Some time ago, I reported a calltrace issue "did not find a suitable aggregator", please see[1]. After a period of analysis and reproduction, I find that this problem is caused by concurrency.
Before the problem occurs, the bond structure is like follows:
bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1 \ port0 \ slaver1(eth1) - agg1.lag_ports -> NULL \ port1
If we run 'ifenslave bond0 -d eth1', the process is like below:
excuting __bond_release_one() | bond_upper_dev_unlink()[step1] | | | | | bond_3ad_lacpdu_recv() | | ->bond_3ad_rx_indication() | | spin_lock_bh() | | ->ad_rx_machine() | | ->__record_pdu()[step2] | | spin_unlock_bh() | | | | bond_3ad_state_machine_handler() | spin_lock_bh() | ->ad_port_selection_logic() | ->try to find free aggregator[step3] | ->try to find suitable aggregator[step4] | ->did not find a suitable aggregator[step5] | spin_unlock_bh() | | | | bond_3ad_unbind_slave() | spin_lock_bh() spin_unlock_bh()
step1: already removed slaver1(eth1) from list, but port1 remains step2: receive a lacpdu and update port0 step3: port0 will be removed from agg0.lag_ports. The struct is "agg0.lag_ports -> port1" now, and agg0 is not free. At the same time, slaver1/agg1 has been removed from the list by step1. So we can't find a free aggregator now. step4: can't find suitable aggregator because of step2 step5: cause a calltrace since port->aggregator is NULL
To solve this concurrency problem, the range of bond->mode_lock is extended from only bond_3ad_unbind_slave() to both bond_upper_dev_unlink() and bond_3ad_unbind_slave().
[1]https://lore.kernel.org/netdev/10374.1611947473@famine/
Signed-off-by: Yufeng Mo moyufeng@huawei.com Acked-by: Jay Vosburgh jay.vosburgh@canonical.com --- drivers/net/bonding/bond_3ad.c | 7 +------ drivers/net/bonding/bond_main.c | 6 +++++- 2 files changed, 6 insertions(+), 7 deletions(-)
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c index 6908822..f0f5adb 100644 --- a/drivers/net/bonding/bond_3ad.c +++ b/drivers/net/bonding/bond_3ad.c @@ -2099,15 +2099,13 @@ void bond_3ad_unbind_slave(struct slave *slave) struct list_head *iter; bool dummy_slave_update; /* Ignore this value as caller updates array */
- /* Sync against bond_3ad_state_machine_handler() */ - spin_lock_bh(&bond->mode_lock); aggregator = &(SLAVE_AD_INFO(slave)->aggregator); port = &(SLAVE_AD_INFO(slave)->port);
/* if slave is null, the whole port is not initialized */ if (!port->slave) { slave_warn(bond->dev, slave->dev, "Trying to unbind an uninitialized port\n"); - goto out; + return; }
slave_dbg(bond->dev, slave->dev, "Unbinding Link Aggregation Group %d\n", @@ -2239,9 +2237,6 @@ void bond_3ad_unbind_slave(struct slave *slave) } } port->slave = NULL; - -out: - spin_unlock_bh(&bond->mode_lock); }
/** diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index 0ff7567..deb019e 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -2129,14 +2129,18 @@ static int __bond_release_one(struct net_device *bond_dev, /* recompute stats just before removing the slave */ bond_get_stats(bond->dev, &bond->bond_stats);
- bond_upper_dev_unlink(bond, slave); /* unregister rx_handler early so bond_handle_frame wouldn't be called * for this slave anymore. */ netdev_rx_handler_unregister(slave_dev);
+ /* Sync against bond_3ad_state_machine_handler() */ + spin_lock_bh(&bond->mode_lock); + bond_upper_dev_unlink(bond, slave); + if (BOND_MODE(bond) == BOND_MODE_8023AD) bond_3ad_unbind_slave(slave); + spin_unlock_bh(&bond->mode_lock);
if (bond_mode_can_use_xmit_hash(bond)) bond_update_slave_arr(bond, slave);
On 28/07/2021 09:19, Yufeng Mo wrote:
Some time ago, I reported a calltrace issue "did not find a suitable aggregator", please see[1]. After a period of analysis and reproduction, I find that this problem is caused by concurrency.
Before the problem occurs, the bond structure is like follows:
bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1 \ port0 \ slaver1(eth1) - agg1.lag_ports -> NULL \ port1
If we run 'ifenslave bond0 -d eth1', the process is like below:
excuting __bond_release_one() | bond_upper_dev_unlink()[step1] | | | | | bond_3ad_lacpdu_recv() | | ->bond_3ad_rx_indication() | | spin_lock_bh() | | ->ad_rx_machine() | | ->__record_pdu()[step2] | | spin_unlock_bh() | | | | bond_3ad_state_machine_handler() | spin_lock_bh() | ->ad_port_selection_logic() | ->try to find free aggregator[step3] | ->try to find suitable aggregator[step4] | ->did not find a suitable aggregator[step5] | spin_unlock_bh() | | | | bond_3ad_unbind_slave() | spin_lock_bh() spin_unlock_bh()
step1: already removed slaver1(eth1) from list, but port1 remains step2: receive a lacpdu and update port0 step3: port0 will be removed from agg0.lag_ports. The struct is "agg0.lag_ports -> port1" now, and agg0 is not free. At the same time, slaver1/agg1 has been removed from the list by step1. So we can't find a free aggregator now. step4: can't find suitable aggregator because of step2 step5: cause a calltrace since port->aggregator is NULL
To solve this concurrency problem, the range of bond->mode_lock is extended from only bond_3ad_unbind_slave() to both bond_upper_dev_unlink() and bond_3ad_unbind_slave().
[1]https://lore.kernel.org/netdev/10374.1611947473@famine/
Signed-off-by: Yufeng Mo moyufeng@huawei.com Acked-by: Jay Vosburgh jay.vosburgh@canonical.com
drivers/net/bonding/bond_3ad.c | 7 +------ drivers/net/bonding/bond_main.c | 6 +++++- 2 files changed, 6 insertions(+), 7 deletions(-)
[snip]
/** diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index 0ff7567..deb019e 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -2129,14 +2129,18 @@ static int __bond_release_one(struct net_device *bond_dev, /* recompute stats just before removing the slave */ bond_get_stats(bond->dev, &bond->bond_stats);
- bond_upper_dev_unlink(bond, slave); /* unregister rx_handler early so bond_handle_frame wouldn't be called
*/ netdev_rx_handler_unregister(slave_dev);
- for this slave anymore.
- /* Sync against bond_3ad_state_machine_handler() */
- spin_lock_bh(&bond->mode_lock);
- bond_upper_dev_unlink(bond, slave);
this calls netdev_upper_dev_unlink() which calls call_netdevice_notifiers_info() for NETDEV_PRECHANGEUPPER and NETDEV_CHANGEUPPER, both of which are allowed to sleep so you cannot hold the mode lock
after netdev_rx_handler_unregister() the bond's recv_probe cannot be executed so you don't really need to unlink it under mode_lock or move mode_lock at all
if (BOND_MODE(bond) == BOND_MODE_8023AD) bond_3ad_unbind_slave(slave);
spin_unlock_bh(&bond->mode_lock);
if (bond_mode_can_use_xmit_hash(bond)) bond_update_slave_arr(bond, slave);
On 28/07/2021 10:34, Nikolay Aleksandrov wrote:
On 28/07/2021 09:19, Yufeng Mo wrote:
Some time ago, I reported a calltrace issue "did not find a suitable aggregator", please see[1]. After a period of analysis and reproduction, I find that this problem is caused by concurrency.
