Skip to contentSkip to conversation

Introduce n-tuple ethtool support

PATCH In review Jeff Kirsher 5 participants 10 messages Feb 11, 2010 – Feb 15, 2010

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v5 cover Introduce n-tuple ethtool support

One more round of fixes, based on feedback from Patrick McHardy

1) Change the list count to an unsigned value
2) Fix a memory leak
3) Removed an unnecessary list traversal in the ethtool core
4) Moved all list destruction to a helper function, allowing the driver to control when it clears the list (aside from when free_netdev() kills the cached list).

---

Peter Waskiewicz (3):
ixgbe: Bump driver version up
ixgbe: Add support for the new ethtool n-tuple programming interface
ethtool: Introduce n-tuple filter programming support

 drivers/net/ixgbe/ixgbe.h         |    4 
 drivers/net/ixgbe/ixgbe_82599.c   |  106 ++++++++++--
 drivers/net/ixgbe/ixgbe_ethtool.c |  111 ++++++++++++
 drivers/net/ixgbe/ixgbe_main.c    |   18 ++
 drivers/net/ixgbe/ixgbe_type.h    |    9 +
 include/linux/ethtool.h           |   50 ++++++
 include/linux/netdevice.h         |    3 
 net/core/dev.c                    |    6 +
 net/core/ethtool.c                |  332 +++++++++++++++++++++++++++++++++++++
 9 files changed, 619 insertions(+), 20 deletions(-)
Signature · 2 lines
Cheers,
Jeff
#1 lore raw

PATCH v5 1/3 ethtool: Introduce n-tuple filter programming support

From: Peter Waskiewicz <[email protected]>

This patchset enables the ethtool layer to program n-tuple filters to an underlying device. The idea is to allow capable hardware to have static rules applied that can assist steering flows into appropriate queues.

Hardware that is known to support these types of filters today are ixgbe and niu.

---

 include/linux/ethtool.h   |   50 +++++++
 include/linux/netdevice.h |    3 
 net/core/dev.c            |    6 +
 net/core/ethtool.c        |  332 +++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 390 insertions(+), 1 deletions(-)
include/linux/ethtool.h +50 -0
index ef4a2d8..a3cac53 100644
@@ -14,6 +14,7 @@
 #define _LINUX_ETHTOOL_H
 
 #include <linux/types.h>
+#include <linux/rculist.h>
 
 /* This should work for both 32 and 64 bit userland. */
 struct ethtool_cmd {
@@ -242,6 +243,7 @@ enum ethtool_stringset {
 	ETH_SS_TEST		= 0,
 	ETH_SS_STATS,
 	ETH_SS_PRIV_FLAGS,
+	ETH_SS_NTUPLE_FILTERS,
 };
 
 /* for passing string sets for data tagging */
@@ -290,6 +292,7 @@ struct ethtool_perm_addr {
  */
 enum ethtool_flags {
 	ETH_FLAG_LRO		= (1 << 15),	/* LRO is enabled */
+	ETH_FLAG_NTUPLE		= (1 << 27),	/* N-tuple filters enabled */
 };
 
 /* The following structures are for supporting RX network flow
@@ -363,6 +366,35 @@ struct ethtool_rxnfc {
 	__u32				rule_locs[0];
 };
 
