diff mbox series

[v2] igb: reinit_locked() should be called with rtnl_lock

Message ID 20200702223906.42B6C95C0494@us180.sjc.aristanetworks.com
State Awaiting Upstream
Delegated to: David Miller
Headers show
Series [v2] igb: reinit_locked() should be called with rtnl_lock | expand

Commit Message

Francesco Ruggeri July 2, 2020, 10:39 p.m. UTC
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 2f90b8657ec9 ("ixgbe: this patch
adds support for DCB to the kernel and ixgbe driver"),
commit 8f4c5c9fb87a ("ixgbe: reinit_locked() should be called with
rtnl_lock") and commit 88adce4ea8f9 ("ixgbe: fix possible race in
reset subtask").

v2: add fix for second race condition above.

Signed-off-by: Francesco Ruggeri <fruggeri@arista.com>

---
 drivers/net/ethernet/intel/igb/igb_main.c | 9 +++++++++
 1 file changed, 9 insertions(+)

Comments

Brown, Aaron F July 28, 2020, 8:37 p.m. UTC | #1
> From: Intel-wired-lan <intel-wired-lan-bounces@osuosl.org> On Behalf Of
> Francesco Ruggeri
> Sent: Thursday, July 2, 2020 3:39 PM
> To: linux-kernel@vger.kernel.org; netdev@vger.kernel.org; intel-wired-
> lan@lists.osuosl.org; kuba@kernel.org; davem@davemloft.net; Kirsher, Jeffrey
> T <jeffrey.t.kirsher@intel.com>; fruggeri@arista.com
> Subject: [Intel-wired-lan] [PATCH v2] 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 2f90b8657ec9 ("ixgbe: this patch
> adds support for DCB to the kernel and ixgbe driver"),
> commit 8f4c5c9fb87a ("ixgbe: reinit_locked() should be called with
> rtnl_lock") and commit 88adce4ea8f9 ("ixgbe: fix possible race in
> reset subtask").
> 
> v2: add fix for second race condition above.
> 
> Signed-off-by: Francesco Ruggeri <fruggeri@arista.com>
> 
> ---
>  drivers/net/ethernet/intel/igb/igb_main.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
diff mbox series

Patch

diff --git a/drivers/net/ethernet/intel/igb/igb_main.c b/drivers/net/ethernet/intel/igb/igb_main.c
index 8bb3db2cbd41..6e5861bfb0fa 100644
--- a/drivers/net/ethernet/intel/igb/igb_main.c
+++ b/drivers/net/ethernet/intel/igb/igb_main.c
@@ -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();
 }
 
 /**