Before the problem occurs, the bond structure is like follows:
bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1 \ port0 \ slaver1(eth1) - agg1.lag_ports -> NULL \ port1
If we run 'ifenslave bond0 -d eth1', the process is like below:
excuting __bond_release_one() | bond_upper_dev_unlink()[step1] | | | | | bond_3ad_lacpdu_recv() | | ->bond_3ad_rx_indication() | | spin_lock_bh() | | ->ad_rx_machine() | | ->__record_pdu()[step2] | | spin_unlock_bh() | | | | bond_3ad_state_machine_handler() | spin_lock_bh() | ->ad_port_selection_logic() | ->try to find free aggregator[step3] | ->try to find suitable aggregator[step4] | ->did not find a suitable aggregator[step5] | spin_unlock_bh() | | | | bond_3ad_unbind_slave() | spin_lock_bh() spin_unlock_bh()
step1: already removed slaver1(eth1) from list, but port1 remains step2: receive a lacpdu and update port0 step3: port0 will be removed from agg0.lag_ports. The struct is "agg0.lag_ports -> port1" now, and agg0 is not free. At the same time, slaver1/agg1 has been removed from the list by step1. So we can't find a free aggregator now. step4: can't find suitable aggregator because of step2 step5: cause a calltrace since port->aggregator is NULL
To solve this concurrency problem, the range of bond->mode_lock is extended from only bond_3ad_unbind_slave() to both bond_upper_dev_unlink() and bond_3ad_unbind_slave().
[1]https://lore.kernel.org/netdev/10374.1611947473@famine/
Signed-off-by: Yufeng Mo moyufeng@huawei.com Acked-by: Jay Vosburgh jay.vosburgh@canonical.com
drivers/net/bonding/bond_3ad.c | 7 +------ drivers/net/bonding/bond_main.c | 6 +++++- 2 files changed, 6 insertions(+), 7 deletions(-)
[snip] after netdev_rx_handler_unregister() the bond's recv_probe cannot be executed so you don't really need to unlink it under mode_lock or move mode_lock at all
^^^^ Forget this part of the comment, I saw later that you don't want to receive lacpdu on the other port
The notifier sleep problem still exists though.
if (BOND_MODE(bond) == BOND_MODE_8023AD) bond_3ad_unbind_slave(slave);
spin_unlock_bh(&bond->mode_lock);
if (bond_mode_can_use_xmit_hash(bond)) bond_update_slave_arr(bond, slave);
Nikolay Aleksandrov nikolay@nvidia.com wrote:
On 28/07/2021 09:19, Yufeng Mo wrote:
Some time ago, I reported a calltrace issue "did not find a suitable aggregator", please see[1]. After a period of analysis and reproduction, I find that this problem is caused by concurrency.
Before the problem occurs, the bond structure is like follows:
bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1 \ port0 \ slaver1(eth1) - agg1.lag_ports -> NULL \ port1
If we run 'ifenslave bond0 -d eth1', the process is like below:
excuting __bond_release_one() | bond_upper_dev_unlink()[step1] | | | | | bond_3ad_lacpdu_recv() | | ->bond_3ad_rx_indication() | | spin_lock_bh() | | ->ad_rx_machine() | | ->__record_pdu()[step2] | | spin_unlock_bh() | | | | bond_3ad_state_machine_handler() | spin_lock_bh() | ->ad_port_selection_logic() | ->try to find free aggregator[step3] | ->try to find suitable aggregator[step4] | ->did not find a suitable aggregator[step5] | spin_unlock_bh() | | | | bond_3ad_unbind_slave() | spin_lock_bh() spin_unlock_bh()
step1: already removed slaver1(eth1) from list, but port1 remains step2: receive a lacpdu and update port0 step3: port0 will be removed from agg0.lag_ports. The struct is "agg0.lag_ports -> port1" now, and agg0 is not free. At the same time, slaver1/agg1 has been removed from the list by step1. So we can't find a free aggregator now. step4: can't find suitable aggregator because of step2 step5: cause a calltrace since port->aggregator is NULL
To solve this concurrency problem, the range of bond->mode_lock is extended from only bond_3ad_unbind_slave() to both bond_upper_dev_unlink() and bond_3ad_unbind_slave().