+struct ethtool_rx_ntuple_flow_spec {
+	__u32		 flow_type;
+	union {
+		struct ethtool_tcpip4_spec		tcp_ip4_spec;
+		struct ethtool_tcpip4_spec		udp_ip4_spec;
+		struct ethtool_tcpip4_spec		sctp_ip4_spec;
+		struct ethtool_ah_espip4_spec		ah_ip4_spec;
+		struct ethtool_ah_espip4_spec		esp_ip4_spec;
+		struct ethtool_rawip4_spec		raw_ip4_spec;
+		struct ethtool_ether_spec		ether_spec;
+		struct ethtool_usrip4_spec		usr_ip4_spec;
+		__u8					hdata[64];
+	} h_u, m_u; /* entry, mask */
+
+	__u16	        vlan_tag;
+	__u16	        vlan_tag_mask;
+	__u64		data;      /* user-defined flow spec data */
+	__u64		data_mask; /* user-defined flow spec mask */
+
+	/* signed to distinguish between queue and actions (DROP) */
+	__s32		action;
+#define ETHTOOL_RXNTUPLE_ACTION_DROP -1
+};
+
+struct ethtool_rx_ntuple {
+	__u32					cmd;
+	struct ethtool_rx_ntuple_flow_spec	fs;
+};
+
 #define ETHTOOL_FLASH_MAX_FILENAME	128
 enum ethtool_flash_op_type {
 	ETHTOOL_FLASH_ALL_REGIONS	= 0,
@@ -377,6 +409,18 @@ struct ethtool_flash {
 
 #ifdef __KERNEL__
 
+struct ethtool_rx_ntuple_flow_spec_container {
+	struct ethtool_rx_ntuple_flow_spec fs;
+	struct list_head list;
+};
+
+struct ethtool_rx_ntuple_list {
+#define ETHTOOL_MAX_NTUPLE_LIST_ENTRY 1024
+#define ETHTOOL_MAX_NTUPLE_STRING_PER_ENTRY 14
+	struct list_head	list;
+	unsigned int		count;
+};
+
 struct net_device;
 
 /* Some generic methods drivers may use in their ethtool_ops */
@@ -394,6 +438,7 @@ u32 ethtool_op_get_ufo(struct net_device *dev);
 int ethtool_op_set_ufo(struct net_device *dev, u32 data);
 u32 ethtool_op_get_flags(struct net_device *dev);
 int ethtool_op_set_flags(struct net_device *dev, u32 data);
+void ethtool_ntuple_flush(struct net_device *dev);
 
 /**
  * &ethtool_ops - Alter and report network device settings
@@ -500,6 +545,8 @@ struct ethtool_ops {
 	int	(*set_rxnfc)(struct net_device *, struct ethtool_rxnfc *);
 	int     (*flash_device)(struct net_device *, struct ethtool_flash *);
 	int	(*reset)(struct net_device *, u32 *);
+	int	(*set_rx_ntuple)(struct net_device *, struct ethtool_rx_ntuple *);
+	int	(*get_rx_ntuple)(struct net_device *, u32 stringset, void *);
 };
 #endif /* __KERNEL__ */
 
@@ -559,6 +606,9 @@ struct ethtool_ops {
 #define	ETHTOOL_FLASHDEV	0x00000033 /* Flash firmware to device */
 #define	ETHTOOL_RESET		0x00000034 /* Reset hardware */
 
+#define ETHTOOL_SRXNTUPLE	0x00000035 /* Add an n-tuple filter to device */
+#define ETHTOOL_GRXNTUPLE	0x00000036 /* Get n-tuple filters from device */
+
 /* compatibility with older code */
 #define SPARC_ETH_GSET		ETHTOOL_GSET
 #define SPARC_ETH_SSET		ETHTOOL_SSET
include/linux/netdevice.h +3 -0
index e535700..cdf53a8 100644
@@ -746,6 +746,7 @@ struct net_device {
 #define NETIF_F_FCOE_CRC	(1 << 24) /* FCoE CRC32 */
 #define NETIF_F_SCTP_CSUM	(1 << 25) /* SCTP checksum offload */
 #define NETIF_F_FCOE_MTU	(1 << 26) /* Supports max FCoE MTU, 2158 bytes*/
+#define NETIF_F_NTUPLE		(1 << 27) /* N-tuple filters supported */
 
