[NET-NEXT PATCH 1/3] net: Init NAPI dev_list on napi_del

[NET-NEXT PATCH 1/3] net: Init NAPI dev_list on napi_del

[Jeff Kirsher]

From: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>

The recent GRO patches introduced the NAPI removal of devices in
free_netdev.  For drivers that can change the number of queues during
driver operation, the NAPI infrastructure doesn't allow the freeing and
re-addition of NAPI entities without reloading the driver.

This change reinitializes the dev_list in each NAPI struct on delete,
instead of just deleting it (and assigning the list pointers to POISON).
Drivers that wish to remove/re-add NAPI will need to re-initialize the
netdev napi_list after removing all NAPI instances, before re-adding NAPI
devices again.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 net/core/dev.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index daca72e..536a8ac 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2548,7 +2548,7 @@ void netif_napi_del(struct napi_struct *napi)
 {
 	struct sk_buff *skb, *next;
 
-	list_del(&napi->dev_list);
+	list_del_init(&napi->dev_list);
 
 	for (skb = napi->gro_list; skb; skb = next) {
 		next = skb->next;

[NET-NEXT PATCH 2/3] ixgbe: Fix NAPI enable/disable path when using DCB

[Jeff Kirsher]

From: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>

This change allows DCB mode to change the number of queues, and presumably
the number of NAPI instances, safely.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/ixgbe/ixgbe_dcb_nl.c |    4 ++--
 drivers/net/ixgbe/ixgbe_main.c   |   11 ++++++++---
 2 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ixgbe/ixgbe_dcb_nl.c b/drivers/net/ixgbe/ixgbe_dcb_nl.c
index 8ac639d..4129976 100644
--- a/drivers/net/ixgbe/ixgbe_dcb_nl.c
+++ b/drivers/net/ixgbe/ixgbe_dcb_nl.c
@@ -148,6 +148,7 @@ static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 			netdev->netdev_ops->ndo_stop(netdev);
 		ixgbe_reset_interrupt_capability(adapter);
 		ixgbe_napi_del_all(adapter);
+		INIT_LIST_HEAD(&netdev->napi_list);
 		kfree(adapter->tx_ring);
 		kfree(adapter->rx_ring);
 		adapter->tx_ring = NULL;
@@ -157,7 +158,6 @@ static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 		adapter->flags &= ~IXGBE_FLAG_RSS_ENABLED;
 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
 		ixgbe_init_interrupt_scheme(adapter);
-		ixgbe_napi_add_all(adapter);
 		if (netif_running(netdev))
 			netdev->netdev_ops->ndo_open(netdev);
 	} else {
@@ -167,6 +167,7 @@ static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 				netdev->netdev_ops->ndo_stop(netdev);
 			ixgbe_reset_interrupt_capability(adapter);
 			ixgbe_napi_del_all(adapter);
+			INIT_LIST_HEAD(&netdev->napi_list);
 			kfree(adapter->tx_ring);
 			kfree(adapter->rx_ring);
 			adapter->tx_ring = NULL;
@@ -176,7 +177,6 @@ static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 			adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
 			adapter->flags |= IXGBE_FLAG_RSS_ENABLED;
 			ixgbe_init_interrupt_scheme(adapter);
-			ixgbe_napi_add_all(adapter);
 			if (netif_running(netdev))
 				netdev->netdev_ops->ndo_open(netdev);
 		}
diff --git a/drivers/net/ixgbe/ixgbe_main.c b/drivers/net/ixgbe/ixgbe_main.c
index b6ae9f6..acef3c6 100644
--- a/drivers/net/ixgbe/ixgbe_main.c
+++ b/drivers/net/ixgbe/ixgbe_main.c
@@ -2064,6 +2064,8 @@ static int ixgbe_up_complete(struct ixgbe_adapter *adapter)
 	else
 		ixgbe_configure_msi_and_legacy(adapter);
 
+	ixgbe_napi_add_all(adapter);
+
 	clear_bit(__IXGBE_DOWN, &adapter->state);
 	ixgbe_napi_enable_all(adapter);
 