[1]https://lore.kernel.org/netdev/10374.1611947473@famine/
Signed-off-by: Yufeng Mo moyufeng@huawei.com Acked-by: Jay Vosburgh jay.vosburgh@canonical.com
drivers/net/bonding/bond_3ad.c | 7 +------ drivers/net/bonding/bond_main.c | 6 +++++- 2 files changed, 6 insertions(+), 7 deletions(-)
[snip]
/** diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index 0ff7567..deb019e 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -2129,14 +2129,18 @@ static int __bond_release_one(struct net_device *bond_dev, /* recompute stats just before removing the slave */ bond_get_stats(bond->dev, &bond->bond_stats);
- bond_upper_dev_unlink(bond, slave); /* unregister rx_handler early so bond_handle_frame wouldn't be called
*/ netdev_rx_handler_unregister(slave_dev);
- for this slave anymore.
- /* Sync against bond_3ad_state_machine_handler() */
- spin_lock_bh(&bond->mode_lock);
- bond_upper_dev_unlink(bond, slave);
this calls netdev_upper_dev_unlink() which calls call_netdevice_notifiers_info() for NETDEV_PRECHANGEUPPER and NETDEV_CHANGEUPPER, both of which are allowed to sleep so you cannot hold the mode lock
Indeed it does, I missed that the callbacks can sleep.
after netdev_rx_handler_unregister() the bond's recv_probe cannot be executed so you don't really need to unlink it under mode_lock or move mode_lock at all
I don't think moving the call to netdev_rx_handler_unregister is sufficient to close the race. If it's moved above the call to bond_upper_dev_unlink, the probe won't be called afterwards, but the LACPDU could have arrived just prior to the unregister and changed the port state in the bond_3ad_lacpdu_recv call sequence ("step 2", something in the LACPDU causes AD_PORT_SELECTED to be cleared). Later, bond_3ad_state_machine_handler runs in a separate work queue context, and could process the effect of the LACPDU after the rx_handler unregister, and still race with the upper_dev_unlink.
I suspect the solution is to rework ad_port_selection_logic to correctly handle the situation where no aggregator is available. Off the top of my head, I think something along the lines of:
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c index 6908822d9773..eb6223e4510e 100644 --- a/drivers/net/bonding/bond_3ad.c +++ b/drivers/net/bonding/bond_3ad.c @@ -1537,6 +1537,10 @@ static void ad_port_selection_logic(struct port *port, bool *update_slave_arr) slave_err(bond->dev, port->slave->dev, "Port %d did not find a suitable aggregator\n", port->actor_port_number); + aggregator = __get_first_agg(port); + ad_agg_selection_logic(aggregator, update_slave_arr); + + return; } } /* if all aggregator's ports are READY_N == TRUE, set ready=TRUE
I've not compiled or tested this, but the theory is that it will reselect a new aggregator for the bond (which happens anyway later in the function), then returns, leaving "port" as not AD_PORT_SELECTED. The next run of the state machine should attempt to select it again, and presumably succeed at that time.
This may leave the bond with no active ports for one interval between runs of the state machine, unfortunately, but it should eliminate the panic.
Another possibility might be netdev_rx_handler_unregister, then bond_3ad_unbind_slave, and finally bond_upper_dev_unlink, but I'm not sure right off if that would have other side effects.
Yufeng, would you be able to test the above and see if it resolves the issue in your test?
-J
if (BOND_MODE(bond) == BOND_MODE_8023AD) bond_3ad_unbind_slave(slave);
spin_unlock_bh(&bond->mode_lock);
if (bond_mode_can_use_xmit_hash(bond)) bond_update_slave_arr(bond, slave);
--- -Jay Vosburgh, jay.vosburgh@canonical.com
On 2021/7/29 3:05, Jay Vosburgh wrote:
Nikolay Aleksandrov nikolay@nvidia.com wrote:
On 28/07/2021 09:19, Yufeng Mo wrote:
Some time ago, I reported a calltrace issue "did not find a suitable aggregator", please see[1]. After a period of analysis and reproduction, I find that this problem is caused by concurrency.
