l = nd_opt->nd_opt_len << 3;
+
+        if (ndoptlen < l || l == 0)
+            return;
+
+        if (lladdr) {
+            memcpy(lladdr, data, dev->addr_len);
+            icmp6h->icmp6_cksum = 0;
+
+            icmp6h->icmp6_cksum = csum_ipv6_magic(&ip6hdr->saddr,
+                                  &ip6hdr->daddr,
+                        ntohs(ip6hdr->payload_len),
+                        IPPROTO_ICMPV6,
+                        csum_partial(icmp6h,
+                                 ntohs(ip6hdr->payload_len), 0));
+        }
+        ndoptlen -= l;
+        nd_opt = ((void *)nd_opt) + l;
+    }
+}
+
+static u8 *alb_get_lladdr(struct sk_buff *skb)
+{
+    struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
+    struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
+    struct net_device *dev = skb->dev;
+    u8 *lladdr = NULL;
+    u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
+                offsetof(struct nd_msg, opt));
+
+    while (ndoptlen) {
+        int l;
+
+        switch (nd_opt->nd_opt_type) {
+        case ND_OPT_SOURCE_LL_ADDR:
+        case ND_OPT_TARGET_LL_ADDR:
+            lladdr = ndisc_opt_addr_data(nd_opt, dev);
+            break;
+
+        default:
+            break;
+        }
+
+        l = nd_opt->nd_opt_len << 3;
+
+        if (ndoptlen < l || l == 0)
+            return NULL;
+
+        if (lladdr)
+            return lladdr;
+
+        ndoptlen -= l;
+        nd_opt = ((void *)nd_opt) + l;
+    }
+
+    return lladdr;
+}
+
+static void alb_set_nd_option(struct sk_buff *skb, struct bonding *bond,
+                  struct slave *tx_slave)
+{
+    struct ipv6hdr *ip6hdr;
+    struct icmp6hdr *hdr = NULL;
+
+    if (skb->protocol == htons(ETH_P_IPV6)) {
+        if (tx_slave && tx_slave !=
+            rcu_access_pointer(bond->curr_active_slave)) {
+            ip6hdr = ipv6_hdr(skb);
+            if (ip6hdr->nexthdr == IPPROTO_ICMPV6) {
+                hdr = icmp6_hdr(skb);
+                if (alb_determine_nd(hdr))
+                    alb_change_nd_option(skb, tx_slave->dev->dev_addr);
+            }
+        }
+    }
+}
+
  /************************ exported alb functions ************************/
    int bond_alb_initialize(struct bonding *bond, int rlb_enabled)

@@ -1415,6 +1529,7 @@ struct slave *bond_xmit_alb_slave_get(struct bonding *bond,

      }
      case ETH_P_IPV6: {
          const struct ipv6hdr *ip6hdr;
+        struct icmp6hdr *hdr = NULL;
            /* IPv6 doesn't really use broadcast mac address, but leave
           * that here just in case.

@@ -1446,6 +1561,21 @@ struct slave *bond_xmit_alb_slave_get(struct bonding *bond,

              break;
          }
  +        if (ip6hdr->nexthdr == IPPROTO_ICMPV6) {
+            hdr = icmp6_hdr(skb);
+            if (alb_determine_nd(hdr)) {
+                u8 *lladdr = NULL;
+
+                lladdr = alb_get_lladdr(skb);
+                if (lladdr) {
+                    if (!bond_slave_has_mac_rx(bond, lladdr)) {
+                        do_tx_balance = false;
+                        break;
+                    }
+                }
+            }
+        }
+
          hash_start = (char *)&ip6hdr->daddr;
          hash_size = sizeof(ip6hdr->daddr);
          break;

@@ -1489,6 +1619,7 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev)

      struct slave *tx_slave = NULL;
        tx_slave = bond_xmit_alb_slave_get(bond, skb);
+    alb_set_nd_option(skb, bond, tx_slave);
      return bond_do_alb_xmit(skb, bond, tx_slave);
  }

base-commit: 6441998e2e37131b0a4c310af9156d79d3351c16

On 12/17/21 8:48 AM, Sun Shouxin wrote:

Since ipv6 neighbor solicitation and advertisement messages
isn't handled gracefully in bonding6 driver, we can see packet
drop due to inconsistency bewteen mac address in the option
message and source MAC .

Another examples is ipv6 neighbor solicitation and advertisement
messages from VM via tap attached to host brighe, the src mac
mighe be changed through balance-alb mode, but it is not synced
with Link-layer address in the option message.

The patch implements bond6's tx handle for ipv6 neighbor
solicitation and advertisement messages.