 	/* Segmentation offload features */
 #define NETIF_F_GSO_SHIFT	16
@@ -954,6 +955,8 @@ struct net_device {
 	/* max exchange id for FCoE LRO by ddp */
 	unsigned int		fcoe_ddp_xid;
 #endif
+	/* n-tuple filter list attached to this device */
+	struct ethtool_rx_ntuple_list ethtool_ntuple_list;
 };
 #define to_net_dev(d) container_of(d, struct net_device, dev)
 
net/core/dev.c +6 -0
index 94c1eee..8b3ad65 100644
@@ -5419,6 +5419,8 @@ struct net_device *alloc_netdev_mq(int sizeof_priv, const char *name,
 
 	netdev_init_queues(dev);
 
+	INIT_LIST_HEAD(&dev->ethtool_ntuple_list.list);
+	dev->ethtool_ntuple_list.count = 0;
 	INIT_LIST_HEAD(&dev->napi_list);
 	INIT_LIST_HEAD(&dev->unreg_list);
 	INIT_LIST_HEAD(&dev->link_watch_list);
@@ -5447,6 +5449,7 @@ EXPORT_SYMBOL(alloc_netdev_mq);
 void free_netdev(struct net_device *dev)
 {
 	struct napi_struct *p, *n;
+	struct ethtool_rx_ntuple_flow_spec_container *fsc, *f;
 
 	release_net(dev_net(dev));
 
@@ -5455,6 +5458,9 @@ void free_netdev(struct net_device *dev)
 	/* Flush device addresses */
 	dev_addr_flush(dev);
 
+	/* Clear ethtool n-tuple list */
+	ethtool_ntuple_flush(dev);
+
 	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
 		netif_napi_del(p);
 
net/core/ethtool.c +331 -1

377 lines not on this page · raw message

#2 lore raw

PATCH v5 2/3 ixgbe: Add support for the new ethtool n-tuple programming interface

In reply to Jeff Kirsher

From: Peter Waskiewicz <[email protected]>

This patch adds n-tuple filter programming to 82599.

---

 drivers/net/ixgbe/ixgbe.h         |    4 +
 drivers/net/ixgbe/ixgbe_82599.c   |  106 +++++++++++++++++++++++++++++++----
 drivers/net/ixgbe/ixgbe_ethtool.c |  111 ++++++++++++++++++++++++++++++++++++-
 drivers/net/ixgbe/ixgbe_main.c    |   16 +++++
 drivers/net/ixgbe/ixgbe_type.h    |    9 +++
 5 files changed, 228 insertions(+), 18 deletions(-)
drivers/net/ixgbe/ixgbe.h +4 -0
index bffbe0d..19e94ee 100644
@@ -453,6 +453,10 @@ extern s32 ixgbe_init_fdir_perfect_82599(struct ixgbe_hw *hw, u32 pballoc);
 extern s32 ixgbe_fdir_add_signature_filter_82599(struct ixgbe_hw *hw,
                                                  struct ixgbe_atr_input *input,
                                                  u8 queue);
+extern s32 ixgbe_fdir_add_perfect_filter_82599(struct ixgbe_hw *hw,
+                                      struct ixgbe_atr_input *input,
+                                      struct ixgbe_atr_input_masks *input_masks,
+                                      u16 soft_id, u8 queue);
 extern s32 ixgbe_atr_set_vlan_id_82599(struct ixgbe_atr_input *input,
                                        u16 vlan_id);
 extern s32 ixgbe_atr_set_src_ipv4_82599(struct ixgbe_atr_input *input,
drivers/net/ixgbe/ixgbe_82599.c +93 -13

197 lines not on this page · raw message

drivers/net/ixgbe/ixgbe_ethtool.c +108 -3

143 lines not on this page · raw message

drivers/net/ixgbe/ixgbe_main.c +14 -2
index 2a3c831..f04b539 100644
@@ -3253,6 +3253,9 @@ void ixgbe_down(struct ixgbe_adapter *adapter)
 
 	netif_carrier_off(netdev);
 
