Commit 024a8168 authored by Francesco Ruggeri's avatar Francesco Ruggeri Committed by Tony Nguyen

igb: reinit_locked() should be called with rtnl_lock

We observed two panics involving races with igb_reset_task.
The first panic is caused by this race condition:

	kworker			reboot -f

	igb_reset_task
	igb_reinit_locked
	igb_down
	napi_synchronize
				__igb_shutdown
				igb_clear_interrupt_scheme
				igb_free_q_vectors
				igb_free_q_vector
				adapter->q_vector[v_idx] = NULL;
	napi_disable
	Panics trying to access
	adapter->q_vector[v_idx].napi_state

The second panic (a divide error) is caused by this race:

kworker		reboot -f	tx packet

igb_reset_task
		__igb_shutdown
		rtnl_lock()
		...
		igb_clear_interrupt_scheme
		igb_free_q_vectors
		adapter->num_tx_queues = 0
		...
		rtnl_unlock()
rtnl_lock()
igb_reinit_locked
igb_down
igb_up
netif_tx_start_all_queues
				dev_hard_start_xmit
				igb_xmit_frame
				igb_tx_queue_mapping
				Panics on
				r_idx % adapter->num_tx_queues

This commit applies to igb_reset_task the same changes that
were applied to ixgbe in commit 2f90b865 ("ixgbe: this patch
adds support for DCB to the kernel and ixgbe driver"),
commit 8f4c5c9f ("ixgbe: reinit_locked() should be called with
rtnl_lock") and commit 88adce4e ("ixgbe: fix possible race in
reset subtask").
Signed-off-by: default avatarFrancesco Ruggeri <fruggeri@arista.com>
Tested-by: default avatarAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: default avatarTony Nguyen <anthony.l.nguyen@intel.com>
parent 1050242f
......@@ -6224,9 +6224,18 @@ static void igb_reset_task(struct work_struct *work)
struct igb_adapter *adapter;
adapter = container_of(work, struct igb_adapter, reset_task);
rtnl_lock();
/* If we're already down or resetting, just bail */
if (test_bit(__IGB_DOWN, &adapter->state) ||
test_bit(__IGB_RESETTING, &adapter->state)) {
rtnl_unlock();
return;
}
igb_dump(adapter);
netdev_err(adapter->netdev, "Reset adapter\n");
igb_reinit_locked(adapter);
rtnl_unlock();
}
/**
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment