The mac address on backup slave should be convert from Solicited-Node Multicast address, not from bonding unicast target address.
v2: fix patch 01's subject
Hangbin Liu (2): bonding: fix incorrect MAC address setting to receive NS messages selftests: bonding: fix incorrect mac address
drivers/net/bonding/bond_options.c | 4 +++- tools/testing/selftests/drivers/net/bonding/bond_options.sh | 4 ++-- 2 files changed, 5 insertions(+), 3 deletions(-)
When validation on the backup slave is enabled, we need to validate the Neighbor Solicitation (NS) messages received on the backup slave. To receive these messages, the correct destination MAC address must be added to the slave. However, the target in bonding is a unicast address, which we cannot use directly. Instead, we should first convert it to a Solicited-Node Multicast Address and then derive the corresponding MAC address.
Fixes: 8eb36164d1a6 ("bonding: add ns target multicast address to slave device") Signed-off-by: Hangbin Liu liuhangbin@gmail.com --- drivers/net/bonding/bond_options.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/net/bonding/bond_options.c b/drivers/net/bonding/bond_options.c index 327b6ecdc77e..63cf209dcdc9 100644 --- a/drivers/net/bonding/bond_options.c +++ b/drivers/net/bonding/bond_options.c @@ -1246,6 +1246,7 @@ static void slave_set_ns_maddrs(struct bonding *bond, struct slave *slave, bool { struct in6_addr *targets = bond->params.ns_targets; char slot_maddr[MAX_ADDR_LEN]; + struct in6_addr mcaddr; int i;
if (!slave_can_set_ns_maddr(bond, slave)) @@ -1255,7 +1256,8 @@ static void slave_set_ns_maddrs(struct bonding *bond, struct slave *slave, bool if (ipv6_addr_any(&targets[i])) break;
- if (!ndisc_mc_map(&targets[i], slot_maddr, slave->dev, 0)) { + addrconf_addr_solict_mult(&targets[i], &mcaddr); + if (!ndisc_mc_map(&mcaddr, slot_maddr, slave->dev, 0)) { if (add) dev_mc_add(slave->dev, slot_maddr); else
Hangbin Liu liuhangbin@gmail.com wrote:
When validation on the backup slave is enabled, we need to validate the Neighbor Solicitation (NS) messages received on the backup slave. To receive these messages, the correct destination MAC address must be added to the slave. However, the target in bonding is a unicast address, which we cannot use directly. Instead, we should first convert it to a Solicited-Node Multicast Address and then derive the corresponding MAC address.
Fixes: 8eb36164d1a6 ("bonding: add ns target multicast address to slave device") Signed-off-by: Hangbin Liu liuhangbin@gmail.com
I think this now deserves some commentary in the code. Not because this function itself is unclear, but because there's the similarly-named slave_set_ns_maddr() (singular, not plural as in this patch) that will behave in a subtly different manner after this patch is applied.
The "maddrs" version here will convert the provided IPv6 address to the IPv6 solicited-nodes multicast address (RFC 4291 section 2.7.1) and thence to the MAC address, via ndisc_mc_map(), whereas the "maddr" version uses the supplied IPv6 address directly for multicast MAC address conversion.
Assuming that I'm following that correctly, I think this distinction deserves explanation. And if I'm getting it wrong, then it definitely deserves some explanation.
FWIW, functionally, I think it's doing the correct thing.
-J
drivers/net/bonding/bond_options.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/net/bonding/bond_options.c b/drivers/net/bonding/bond_options.c index 327b6ecdc77e..63cf209dcdc9 100644 --- a/drivers/net/bonding/bond_options.c +++ b/drivers/net/bonding/bond_options.c @@ -1246,6 +1246,7 @@ static void slave_set_ns_maddrs(struct bonding *bond, struct slave *slave, bool { struct in6_addr *targets = bond->params.ns_targets; char slot_maddr[MAX_ADDR_LEN];
struct in6_addr mcaddr; int i;
if (!slave_can_set_ns_maddr(bond, slave))
@@ -1255,7 +1256,8 @@ static void slave_set_ns_maddrs(struct bonding *bond, struct slave *slave, bool if (ipv6_addr_any(&targets[i])) break;
if (!ndisc_mc_map(&targets[i], slot_maddr, slave->dev, 0)) {
addrconf_addr_solict_mult(&targets[i], &mcaddr);if (!ndisc_mc_map(&mcaddr, slot_maddr, slave->dev, 0)) { if (add) dev_mc_add(slave->dev, slot_maddr); else-- 2.46.0
--- -Jay Vosburgh, jv@jvosburgh.net
On Thu, Feb 06, 2025 at 05:19:34PM -0800, Jay Vosburgh wrote:
I think this now deserves some commentary in the code. Not because this function itself is unclear, but because there's the similarly-named slave_set_ns_maddr() (singular, not plural as in this patch) that will behave in a subtly different manner after this patch is applied.
Yes, slave_set_ns_maddr() also need to be changed. The names are similar. I will add some comments for the functions.
Thanks Hangbin
The correct mac address for NS target 2001:db8::254 is 33:33:ff:00:02:54, not 33:33:00:00:02:54. The same with client maddress.
Fixes: 86fb6173d11e ("selftests: bonding: add ns multicast group testing") Signed-off-by: Hangbin Liu liuhangbin@gmail.com --- tools/testing/selftests/drivers/net/bonding/bond_options.sh | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tools/testing/selftests/drivers/net/bonding/bond_options.sh b/tools/testing/selftests/drivers/net/bonding/bond_options.sh index edc56e2cc606..7bc148889ca7 100755 --- a/tools/testing/selftests/drivers/net/bonding/bond_options.sh +++ b/tools/testing/selftests/drivers/net/bonding/bond_options.sh @@ -11,8 +11,8 @@ ALL_TESTS="
lib_dir=$(dirname "$0") source ${lib_dir}/bond_topo_3d1c.sh -c_maddr="33:33:00:00:00:10" -g_maddr="33:33:00:00:02:54" +c_maddr="33:33:ff:00:00:10" +g_maddr="33:33:ff:00:02:54"
skip_prio() {
linux-kselftest-mirror@lists.linaro.org
-
Hangbin Liu -
Jay Vosburgh