diff mbox

[PATCHv2,NEXT,1/1] netxen: add vlan accel support

Message ID 1312536015-21719-2-git-send-email-amit.salecha@qlogic.com
State Changes Requested, archived
Delegated to: David Miller
Headers show

Commit Message

amit salecha Aug. 5, 2011, 9:20 a.m. UTC 
From: Rajesh K Borundia <rajesh.borundia@qlogic.com>

o This increases performance on vlan interface.
o In case of fw reset, need to reprogram the ip addresses.
o Support LRO on vlan interface.

Signed-off-by: Rajesh K Borundia <rajesh.borundia@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
---
 drivers/net/netxen/netxen_nic.h      |    2 +
 drivers/net/netxen/netxen_nic_init.c |   22 ++++++++++++
 drivers/net/netxen/netxen_nic_main.c |   63 +++++++++++++++++++++++++++++----
 3 files changed, 79 insertions(+), 8 deletions(-)

Comments

David Miller Aug. 5, 2011, 9:42 a.m. UTC | #1 
From: <amit.salecha@qlogic.com>
Date: Fri, 5 Aug 2011 02:20:15 -0700

> From: Rajesh K Borundia <rajesh.borundia@qlogic.com>
> 
> o This increases performance on vlan interface.
> o In case of fw reset, need to reprogram the ip addresses.
> o Support LRO on vlan interface.
> 
> Signed-off-by: Rajesh K Borundia <rajesh.borundia@qlogic.com>
> Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>

As far as I can tell your card doesn't accelerate VLANs in hardware
at all, is this true?

I think it's very cumbersome and it doesn't make sense for every
driver in your situation to add this same sequence of software calls.

Some generic facility and interfaces should be created so that the
code doing this software VLAN acceleration is consolidated into
one place and any driver can follow a simple to use API in order
to make use of it.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
amit salecha Aug. 5, 2011, 10:12 a.m. UTC | #2 
> From: David Miller [mailto:davem@davemloft.net]
> Subject: Re: [PATCHv2 NEXT 1/1] netxen: add vlan accel support
> 
> From: <amit.salecha@qlogic.com>
> Date: Fri, 5 Aug 2011 02:20:15 -0700
> 
> > From: Rajesh K Borundia <rajesh.borundia@qlogic.com>
> >
> > o This increases performance on vlan interface.
> > o In case of fw reset, need to reprogram the ip addresses.
> > o Support LRO on vlan interface.
> >
> > Signed-off-by: Rajesh K Borundia <rajesh.borundia@qlogic.com>
> > Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
> 
> As far as I can tell your card doesn't accelerate VLANs in hardware
> at all, is this true?
> 
True.

> I think it's very cumbersome and it doesn't make sense for every
> driver in your situation to add this same sequence of software calls.
> 
> Some generic facility and interfaces should be created so that the
> code doing this software VLAN acceleration is consolidated into
> one place and any driver can follow a simple to use API in order
> to make use of it.

Basically netxen_nic LRO requirement is to program ip addresses in HW.
After fw recovery we don't get vlan devices by any kernel api, so we are adding NETIF_VLAN_ACC support.
Earlier kernel had vlan group support, through that we get vlan devices and then their ip addresses.

So please suggest from below option:
   1) We can add kernel api to get vlan devices belongs to base interface
   2) we can maintain linked list of ip addresses belongs to vlan device in driver.
   3) Have same patch and write kernel api for vlan acceleration code.

-Amit

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller Aug. 5, 2011, 10:26 a.m. UTC | #3 
From: Amit Salecha <amit.salecha@qlogic.com>
Date: Fri, 5 Aug 2011 05:12:37 -0500

> After fw recovery we don't get vlan devices by any kernel api, so we
> are adding NETIF_VLAN_ACC support.  Earlier kernel had vlan group
> support, through that we get vlan devices and then their ip
> addresses.

Every time an IP address is added or removed there is a notification
generated on the "inetaddr_chain", and you seem to be properly using
this.

You have the VLAN device expansion in there as well.

Why doesn't this work?

The IP addresses cannot be added or removed from the VLAN device
until it is attached to your device.  So you should see any IP
address manipulation that occurs for VLANs stacked on top of
your device.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
amit salecha Aug. 5, 2011, 11:04 a.m. UTC | #4 
> From: David Miller [mailto:davem@davemloft.net]
> Sent: Friday, August 05, 2011 3:56 PM
> 
> From: Amit Salecha <amit.salecha@qlogic.com>
> Date: Fri, 5 Aug 2011 05:12:37 -0500
> 
> > After fw recovery we don't get vlan devices by any kernel api, so we
> > are adding NETIF_VLAN_ACC support.  Earlier kernel had vlan group
> > support, through that we get vlan devices and then their ip
> > addresses.
> 
> Every time an IP address is added or removed there is a notification
> generated on the "inetaddr_chain", and you seem to be properly using
> this.
> 
> You have the VLAN device expansion in there as well.
> 
> Why doesn't this work?
> 
> The IP addresses cannot be added or removed from the VLAN device
> until it is attached to your device.  So you should see any IP
> address manipulation that occurs for VLANs stacked on top of
> your device.
>

This is true, but problem is after fw recovery. So flow is 
1) vlan device added
2) Driver program ip address of vlan device (through inet events)
3) Fw recovery
4) After fw recovery driver need to program ip addresses again which were already programmed.
   After fw recovery there won't be any inet/netdev event.

Simple solution is to maintain those ip addresses in driver.

It can be more easier, if we can get vlan ip addresses(or vlan device) by some kernel api.

-Amit
 


--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller Aug. 5, 2011, 11:27 a.m. UTC | #5 
From: Amit Salecha <amit.salecha@qlogic.com>
Date: Fri, 5 Aug 2011 06:04:10 -0500

> This is true, but problem is after fw recovery. So flow is 
> 1) vlan device added
> 2) Driver program ip address of vlan device (through inet events)
> 3) Fw recovery
> 4) After fw recovery driver need to program ip addresses again which were already programmed.
>    After fw recovery there won't be any inet/netdev event.
> 
> Simple solution is to maintain those ip addresses in driver.

When you program the device, in response to inet events, maintain a
software list in the driver that remembers the IP address state.

Then on FW recovery, walk the special list to reprogram the device.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
amit salecha Aug. 5, 2011, 12:13 p.m. UTC | #6 
> From: David Miller [mailto:davem@davemloft.net]
> Sent: Friday, August 05, 2011 4:58 PM
> 
> From: Amit Salecha <amit.salecha@qlogic.com>
> Date: Fri, 5 Aug 2011 06:04:10 -0500
> 
> > This is true, but problem is after fw recovery. So flow is
> > 1) vlan device added
> > 2) Driver program ip address of vlan device (through inet events)
> > 3) Fw recovery
> > 4) After fw recovery driver need to program ip addresses again which
> were already programmed.
> >    After fw recovery there won't be any inet/netdev event.
> >
> > Simple solution is to maintain those ip addresses in driver.
> 
> When you program the device, in response to inet events, maintain a
> software list in the driver that remembers the IP address state.
> 
> Then on FW recovery, walk the special list to reprogram the device.

Make sense. Will repost patch with above logic
-Thanks.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/net/netxen/netxen_nic.h b/drivers/net/netxen/netxen_nic.h
index f744d29..b37cf4e 100644
--- a/drivers/net/netxen/netxen_nic.h
+++ b/drivers/net/netxen/netxen_nic.h
@@ -44,6 +44,7 @@ 
 #include <linux/timer.h>
 
 #include <linux/vmalloc.h>
+#include <linux/if_vlan.h>
 
 #include <asm/io.h>
 #include <asm/byteorder.h>
@@ -1212,6 +1213,7 @@  struct netxen_adapter {
 
 	u8 mac_addr[ETH_ALEN];
 
+	unsigned long vlans[BITS_TO_LONGS(VLAN_N_VID)];
 	struct netxen_adapter_stats stats;
 
 	struct netxen_recv_context recv_ctx;
diff --git a/drivers/net/netxen/netxen_nic_init.c b/drivers/net/netxen/netxen_nic_init.c
index e8993a7..132d8fe 100644
--- a/drivers/net/netxen/netxen_nic_init.c
+++ b/drivers/net/netxen/netxen_nic_init.c
@@ -25,6 +25,7 @@ 
 
 #include <linux/netdevice.h>
 #include <linux/delay.h>
+#include <linux/if_vlan.h>
 #include <linux/slab.h>
 #include "netxen_nic.h"
 #include "netxen_nic_hw.h"
@@ -1558,7 +1559,9 @@  netxen_process_rcv(struct netxen_adapter *adapter,
 	struct netxen_rx_buffer *buffer;
 	struct sk_buff *skb;
 	struct nx_host_rds_ring *rds_ring;
+	struct ethhdr *eth_hdr;
 	int index, length, cksum, pkt_offset;
+	u16 vid = 0xffff;
 
 	if (unlikely(ring >= adapter->max_rds_rings))
 		return NULL;
@@ -1588,8 +1591,16 @@  netxen_process_rcv(struct netxen_adapter *adapter,
 	if (pkt_offset)
 		skb_pull(skb, pkt_offset);
 
+	if (!__vlan_get_tag(skb, &vid)) {
+		eth_hdr = (struct ethhdr *) skb->data;
+		memmove(skb->data + VLAN_HLEN, eth_hdr, ETH_ALEN * 2);
+		skb_pull(skb, VLAN_HLEN);
+	}
+
 	skb->protocol = eth_type_trans(skb, netdev);
 
+	if (vid != 0xffff)
+		__vlan_hwaccel_put_tag(skb, vid);
 	napi_gro_receive(&sds_ring->napi, skb);
 
 	adapter->stats.rx_pkts++;
@@ -1614,10 +1625,12 @@  netxen_process_lro(struct netxen_adapter *adapter,
 	struct nx_host_rds_ring *rds_ring;
 	struct iphdr *iph;
 	struct tcphdr *th;
+	struct ethhdr *eth_hdr;
 	bool push, timestamp;
 	int l2_hdr_offset, l4_hdr_offset;
 	int index;
 	u16 lro_length, length, data_offset;
+	u16 vid = 0xffff;
 	u32 seq_number;
 
 	if (unlikely(ring > adapter->max_rds_rings))
@@ -1650,6 +1663,13 @@  netxen_process_lro(struct netxen_adapter *adapter,
 	skb_put(skb, lro_length + data_offset);
 
 	skb_pull(skb, l2_hdr_offset);
+
+	if (!__vlan_get_tag(skb, &vid)) {
+		eth_hdr = (struct ethhdr *) skb->data;
+		memmove(skb->data + VLAN_HLEN, eth_hdr, ETH_ALEN * 2);
+		skb_pull(skb, VLAN_HLEN);
+	}
+
 	skb->protocol = eth_type_trans(skb, netdev);
 
 	iph = (struct iphdr *)skb->data;
@@ -1664,6 +1684,8 @@  netxen_process_lro(struct netxen_adapter *adapter,
 
 	length = skb->len;
 
+	if (vid != 0xffff)
+		__vlan_hwaccel_put_tag(skb, vid);
 	netif_receive_skb(skb);
 
 	adapter->stats.lro_pkts++;
diff --git a/drivers/net/netxen/netxen_nic_main.c b/drivers/net/netxen/netxen_nic_main.c
index f574edf..a6d6860 100644
--- a/drivers/net/netxen/netxen_nic_main.c
+++ b/drivers/net/netxen/netxen_nic_main.c
@@ -91,7 +91,11 @@  static irqreturn_t netxen_intr(int irq, void *data);
 static irqreturn_t netxen_msi_intr(int irq, void *data);
 static irqreturn_t netxen_msix_intr(int irq, void *data);
 
-static void netxen_config_indev_addr(struct net_device *dev, unsigned long);
+static void netxen_restore_indev_addr(struct net_device *dev, unsigned long);
+#ifdef CONFIG_INET
+static void netxen_config_indev_addr(struct netxen_adapter *,
+				struct net_device *, unsigned long);
+#endif
 static struct rtnl_link_stats64 *netxen_nic_get_stats(struct net_device *dev,
 						      struct rtnl_link_stats64 *stats);
 static int netxen_nic_set_mac(struct net_device *netdev, void *p);
@@ -517,6 +521,25 @@  static int netxen_set_features(struct net_device *dev, u32 features)
 	return 0;
 }
 
+static void
+netxen_vlan_rx_add(struct net_device *netdev, u16 vid)
+{
+	struct netxen_adapter *adapter = netdev_priv(netdev);
+	set_bit(vid, adapter->vlans);
+}
+
+static void
+netxen_vlan_rx_del(struct net_device *netdev, u16 vid)
+{
+	struct netxen_adapter *adapter = netdev_priv(netdev);
+	struct net_device *dev;
+
+	dev = __vlan_find_dev_deep(netdev, vid);
+	if (dev)
+		netxen_config_indev_addr(adapter, dev, NETDEV_DOWN);
+	clear_bit(vid, adapter->vlans);
+}
+
 static const struct net_device_ops netxen_netdev_ops = {
 	.ndo_open	   = netxen_nic_open,
 	.ndo_stop	   = netxen_nic_close,
@@ -529,6 +552,8 @@  static const struct net_device_ops netxen_netdev_ops = {
 	.ndo_tx_timeout	   = netxen_tx_timeout,
 	.ndo_fix_features = netxen_fix_features,
 	.ndo_set_features = netxen_set_features,
+	.ndo_vlan_rx_add_vid	= netxen_vlan_rx_add,
+	.ndo_vlan_rx_kill_vid	= netxen_vlan_rx_del,
 #ifdef CONFIG_NET_POLL_CONTROLLER
 	.ndo_poll_controller = netxen_nic_poll_controller,
 #endif
@@ -1259,6 +1284,9 @@  netxen_setup_netdev(struct netxen_adapter *adapter,
 	if (adapter->capabilities & NX_FW_CAPABILITY_FVLANTX)
 		netdev->hw_features |= NETIF_F_HW_VLAN_TX;
 
+	netdev->hw_features |= NETIF_F_HW_VLAN_RX;
+	netdev->hw_features |= NETIF_F_HW_VLAN_FILTER;
+
 	if (adapter->capabilities & NX_FW_CAPABILITY_HW_LRO)
 		netdev->hw_features |= NETIF_F_LRO;
 
@@ -1563,7 +1591,7 @@  static int netxen_nic_attach_func(struct pci_dev *pdev)
 		if (err)
 			goto err_out_detach;
 
-		netxen_config_indev_addr(netdev, NETDEV_UP);
+		netxen_restore_indev_addr(netdev, NETDEV_UP);
 	}
 
 	netif_device_attach(netdev);
@@ -2374,7 +2402,7 @@  netxen_attach_work(struct work_struct *work)
 			goto done;
 		}
 
-		netxen_config_indev_addr(netdev, NETDEV_UP);
+		netxen_restore_indev_addr(netdev, NETDEV_UP);
 	}
 
 	netif_device_attach(netdev);
@@ -2848,10 +2876,10 @@  netxen_destip_supported(struct netxen_adapter *adapter)
 }
 
 static void
-netxen_config_indev_addr(struct net_device *dev, unsigned long event)
+netxen_config_indev_addr(struct netxen_adapter *adapter,
+		struct net_device *dev, unsigned long event)
 {
 	struct in_device *indev;
-	struct netxen_adapter *adapter = netdev_priv(dev);
 
 	if (!netxen_destip_supported(adapter))
 		return;
@@ -2878,6 +2906,24 @@  netxen_config_indev_addr(struct net_device *dev, unsigned long event)
 	in_dev_put(indev);
 }
 
+static void
+netxen_restore_indev_addr(struct net_device *netdev, unsigned long event)
+
+{
+	struct netxen_adapter *adapter = netdev_priv(netdev);
+	struct net_device *dev;
+	u16 vid;
+
+	netxen_config_indev_addr(adapter, netdev, event);
+
+	for_each_set_bit(vid, adapter->vlans, VLAN_N_VID) {
+		dev = __vlan_find_dev_deep(netdev, vid);
+		if (!dev)
+			continue;
+		netxen_config_indev_addr(adapter, dev, event);
+	}
+}
+
 static int netxen_netdev_event(struct notifier_block *this,
 				 unsigned long event, void *ptr)
 {
@@ -2904,7 +2950,7 @@  recheck:
 	if (adapter->is_up != NETXEN_ADAPTER_UP_MAGIC)
 		goto done;
 
-	netxen_config_indev_addr(dev, event);
+	netxen_config_indev_addr(adapter, dev, event);
 done:
 	return NOTIFY_DONE;
 }
@@ -2921,7 +2967,7 @@  netxen_inetaddr_event(struct notifier_block *this,
 	dev = ifa->ifa_dev ? ifa->ifa_dev->dev : NULL;
 
 recheck:
-	if (dev == NULL || !netif_running(dev))
+	if (dev == NULL)
 		goto done;
 
 	if (dev->priv_flags & IFF_802_1Q_VLAN) {
@@ -2964,8 +3010,9 @@  static struct notifier_block netxen_inetaddr_cb = {
 };
 #else
 static void
-netxen_config_indev_addr(struct net_device *dev, unsigned long event)
+netxen_restore_indev_addr(struct net_device *dev, unsigned long event)
 { }
+static void
 #endif
 
 static struct pci_error_handlers netxen_err_handler = {