Before the problem occurs, the bond structure is like follows:
bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1 \ port0 \ slaver1(eth1) - agg1.lag_ports -> NULL \ port1
If we run 'ifenslave bond0 -d eth1', the process is like below:
excuting __bond_release_one() | bond_upper_dev_unlink()[step1] | | | | | bond_3ad_lacpdu_recv() | | ->bond_3ad_rx_indication() | | spin_lock_bh() | | ->ad_rx_machine() | | ->__record_pdu()[step2] | | spin_unlock_bh() | | | | bond_3ad_state_machine_handler() | spin_lock_bh() | ->ad_port_selection_logic() | ->try to find free aggregator[step3] | ->try to find suitable aggregator[step4] | ->did not find a suitable aggregator[step5] | spin_unlock_bh() | | | | bond_3ad_unbind_slave() | spin_lock_bh() spin_unlock_bh()
step1: already removed slaver1(eth1) from list, but port1 remains step2: receive a lacpdu and update port0 step3: port0 will be removed from agg0.lag_ports. The struct is "agg0.lag_ports -> port1" now, and agg0 is not free. At the same time, slaver1/agg1 has been removed from the list by step1. So we can't find a free aggregator now. step4: can't find suitable aggregator because of step2 step5: cause a calltrace since port->aggregator is NULL
To solve this concurrency problem, the range of bond->mode_lock is extended from only bond_3ad_unbind_slave() to both bond_upper_dev_unlink() and bond_3ad_unbind_slave().
[1]https://lore.kernel.org/netdev/10374.1611947473@famine/
Signed-off-by: Yufeng Mo moyufeng@huawei.com Acked-by: Jay Vosburgh jay.vosburgh@canonical.com
drivers/net/bonding/bond_3ad.c | 7 +------ drivers/net/bonding/bond_main.c | 6 +++++- 2 files changed, 6 insertions(+), 7 deletions(-)
[snip]
/** diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index 0ff7567..deb019e 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -2129,14 +2129,18 @@ static int __bond_release_one(struct net_device *bond_dev, /* recompute stats just before removing the slave */ bond_get_stats(bond->dev, &bond->bond_stats);
- bond_upper_dev_unlink(bond, slave); /* unregister rx_handler early so bond_handle_frame wouldn't be called
*/ netdev_rx_handler_unregister(slave_dev);
- for this slave anymore.
- /* Sync against bond_3ad_state_machine_handler() */
- spin_lock_bh(&bond->mode_lock);
- bond_upper_dev_unlink(bond, slave);
this calls netdev_upper_dev_unlink() which calls call_netdevice_notifiers_info() for NETDEV_PRECHANGEUPPER and NETDEV_CHANGEUPPER, both of which are allowed to sleep so you cannot hold the mode lock
Indeed it does, I missed that the callbacks can sleep.
Yes, I missed that too.
after netdev_rx_handler_unregister() the bond's recv_probe cannot be executed so you don't really need to unlink it under mode_lock or move mode_lock at all
I don't think moving the call to netdev_rx_handler_unregister is sufficient to close the race. If it's moved above the call to bond_upper_dev_unlink, the probe won't be called afterwards, but the LACPDU could have arrived just prior to the unregister and changed the port state in the bond_3ad_lacpdu_recv call sequence ("step 2", something in the LACPDU causes AD_PORT_SELECTED to be cleared). Later, bond_3ad_state_machine_handler runs in a separate work queue context, and could process the effect of the LACPDU after the rx_handler unregister, and still race with the upper_dev_unlink.
I suspect the solution is to rework ad_port_selection_logic to correctly handle the situation where no aggregator is available. Off the top of my head, I think something along the lines of:
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c index 6908822d9773..eb6223e4510e 100644 --- a/drivers/net/bonding/bond_3ad.c +++ b/drivers/net/bonding/bond_3ad.c @@ -1537,6 +1537,10 @@ static void ad_port_selection_logic(struct port *port, bool *update_slave_arr) slave_err(bond->dev, port->slave->dev, "Port %d did not find a suitable aggregator\n", port->actor_port_number);
aggregator = __get_first_agg(port);
ad_agg_selection_logic(aggregator, update_slave_arr);
return;
} } /* if all aggregator's ports are READY_N == TRUE, set ready=TRUE
I've not compiled or tested this, but the theory is that it will
reselect a new aggregator for the bond (which happens anyway later in the function), then returns, leaving "port" as not AD_PORT_SELECTED. The next run of the state machine should attempt to select it again, and presumably succeed at that time.
