Hangbin Liu <[email protected]> wrote:
>Hi Jay,
>
>On Mon, Aug 25, 2025 at 04:07:55AM +0000, Hangbin Liu wrote:
>> On Fri, Aug 22, 2025 at 05:21:30PM -0700, Jay Vosburgh wrote:
>> > Hangbin Liu <[email protected]> wrote:
>> >
>> > >Commit 5c3bf6cba791 ("bonding: assign random address if device address is
>> > >same as bond") fixed an issue where, after releasing the first slave and
>> > >re-adding it to the bond with fail_over_mac=follow, both the active and
>> > >backup slaves could end up with duplicate MAC addresses. To avoid this,
>> > >the new slave was assigned a random address.
>> > >
>> > >However, if this happens when adding the very first slave, the bond’s
>> > >hardware address is set to match the slave’s. Later, during the
>> > >fail_over_mac=follow check, the slave’s MAC is randomized because it
>> > >naturally matches the bond, which is incorrect.
>> >
>> > The description here seems confusing to me; what does "this"
>> > refer to? I don't understand the sequence of events that lead to the
>> > issue being fixed here.
>> >
>> > I wonder if there's another bug somewhere, since nominally when
>> > releasing the last interface in the bond, __bond_release_one() should
>> > randomize the bond's MAC address, so it shouldn't match when adding (or
>> > re-adding ?) the first interface to the bond.
>> >
>>
>> Sorry I didn't make it clear. A easy reproducer would describe the issue.
>> e.g.
>> (omit the lo interface)
>>
>> [root@virtme-ng net]# ip link add type veth
>> [root@virtme-ng net]# ip link add bond0 type bond mode 1 miimon 100
>> fail_over_mac 2
>> [root@virtme-ng net]# ip link show
>> 3: veth0@veth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN
>> mode DEFAULT group default qlen 1000
>> link/ether 02:0a:04:c2:d6:21 brd ff:ff:ff:ff:ff:ff
>> 4: veth1@veth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN
>> mode DEFAULT group default qlen 1000
>> link/ether 82:a8:52:f4:81:4e brd ff:ff:ff:ff:ff:ff
>> 5: bond0: <BROADCAST,MULTICAST,MASTER> mtu 1500 qdisc noop state DOWN mode
>> DEFAULT group default qlen 1000
>> link/ether 92:5d:9c:47:e7:53 brd ff:ff:ff:ff:ff:ff
>> [root@virtme-ng net]# ip link set veth0 master bond0
>> [root@virtme-ng net]# ip link show
>> 3: veth0@veth1: <NO-CARRIER,BROADCAST,MULTICAST,SLAVE,UP,M-DOWN> mtu 1500
>> qdisc noqueue master bond0 state LOWERLAYERDOWN mode DEFAULT group default
>> qlen 1000
>> link/ether 4e:b5:4a:b4:03:18 brd ff:ff:ff:ff:ff:ff
>> 4: veth1@veth0: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode
>> DEFAULT group default qlen 1000
>> link/ether 82:a8:52:f4:81:4e brd ff:ff:ff:ff:ff:ff
>> 5: bond0: <BROADCAST,MULTICAST,MASTER> mtu 1500 qdisc noop state DOWN mode
>> DEFAULT group default qlen 1000
>> link/ether 02:0a:04:c2:d6:21 brd ff:ff:ff:ff:ff:ff
>>
>> Here we can see the veth0's mac address is randomized. The reason is in
>> function bond_enslave(), we set the bond mac address to the same as slave's
>> if it's the first one.
>>
>> /* If this is the first slave, then we need to set the master's
>> hardware
>> * address to be the same as the slave's.
>> */
>> if (!bond_has_slaves(bond) &&
>> bond->dev->addr_assign_type == NET_ADDR_RANDOM) {
>> res = bond_set_dev_addr(bond->dev, slave_dev);
>> if (res)
>> goto err_undo_flags;
>> }
>>
>> And later
>>
>> } else if (bond->params.fail_over_mac == BOND_FOM_FOLLOW &&
>> BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP &&
>> memcmp(slave_dev->dev_addr, bond_dev->dev_addr,
>> bond_dev->addr_len) == 0) {
>> /* Set slave to random address to avoid duplicate mac
>> * address in later fail over.
>> */
>> eth_random_addr(ss.__data);
>> } else {
>>
>> Here we check the bond and slave's mac address, which would be the same
>> definitely, which cause the first slave's mac got changed.
>
>Any comments for this?
Sorry, fell off the radar.
I follow what's going on now, and it's actually a lot simpler
than the description suggests, at least to my reading. Perhaps language
like:
After commit 5c3bf6cba791 ("bonding: assign random address if device
address is same as bond"), bonding will erroneously randomize the MAC
address of the first interface added to the bond if fail_over_mac =
follow.
Correct this by additionally testing for the bond being empty before
randomizing the MAC.
Does that sound ok to you?
-J
>Thanks
>Hangbin
>
>>
>> >
>> > >The issue is normally hidden since the first slave usually becomes the
>> > >active one, which restores the bond's MAC address. However, if another
>> > >slave is selected as the initial active interface, the issue becomes
>> > >visible.
>> > >
>> > >Fix this by assigning a random address only when slaves already exist in
>> > >the bond.
>> > >
>> > >Fixes: 5c3bf6cba791 ("bonding: assign random address if device address is
>> > >same as bond")
>> > >Reported-by: Qiuling Ren <[email protected]>
>> > >Signed-off-by: Hangbin Liu <[email protected]>
>> > >---
>> > > drivers/net/bonding/bond_main.c | 1 +
>> > > 1 file changed, 1 insertion(+)
>> > >
>> > >diff --git a/drivers/net/bonding/bond_main.c
>> > >b/drivers/net/bonding/bond_main.c
>> > >index 257333c88710..8832bc9f107b 100644
>> > >--- a/drivers/net/bonding/bond_main.c
>> > >+++ b/drivers/net/bonding/bond_main.c
>> > >@@ -2132,6 +2132,7 @@ int bond_enslave(struct net_device *bond_dev,
>> > >struct net_device *slave_dev,
>> > > memcpy(ss.__data, bond_dev->dev_addr, bond_dev->addr_len);
>> > > } else if (bond->params.fail_over_mac == BOND_FOM_FOLLOW &&
>> > > BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP &&
>> > >+ bond_has_slaves(bond) &&
>> > > memcmp(slave_dev->dev_addr, bond_dev->dev_addr,
>> > > bond_dev->addr_len) == 0) {
>> > > /* Set slave to random address to avoid duplicate mac
>> > > * address in later fail over.
>> > >--
>> > >2.50.1
>> > >
>> >
>> > ---
>> > -Jay Vosburgh, [email protected]
>> >
---
-Jay Vosburgh, [email protected]