+	/* clear n-tuple filters that are cached */
+	ethtool_ntuple_flush(netdev);
+
 	if (!pci_channel_offline(adapter->pdev))
 		ixgbe_reset(adapter);
 	ixgbe_clean_all_tx_rings(adapter);
@@ -4187,6 +4190,7 @@ static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
 	struct pci_dev *pdev = adapter->pdev;
+	struct net_device *dev = adapter->netdev;
 	unsigned int rss;
 #ifdef CONFIG_IXGBE_DCB
 	int j;
@@ -4214,10 +4218,18 @@ static int __devinit ixgbe_sw_init(struct ixgbe_adapter *adapter)
 		adapter->max_msix_q_vectors = MAX_MSIX_Q_VECTORS_82599;
 		adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
 		adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
-		adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
+		if (dev->features & NETIF_F_NTUPLE) {
+			/* Flow Director perfect filter enabled */
+			adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
+			adapter->atr_sample_rate = 0;
+			spin_lock_init(&adapter->fdir_perfect_lock);
+		} else {
+			/* Flow Director hash filters enabled */
+			adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
+			adapter->atr_sample_rate = 20;
+		}
 		adapter->ring_feature[RING_F_FDIR].indices =
 		                                         IXGBE_MAX_FDIR_INDICES;
-		adapter->atr_sample_rate = 20;
 		adapter->fdir_pballoc = 0;
 #ifdef IXGBE_FCOE
 		adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
drivers/net/ixgbe/ixgbe_type.h +9 -0
index 0db67c1..2be9074 100644
@@ -2129,6 +2129,15 @@ struct ixgbe_atr_input {
 	u8 byte_stream[42];
 };
 
+struct ixgbe_atr_input_masks {
+	u32 src_ip_mask;
+	u32 dst_ip_mask;
+	u16 src_port_mask;
+	u16 dst_port_mask;
+	u16 vlan_id_mask;
+	u16 data_mask;
+};
+
 enum ixgbe_eeprom_type {
 	ixgbe_eeprom_uninitialized = 0,
 	ixgbe_eeprom_spi,
#3 lore raw

PATCH v5 3/3 ixgbe: Bump driver version up

In reply to Jeff Kirsher

From: Peter Waskiewicz <[email protected]>

Driver has gone under significant changes, the version should reflect that.

---

 drivers/net/ixgbe/ixgbe_main.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/net/ixgbe/ixgbe_main.c +1 -1
index f04b539..382bf89 100644
@@ -51,7 +51,7 @@ char ixgbe_driver_name[] = "ixgbe";
 static const char ixgbe_driver_string[] =
                               "Intel(R) 10 Gigabit PCI Express Network Driver";
 
-#define DRV_VERSION "2.0.44-k2"
+#define DRV_VERSION "2.0.62-k2"
 const char ixgbe_driver_version[] = DRV_VERSION;
 static char ixgbe_copyright[] = "Copyright (c) 1999-2010 Intel Corporation.";
 
maintainer
#4 lore raw
In reply to Jeff Kirsher

From: Jeff Kirsher <[email protected]>
Date: Wed, 10 Feb 2010 18:07:18 -0800

Quoted 8 lines

One more round of fixes, based on feedback from Patrick McHardy

1) Change the list count to an unsigned value
2) Fix a memory leak
3) Removed an unnecessary list traversal in the ethtool core
4) Moved all list destruction to a helper function, allowing the driver to control when it clears the list (aside from when free_netdev() kills the cached list).

All applied, thanks.

Pj, can you look a shoring up the behavior of one more thing for me?

When rules get added, it first installs the filter by calling into the driver.

_Then_ is tries to add the software copy of the rule to the device list. This does an allocation which may fail.

If it does fail, we return -ENOMEM but we left the device configured with the new rule.

Probably better to do it something like:

	p = NULL;
	if (need_sw_rule_list)
		p = alloc_rule();
		if (!p)
			return -ENOMEM;
	}

	ret = op->install_rule();
	if (ret) {
		kfree(p);
		return ret;
	}
	
	if (need_sw_rule_list)
		sw_rule_insert(dev, p);

You get the idea, we can do the kfree() unconditonally because kfree(NULL) is OK, etc.

maintainer
#5 lore raw

From: David Miller <[email protected]>
Date: Wed, 10 Feb 2010 19:54:45 -0800 (PST)

Quoted 1 line

All applied, thanks.