                         Border-Leaf
                         /        \
                        /          \
                     Tunnel1    Tunnel2
                      /              \
                     /                \
                   Leaf-1--Tunnel3--Leaf-2
                     \                /
                      \              /
                       \            /
                        \          /
                        NIC1    NIC2
                         \      /
                         server

We can see in our lab the Border-Leaf receives occasionally
a NA packet which is assigned to NIC1 mac in ND/NS option
message, but actaully send out via NIC2 mac due to tx-alb,
as a result, it will cause inconsistency between MAC table
and ND Table in Border-Leaf, i.e, NIC1 = Tunnel2 in ND table
and  NIC1 = Tunnel1 in mac table.

And then, Border-Leaf starts to forward packet destinated
to the Server, it will only check the ND table entry in some
switch to encapsulate the destination MAC of the message as
NIC1 MAC, and then send it out from Tunnel2 by ND table.
Then, Leaf-2 receives the packet, it notices the destination
MAC of message is NIC1 MAC and should forword it to Tunne1
by Tunnel3.

However, this traffic forward will be failure due to split
horizon of VxLAN tunnels.

Suggested-by: Hu Yadi <redacted>
Reviewed-by: Jay Vosburgh<redacted>
Reviewed-by: Eric Dumazet<redacted>

??? I do not recall giving my Reviewed-by:

Hi Sun,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on 6441998e2e37131b0a4c310af9156d79d3351c16]

url:    https://github.com/0day-ci/linux/commits/Sun-Shouxin/net-bonding-Add-support-for-IPV6-ns-na/20211218-005147
base:   6441998e2e37131b0a4c310af9156d79d3351c16
config: riscv-allyesconfig (https://download.01.org/0day-ci/archive/20211221/202112211539.hiXV7ML0-lkp@intel.com/config)
compiler: riscv64-linux-gcc (GCC) 11.2.0
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # https://github.com/0day-ci/linux/commit/ba063ac377ce5e5000b84a7f9bc127e4f72d0b00
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Sun-Shouxin/net-bonding-Add-support-for-IPV6-ns-na/20211218-005147
        git checkout ba063ac377ce5e5000b84a7f9bc127e4f72d0b00
        # save the config file to linux build tree
        mkdir build_dir
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-11.2.0 make.cross O=build_dir ARCH=riscv SHELL=/bin/bash

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <redacted>

All errors (new ones prefixed by >>):

   drivers/net/bonding/bond_alb.c: In function 'alb_change_nd_option':

quoted

drivers/net/bonding/bond_alb.c:1318:47: error: implicit declaration of function 'csum_ipv6_magic'; did you mean 'csum_tcpudp_magic'? [-Werror=implicit-function-declaration]

    1318 |                         icmp6h->icmp6_cksum = csum_ipv6_magic(&ip6hdr->saddr,
         |                                               ^~~~~~~~~~~~~~~
         |                                               csum_tcpudp_magic
   cc1: some warnings being treated as errors


vim +1318 drivers/net/bonding/bond_alb.c

  1283	
  1284	static void alb_change_nd_option(struct sk_buff *skb, void *data)
  1285	{
  1286		struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
  1287		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)msg->opt;
  1288		struct net_device *dev = skb->dev;
  1289		struct icmp6hdr *icmp6h = icmp6_hdr(skb);
  1290		struct ipv6hdr *ip6hdr = ipv6_hdr(skb);
  1291		u8 *lladdr = NULL;
  1292		u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
  1293					offsetof(struct nd_msg, opt));
  1294	
  1295		while (ndoptlen) {
  1296			int l;
  1297	
  1298			switch (nd_opt->nd_opt_type) {
  1299			case ND_OPT_SOURCE_LL_ADDR:
  1300			case ND_OPT_TARGET_LL_ADDR:
  1301			lladdr = ndisc_opt_addr_data(nd_opt, dev);
  1302			break;
  1303	
  1304			default:
  1305			lladdr = NULL;
  1306			break;
  1307			}
  1308	
  1309			l = nd_opt->nd_opt_len << 3;
  1310	
  1311			if (ndoptlen < l || l == 0)
  1312				return;
  1313	
  1314			if (lladdr) {
  1315				memcpy(lladdr, data, dev->addr_len);
  1316				icmp6h->icmp6_cksum = 0;
  1317	

1318				icmp6h->icmp6_cksum = csum_ipv6_magic(&ip6hdr->saddr,

  1319								      &ip6hdr->daddr,
  1320							ntohs(ip6hdr->payload_len),
  1321							IPPROTO_ICMPV6,
  1322							csum_partial(icmp6h,
  1323								     ntohs(ip6hdr->payload_len), 0));
  1324			}
  1325			ndoptlen -= l;
  1326			nd_opt = ((void *)nd_opt) + l;
  1327		}
  1328	}
  1329	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org