This may leave the bond with no active ports for one interval between runs of the state machine, unfortunately, but it should eliminate the panic.
Another possibility might be netdev_rx_handler_unregister, then , and finally bond_upper_dev_unlink, but I'm not sure right off if that would have other side effects.
This may cause "%s: Warning: Found an uninitialized port\n" to be printed in bond_3ad_state_machine_handler(). But it doesn't matter.
In addition, I have analyzed the code in bond_3ad_unbind_slave(). Even if the slaver is not deleted from the list, the process is not affected. This seems to work. Anyway, I will test it.
Yufeng, would you be able to test the above and see if it resolves the issue in your test?
Sure,I will test both these two solution and report then.
Thanks Nikolay and Jay for the comments.
-J
if (BOND_MODE(bond) == BOND_MODE_8023AD) bond_3ad_unbind_slave(slave);
spin_unlock_bh(&bond->mode_lock);
if (bond_mode_can_use_xmit_hash(bond)) bond_update_slave_arr(bond, slave);
-Jay Vosburgh, jay.vosburgh@canonical.com .
On 2021/7/29 10:32, moyufeng wrote:
On 2021/7/29 3:05, Jay Vosburgh wrote:
Nikolay Aleksandrov nikolay@nvidia.com wrote:
On 28/07/2021 09:19, Yufeng Mo wrote:
Some time ago, I reported a calltrace issue "did not find a suitable aggregator", please see[1]. After a period of analysis and reproduction, I find that this problem is caused by concurrency.
Before the problem occurs, the bond structure is like follows:
bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1 \ port0 \ slaver1(eth1) - agg1.lag_ports -> NULL \ port1
If we run 'ifenslave bond0 -d eth1', the process is like below:
excuting __bond_release_one() | bond_upper_dev_unlink()[step1] | | | | | bond_3ad_lacpdu_recv() | | ->bond_3ad_rx_indication() | | spin_lock_bh() | | ->ad_rx_machine() | | ->__record_pdu()[step2] | | spin_unlock_bh() | | | | bond_3ad_state_machine_handler() | spin_lock_bh() | ->ad_port_selection_logic() | ->try to find free aggregator[step3] | ->try to find suitable aggregator[step4] | ->did not find a suitable aggregator[step5] | spin_unlock_bh() | | | | bond_3ad_unbind_slave() | spin_lock_bh() spin_unlock_bh()
step1: already removed slaver1(eth1) from list, but port1 remains step2: receive a lacpdu and update port0 step3: port0 will be removed from agg0.lag_ports. The struct is "agg0.lag_ports -> port1" now, and agg0 is not free. At the same time, slaver1/agg1 has been removed from the list by step1. So we can't find a free aggregator now. step4: can't find suitable aggregator because of step2 step5: cause a calltrace since port->aggregator is NULL
To solve this concurrency problem, the range of bond->mode_lock is extended from only bond_3ad_unbind_slave() to both bond_upper_dev_unlink() and bond_3ad_unbind_slave().
[1]https://lore.kernel.org/netdev/10374.1611947473@famine/
Signed-off-by: Yufeng Mo moyufeng@huawei.com Acked-by: Jay Vosburgh jay.vosburgh@canonical.com
drivers/net/bonding/bond_3ad.c | 7 +------ drivers/net/bonding/bond_main.c | 6 +++++- 2 files changed, 6 insertions(+), 7 deletions(-)
[snip]
/** diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index 0ff7567..deb019e 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -2129,14 +2129,18 @@ static int __bond_release_one(struct net_device *bond_dev, /* recompute stats just before removing the slave */ bond_get_stats(bond->dev, &bond->bond_stats);
- bond_upper_dev_unlink(bond, slave); /* unregister rx_handler early so bond_handle_frame wouldn't be called
*/ netdev_rx_handler_unregister(slave_dev);
- for this slave anymore.
- /* Sync against bond_3ad_state_machine_handler() */
- spin_lock_bh(&bond->mode_lock);
- bond_upper_dev_unlink(bond, slave);
this calls netdev_upper_dev_unlink() which calls call_netdevice_notifiers_info() for NETDEV_PRECHANGEUPPER and NETDEV_CHANGEUPPER, both of which are allowed to sleep so you cannot hold the mode lock
Indeed it does, I missed that the callbacks can sleep.
Yes, I missed that too.
after netdev_rx_handler_unregister() the bond's recv_probe cannot be executed so you don't really need to unlink it under mode_lock or move mode_lock at all
I don't think moving the call to netdev_rx_handler_unregister is sufficient to close the race. If it's moved above the call to bond_upper_dev_unlink, the probe won't be called afterwards, but the LACPDU could have arrived just prior to the unregister and changed the port state in the bond_3ad_lacpdu_recv call sequence ("step 2", something in the LACPDU causes AD_PORT_SELECTED to be cleared). Later, bond_3ad_state_machine_handler runs in a separate work queue context, and could process the effect of the LACPDU after the rx_handler unregister, and still race with the upper_dev_unlink.
I suspect the solution is to rework ad_port_selection_logic to correctly handle the situation where no aggregator is available. Off the top of my head, I think something along the lines of:
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c index 6908822d9773..eb6223e4510e 100644 --- a/drivers/net/bonding/bond_3ad.c +++ b/drivers/net/bonding/bond_3ad.c @@ -1537,6 +1537,10 @@ static void ad_port_selection_logic(struct port *port, bool *update_slave_arr) slave_err(bond->dev, port->slave->dev, "Port %d did not find a suitable aggregator\n", port->actor_port_number);
aggregator = __get_first_agg(port);
ad_agg_selection_logic(aggregator, update_slave_arr);
return;
} } /* if all aggregator's ports are READY_N == TRUE, set ready=TRUE
I've not compiled or tested this, but the theory is that it will
reselect a new aggregator for the bond (which happens anyway later in the function), then returns, leaving "port" as not AD_PORT_SELECTED. The next run of the state machine should attempt to select it again, and presumably succeed at that time.
This may leave the bond with no active ports for one interval between runs of the state machine, unfortunately, but it should eliminate the panic.
Another possibility might be netdev_rx_handler_unregister, then , and finally bond_upper_dev_unlink, but I'm not sure right off if that would have other side effects.
This may cause "%s: Warning: Found an uninitialized port\n" to be printed in bond_3ad_state_machine_handler(). But it doesn't matter.
In addition, I have analyzed the code in bond_3ad_unbind_slave(). Even if the slaver is not deleted from the list, the process is not affected. This seems to work. Anyway, I will test it.
Yufeng, would you be able to test the above and see if it resolves the issue in your test?
Sure,I will test both these two solution and report then.
Thanks Nikolay and Jay for the comments.
I have tested these two solution and got result below:
solution 1: handle the situation where no aggregator is available result: failed
I got a calltrace similar to the previous one. I think this is because port->aggregator is still NULL after the modification. The calltrace still occurs in the subsequent process.
