Blob Blame History Raw
From: Jacob Keller <jacob.e.keller@intel.com>
Date: Fri, 27 Oct 2017 11:06:52 -0400
Subject: i40evf: hold the critical task bit lock while opening
Patch-mainline: v4.16-rc1
Git-commit: 9b2aef128bff371c1b91b492a92cf42dad8c739b
References: bsc#1101816 FATE#325147 FATE#325149

If i40evf_open() is called quickly at the same time as a reset occurs
(such as via ethtool) it is possible for the device to attempt to open
while a reset is in progress. This occurs because the driver was not
holding the critical task bit lock during i40evf_open, nor was it
holding it around the call to i40evf_up_complete() in
i40evf_reset_task().

We didn't hold the lock previously because calls to i40evf_down() would
take the bit lock directly, and this would have caused a deadlock.

To avoid this, we'll move the bit lock handling out of i40evf_down() and
into the callers of this function. Additionally, we'll now hold the bit
lock over the entire set of steps when going up or down, to ensure that
we remain consistent.

Ultimately this causes us to serialize the transitions between down and
up properly, and avoid changing status while we're resetting.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 drivers/net/ethernet/intel/i40evf/i40evf_main.c |   40 ++++++++++++++++++------
 1 file changed, 31 insertions(+), 9 deletions(-)

--- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
@@ -1035,6 +1035,8 @@ static void i40evf_configure(struct i40e
 /**
  * i40evf_up_complete - Finish the last steps of bringing up a connection
  * @adapter: board private structure
+ *
+ * Expects to be called while holding the __I40EVF_IN_CRITICAL_TASK bit lock.
  **/
 static void i40evf_up_complete(struct i40evf_adapter *adapter)
 {
@@ -1052,6 +1054,8 @@ static void i40evf_up_complete(struct i4
 /**
  * i40e_down - Shutdown the connection processing
  * @adapter: board private structure
+ *
+ * Expects to be called while holding the __I40EVF_IN_CRITICAL_TASK bit lock.
  **/
 void i40evf_down(struct i40evf_adapter *adapter)
 {
@@ -1061,10 +1065,6 @@ void i40evf_down(struct i40evf_adapter *
 	if (adapter->state <= __I40EVF_DOWN_PENDING)
 		return;
 
-	while (test_and_set_bit(__I40EVF_IN_CRITICAL_TASK,
-				&adapter->crit_section))
-		usleep_range(500, 1000);
-
 	netif_carrier_off(netdev);
 	netif_tx_disable(netdev);
 	adapter->link_up = false;
@@ -1097,7 +1097,6 @@ void i40evf_down(struct i40evf_adapter *
 		adapter->aq_required |= I40EVF_FLAG_AQ_DISABLE_QUEUES;
 	}
 
-	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
 	mod_timer_pending(&adapter->watchdog_timer, jiffies + 1);
 }
 
@@ -1959,8 +1958,6 @@ continue_reset:
 
 	adapter->aq_required |= I40EVF_FLAG_AQ_ADD_MAC_FILTER;
 	adapter->aq_required |= I40EVF_FLAG_AQ_ADD_VLAN_FILTER;
-	clear_bit(__I40EVF_IN_CLIENT_TASK, &adapter->crit_section);
-	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
 	i40evf_misc_irq_enable(adapter);
 
 	mod_timer(&adapter->watchdog_timer, jiffies + 2);
@@ -1997,9 +1994,13 @@ continue_reset:
 		adapter->state = __I40EVF_DOWN;
 		wake_up(&adapter->down_waitqueue);
 	}
+	clear_bit(__I40EVF_IN_CLIENT_TASK, &adapter->crit_section);
+	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
 
 	return;
 reset_err:
+	clear_bit(__I40EVF_IN_CLIENT_TASK, &adapter->crit_section);
+	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
 	dev_err(&adapter->pdev->dev, "failed to allocate resources during reinit\n");
 	i40evf_close(netdev);
 }
@@ -2243,8 +2244,14 @@ static int i40evf_open(struct net_device
 		return -EIO;
 	}
 
-	if (adapter->state != __I40EVF_DOWN)
-		return -EBUSY;
+	while (test_and_set_bit(__I40EVF_IN_CRITICAL_TASK,
+				&adapter->crit_section))
+		usleep_range(500, 1000);
+
+	if (adapter->state != __I40EVF_DOWN) {
+		err = -EBUSY;
+		goto err_unlock;
+	}
 
 	/* allocate transmit descriptors */
 	err = i40evf_setup_all_tx_resources(adapter);
@@ -2268,6 +2275,8 @@ static int i40evf_open(struct net_device
 
 	i40evf_irq_enable(adapter, true);
 
+	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
+
 	return 0;
 
 err_req_irq:
@@ -2277,6 +2286,8 @@ err_setup_rx:
 	i40evf_free_all_rx_resources(adapter);
 err_setup_tx:
 	i40evf_free_all_tx_resources(adapter);
+err_unlock:
+	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
 
 	return err;
 }
@@ -2300,6 +2311,9 @@ static int i40evf_close(struct net_devic
 	if (adapter->state <= __I40EVF_DOWN_PENDING)
 		return 0;
 
+	while (test_and_set_bit(__I40EVF_IN_CRITICAL_TASK,
+				&adapter->crit_section))
+		usleep_range(500, 1000);
 
 	set_bit(__I40E_VSI_DOWN, adapter->vsi.state);
 	if (CLIENT_ENABLED(adapter))
@@ -2309,6 +2323,8 @@ static int i40evf_close(struct net_devic
 	adapter->state = __I40EVF_DOWN_PENDING;
 	i40evf_free_traffic_irqs(adapter);
 
+	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
+
 	/* We explicitly don't free resources here because the hardware is
 	 * still active and can DMA into memory. Resources are cleared in
 	 * i40evf_virtchnl_completion() after we get confirmation from the PF
@@ -2992,6 +3008,10 @@ static int i40evf_suspend(struct pci_dev
 
 	netif_device_detach(netdev);
 
+	while (test_and_set_bit(__I40EVF_IN_CRITICAL_TASK,
+				&adapter->crit_section))
+		usleep_range(500, 1000);
+
 	if (netif_running(netdev)) {
 		rtnl_lock();
 		i40evf_down(adapter);
@@ -3000,6 +3020,8 @@ static int i40evf_suspend(struct pci_dev
 	i40evf_free_misc_irq(adapter);
 	i40evf_reset_interrupt_capability(adapter);
 
+	clear_bit(__I40EVF_IN_CRITICAL_TASK, &adapter->crit_section);
+
 	retval = pci_save_state(pdev);
 	if (retval)
 		return retval;