Re: [PATCH] bonding: 3ad: update slave arr after initialize

[Jay Vosburgh]

jinyiting <jinyiting@xxxxxxxxxx> wrote:

>From: jin yiting <jinyiting@xxxxxxxxxx>
>
>The bond works in mode 4, and performs down/up operations on the bond
>that is normally negotiated. The probability of bond-> slave_arr is NULL
>
>Test commands:
>    ifconfig bond1 down
>    ifconfig bond1 up
>
>The conflict occurs in the following process：
>
>__dev_open (CPU A)
> --bond_open
>   --queue_delayed_work(bond->wq,&bond->ad_work,0);
>   --bond_update_slave_arr
>     --bond_3ad_get_active_agg_info
>
>ad_work(CPU B)
> --bond_3ad_state_machine_handler
>   --ad_agg_selection_logic
>
>ad_work runs on cpu B. In the function ad_agg_selection_logic, all
>agg->is_active will be cleared. Before the new active aggregator is
>selected on CPU B, bond_3ad_get_active_agg_info failed on CPU A,
>bond->slave_arr will be set to NULL. The best aggregator in
>ad_agg_selection_logic has not changed, no need to update slave arr.
>
>Signed-off-by: jin yiting <jinyiting@xxxxxxxxxx>
>---
> drivers/net/bonding/bond_3ad.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
>diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
>index 6908822..d100079 100644
>--- a/drivers/net/bonding/bond_3ad.c
>+++ b/drivers/net/bonding/bond_3ad.c
>@@ -2327,6 +2327,12 @@ void bond_3ad_state_machine_handler(struct work_struct *work)
> 
> 			aggregator = __get_first_agg(port);
> 			ad_agg_selection_logic(aggregator, &update_slave_arr);
>+			if (!update_slave_arr) {
>+				struct aggregator *active = __get_active_agg(aggregator);
>+
>+				if (active && active->is_active)
>+					update_slave_arr = true;
>+			}
> 		}
> 		bond_3ad_set_carrier(bond);
> 	}

	The described issue is a race condition (in that
ad_agg_selection_logic clears agg->is_active under mode_lock, but
bond_open -> bond_update_slave_arr is inspecting agg->is_active outside
the lock).  I don't see how the above change will reliably manage this;
the real issue looks to be that bond_update_slave_arr is committing
changes to the array (via bond_reset_slave_arr) based on a racy
inspection of the active aggregator state while it is in flux.

	Also, the description of the issue says "The best aggregator in
ad_agg_selection_logic has not changed, no need to update slave arr,"
but the change above does the opposite, and will set update_slave_arr
when the aggregator has not changed (update_slave_arr remains false at
return of ad_agg_selection_logic).

	I believe I understand the described problem, but I don't see
how the patch fixes it.  I suspect (but haven't tested) that the proper
fix is to acquire mode_lock in bond_update_slave_arr while calling
bond_3ad_get_active_agg_info to avoid conflict with the state machine.

	-J

---
	-Jay Vosburgh, jay.vosburgh@xxxxxxxxxxxxx