@@ -3176,8 +3178,13 @@ static int ixgbe_close(struct net_device *netdev)
 void ixgbe_napi_add_all(struct ixgbe_adapter *adapter)
 {
 	int q_idx, q_vectors;
+	struct net_device *netdev = adapter->netdev;
 	int (*poll)(struct napi_struct *, int);
 
+	/* check if we already have our netdev->napi_list populated */
+	if (&netdev->napi_list != netdev->napi_list.next)
+		return;
+
 	if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
 		poll = &ixgbe_clean_rxonly;
 		/* Only enable as many vectors as we have rx queues. */
@@ -3271,6 +3278,7 @@ static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
 	}
 	ixgbe_reset_interrupt_capability(adapter);
 	ixgbe_napi_del_all(adapter);
+	INIT_LIST_HEAD(&netdev->napi_list);
 	kfree(adapter->tx_ring);
 	kfree(adapter->rx_ring);
 
@@ -4212,8 +4220,6 @@ static int __devinit ixgbe_probe(struct pci_dev *pdev,
 	netif_carrier_off(netdev);
 	netif_tx_stop_all_queues(netdev);
 
-	ixgbe_napi_add_all(adapter);
-
 	strcpy(netdev->name, "eth%d");
 	err = register_netdev(netdev);
 	if (err)
@@ -4299,7 +4305,6 @@ static void __devexit ixgbe_remove(struct pci_dev *pdev)
 	pci_release_regions(pdev);
 
 	DPRINTK(PROBE, INFO, "complete\n");
-	ixgbe_napi_del_all(adapter);
 	kfree(adapter->tx_ring);
 	kfree(adapter->rx_ring);
 

[NET-NEXT PATCH 3/3] ixgbe: Fix set_ringparam in ixgbe to use the same memory pools.

[Jeff Kirsher]

From: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>

The adapter rings are kcalloc()'d, but in set_ringparam() in ixgbe_ethtool,
we replace that memory from the vmalloc() pool.  This can result in a NULL
pointer reference when trying to modify the rings at a later time, or on
device removal.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/ixgbe/ixgbe_ethtool.c |   64 +++++++++++++++++--------------------
 1 files changed, 30 insertions(+), 34 deletions(-)

diff --git a/drivers/net/ixgbe/ixgbe_ethtool.c b/drivers/net/ixgbe/ixgbe_ethtool.c
index 849c1fe..67f87a7 100644
--- a/drivers/net/ixgbe/ixgbe_ethtool.c
+++ b/drivers/net/ixgbe/ixgbe_ethtool.c
@@ -712,30 +712,15 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 		return 0;
 	}
 
-	if (adapter->num_tx_queues > adapter->num_rx_queues)
-		temp_ring = vmalloc(adapter->num_tx_queues *
-		                    sizeof(struct ixgbe_ring));
-	else
-		temp_ring = vmalloc(adapter->num_rx_queues *
-		                    sizeof(struct ixgbe_ring));
+	temp_ring = kcalloc(adapter->num_tx_queues,
+	                    sizeof(struct ixgbe_ring), GFP_KERNEL);
 	if (!temp_ring)
 		return -ENOMEM;
 
 	while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
 		msleep(1);
 
-	if (netif_running(netdev))
-		ixgbe_down(adapter);
-
-	/*
-	 * We can't just free everything and then setup again,
-	 * because the ISRs in MSI-X mode get passed pointers
-	 * to the tx and rx ring structs.
-	 */
 	if (new_tx_count != adapter->tx_ring->count) {
-		memcpy(temp_ring, adapter->tx_ring,
-		       adapter->num_tx_queues * sizeof(struct ixgbe_ring));
-
 		for (i = 0; i < adapter->num_tx_queues; i++) {
 			temp_ring[i].count = new_tx_count;
 			err = ixgbe_setup_tx_resources(adapter, &temp_ring[i]);
@@ -747,21 +732,28 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 				}
 				goto err_setup;
 			}
+			temp_ring[i].v_idx = adapter->tx_ring[i].v_idx;
 		}
-
-		for (i = 0; i < adapter->num_tx_queues; i++)
-			ixgbe_free_tx_resources(adapter, &adapter->tx_ring[i]);
-
-		memcpy(adapter->tx_ring, temp_ring,
-		       adapter->num_tx_queues * sizeof(struct ixgbe_ring));
-
+		if (netif_running(netdev))
+			netdev->netdev_ops->ndo_stop(netdev);
+		ixgbe_reset_interrupt_capability(adapter);
+		ixgbe_napi_del_all(adapter);
+		INIT_LIST_HEAD(&netdev->napi_list);
+		kfree(adapter->tx_ring);
+		adapter->tx_ring = temp_ring;
+		temp_ring = NULL;
 		adapter->tx_ring_count = new_tx_count;
 	}
 