log as below(bond0 with two slaver:eth0 and eth3):
$ ifenslave bond0 -d eth3 [87113.498148] bond0: (slave eth0): Port 1 did not find a suitable aggregator [87113.504996] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000030 [87113.513741] Mem abort info: [87113.516524] ESR = 0x96000004 [87113.519567] EC = 0x25: DABT (current EL), IL = 32 bits [87113.524856] SET = 0, FnV = 0 [87113.527898] EA = 0, S1PTW = 0 [87113.531026] Data abort info: [87113.533894] ISV = 0, ISS = 0x00000004 [87113.537713] CM = 0, WnR = 0 [87113.540667] user pgtable: 4k pages, 48-bit VAs, pgdp=00000020bfe17000 [87113.547078] [0000000000000030] pgd=0000000000000000, p4d=0000000000000000 [87113.553840] Internal error: Oops: 96000004 [#1] PREEMPT SMP [87113.559387] Modules linked in: bonding hclgevf hns3 hclge hnae3 [last unloaded: bonding] [87113.567445] CPU: 65 PID: 7 Comm: kworker/u256:0 Not tainted 5.13.0-rc4+ #1 [87113.574287] Hardware name: Huawei TaiShan 2280 V2/BC82AMDC, BIOS 2280-V2 CS V5.B110.01 01/07/2021 [87113.583116] Workqueue: bond0 bond_3ad_state_machine_handler [bonding] [87113.589540] pstate: 80400009 (Nzcv daif +PAN -UAO -TCO BTYPE=--) [87113.595518] pc : bond_3ad_state_machine_handler+0x5b0/0xe40 [bonding] [87113.601934] lr : bond_3ad_state_machine_handler+0x700/0xe40 [bonding] [87113.608348] sp : ffff800010533d10 [87113.611648] x29: ffff800010533d10 x28: ffff800010533d90 x27: ffff0020bfe2d638 [87113.618750] x26: ffff00400166e940 x25: ffff00400166ebf0 x24: ffffdf65e83a8524 [87113.625852] x23: ffff800010533d88 x22: ffff00400166e900 x21: ffff0020bfe2d600 [87113.632956] x20: 0000000000000000 x19: ffff00400166e900 x18: 0000000000000030 [87113.640059] x17: 0000000000000000 x16: ffffdf66343c1350 x15: ffff00208d685b68 [87113.647162] x14: ffffffffffffffff x13: ffff800090533927 x12: ffff80001053392f [87113.654264] x11: 0000000000000000 x10: ffff2047b7940000 x9 : ffffdf65e8395f9c [87113.661368] x8 : ffff2047b7680000 x7 : ffff2047b7940000 x6 : 0000000000000000 [87113.668470] x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000 [87113.675574] x2 : 000000000000003f x1 : 0000000000000004 x0 : 0000000000000003 [87113.682676] Call trace: [87113.685113] bond_3ad_state_machine_handler+0x5b0/0xe40 [bonding] [87113.691183] process_one_work+0x1dc/0x48c [87113.695176] worker_thread+0x15c/0x464 [87113.698908] kthread+0x168/0x16c [87113.702122] ret_from_fork+0x10/0x18 [87113.705685] Code: 7104009f 54001820 52800060 b9004f60 (79406064) [87113.711804] ---[ end trace 5bf403daf9e444eb ]--- [87113.721609] Kernel panic - not syncing: Oops: Fatal exception in interrupt [87113.728476] SMP: stopping secondary CPUs [87114.054358] Kernel Offset: 0x5f6624290000 from 0xffff800010000000 [87114.060423] PHYS_OFFSET: 0x0 [87114.063291] CPU features: 0x00000241,a3002c40 [87114.067628] Memory Limit: none [87114.075727] ---[ end Kernel panic - not syncing: Oops: Fatal exception in interrupt ]---
solution 2: put bond_upper_dev_unlink() after bond_3ad_unbind_slave() result: passed
The result is passed, except for a previously mentioned warning print. In normal cases, this warning is not printed.
log as below(bond0 with two slaver:eth0 and eth3):
$ ifenslave bond0 -d eth3 [86653.902168] bond0: Warning: Found an uninitialized port [86654.003515] bond0: (slave eth3): Releasing backup interface [86654.031183] hns3 0000:7d:00.3 eth3: net stop [86654.035823] hns3 0000:7d:00.3 eth3: link down
The solution 2 avoids the failure to find a suitable aggregator. So I think the solution 2 seems to solve the problem better.
-J
if (BOND_MODE(bond) == BOND_MODE_8023AD) bond_3ad_unbind_slave(slave);
spin_unlock_bh(&bond->mode_lock);
if (bond_mode_can_use_xmit_hash(bond)) bond_update_slave_arr(bond, slave);
-Jay Vosburgh, jay.vosburgh@canonical.com .
.