-	if (new_rx_count != adapter->rx_ring->count) {
-		memcpy(temp_ring, adapter->rx_ring,
-		       adapter->num_rx_queues * sizeof(struct ixgbe_ring));
+	temp_ring = kcalloc(adapter->num_rx_queues,
+	                    sizeof(struct ixgbe_ring), GFP_KERNEL);
+	if (!temp_ring) {
+		if (netif_running(netdev))
+			netdev->netdev_ops->ndo_open(netdev);
+		return -ENOMEM;
+	}
 
+	if (new_rx_count != adapter->rx_ring->count) {
 		for (i = 0; i < adapter->num_rx_queues; i++) {
 			temp_ring[i].count = new_rx_count;
 			err = ixgbe_setup_rx_resources(adapter, &temp_ring[i]);
@@ -773,13 +765,16 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 				}
 				goto err_setup;
 			}
+			temp_ring[i].v_idx = adapter->rx_ring[i].v_idx;
 		}
-
-		for (i = 0; i < adapter->num_rx_queues; i++)
-			ixgbe_free_rx_resources(adapter, &adapter->rx_ring[i]);
-
-		memcpy(adapter->rx_ring, temp_ring,
-		       adapter->num_rx_queues * sizeof(struct ixgbe_ring));
+		if (netif_running(netdev))
+			netdev->netdev_ops->ndo_stop(netdev);
+		ixgbe_reset_interrupt_capability(adapter);
+		ixgbe_napi_del_all(adapter);
+		INIT_LIST_HEAD(&netdev->napi_list);
+		kfree(adapter->rx_ring);
+		adapter->rx_ring = temp_ring;
+		temp_ring = NULL;
 
 		adapter->rx_ring_count = new_rx_count;
 	}
@@ -787,8 +782,9 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 	/* success! */
 	err = 0;
 err_setup:
+	ixgbe_init_interrupt_scheme(adapter);
 	if (netif_running(netdev))
-		ixgbe_up(adapter);
+		netdev->netdev_ops->ndo_open(netdev);
 
 	clear_bit(__IXGBE_RESETTING, &adapter->state);
 	return err;

Re: [NET-NEXT PATCH 1/3] net: Init NAPI dev_list on napi_del

[David Miller]

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Date: Thu, 25 Dec 2008 01:02:04 -0800

> The recent GRO patches introduced the NAPI removal of devices in
> free_netdev.  For drivers that can change the number of queues during
> driver operation, the NAPI infrastructure doesn't allow the freeing and
> re-addition of NAPI entities without reloading the driver.
> 
> This change reinitializes the dev_list in each NAPI struct on delete,
> instead of just deleting it (and assigning the list pointers to POISON).
> Drivers that wish to remove/re-add NAPI will need to re-initialize the
> netdev napi_list after removing all NAPI instances, before re-adding NAPI
> devices again.
> 
> Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Applied.

Re: [NET-NEXT PATCH 2/3] ixgbe: Fix NAPI enable/disable path when using DCB

[David Miller]

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Date: Thu, 25 Dec 2008 01:02:23 -0800

> This change allows DCB mode to change the number of queues, and presumably
> the number of NAPI instances, safely.
> 
> Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Applied.

Re: [NET-NEXT PATCH 3/3] ixgbe: Fix set_ringparam in ixgbe to use the same memory pools.

[David Miller]

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Date: Thu, 25 Dec 2008 01:02:41 -0800

> The adapter rings are kcalloc()'d, but in set_ringparam() in ixgbe_ethtool,
> we replace that memory from the vmalloc() pool.  This can result in a NULL
> pointer reference when trying to modify the rings at a later time, or on
> device removal.
> 
> Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Applied.