Guys please check your changes for warnings in the build:

net/core/dev.c: In function ‘free_netdev’:
net/core/dev.c:5452: warning: unused variable ‘f’ net/core/dev.c:5452: warning: unused variable ‘fsc’

  CC [M]  drivers/net/e1000e/lib.o
  CC [M]  drivers/net/e1000e/phy.o
  CC      net/core/ethtool.o
  CC [M]  drivers/net/e1000e/param.o

net/core/ethtool.c: In function ‘ethtool_reset’:
net/core/ethtool.c:626: warning: unused variable ‘f’ net/core/ethtool.c:626: warning: unused variable ‘fsc’ net/core/ethtool.c: In function ‘ethtool_nway_reset’:
net/core/ethtool.c:676: warning: unused variable ‘f’ net/core/ethtool.c:676: warning: unused variable ‘fsc’

I'll fix this up, but you know... :-/

#6 lore raw
In reply to David Miller
Quoted 43 lines · David Miller

From: Jeff Kirsher <[email protected]>
Date: Wed, 10 Feb 2010 18:07:18 -0800

Quoted 8 lines

One more round of fixes, based on feedback from Patrick McHardy

1) Change the list count to an unsigned value
2) Fix a memory leak
3) Removed an unnecessary list traversal in the ethtool core
4) Moved all list destruction to a helper function, allowing the driver to control when it clears the list (aside from when free_netdev() kills the cached list).

All applied, thanks.

Pj, can you look a shoring up the behavior of one more thing for me?

When rules get added, it first installs the filter by calling into the driver.

_Then_ is tries to add the software copy of the rule to the device list. This does an allocation which may fail.

If it does fail, we return -ENOMEM but we left the device configured with the new rule.

Probably better to do it something like:

	p = NULL;
	if (need_sw_rule_list)
		p = alloc_rule();
		if (!p)
			return -ENOMEM;
	}

	ret = op->install_rule();
	if (ret) {
		kfree(p);
		return ret;
	}
	
	if (need_sw_rule_list)
		sw_rule_insert(dev, p);
Quoted 2 lines

You get the idea, we can do the kfree() unconditonally because kfree(NULL) is OK, etc.

No problem. I'll get a patch together asap for 2.6.34.

Also, probably for 2.6.35, I'm going to try and add a remove capability.

Cheers,
-PJ

#7 lore raw

includes build break - was Re: [net-next-2.6 PATCH v5 0/3] Introduce n-tuple ethtool support

In reply to David Miller
Quoted 5 lines · David Miller

From: Jeff Kirsher <[email protected]>
Date: Wed, 10 Feb 2010 18:07:18 -0800

Quoted 2 lines

One more round of fixes, based on feedback from Patrick McHardy

Quoted 9 lines
Quoted 6 lines

1) Change the list count to an unsigned value
2) Fix a memory leak
3) Removed an unnecessary list traversal in the ethtool core
4) Moved all list destruction to a helper function, allowing the driver to control when it clears the list (aside from when free_netdev() kills the cached list).

All applied, thanks.

Hi Peter,

compiling the latest net-next-2.6 i got this problem:

CHECK include/linux (358 files)
/home/hartko/net-next-2.6/usr/include/linux/ethtool.h:17: included file 'linux/rculist.h' is not exported

Which has been introduced by

http://git.kernel.org/?p=linux/kernel/git/davem/net-next-2.6.git;a=commitdiff;h=15682bc488d4af8c9bb998844a94281025e0a333

("ethtool: Introduce n-tuple filter programming support")

The patch below fixes the issue.

---

include/linux/ethtool.h +2 -0
index a3cac53..83a9a53 100644
@@ -14,7 +14,9 @@
 #define _LINUX_ETHTOOL_H
 
 #include <linux/types.h>
+#ifdef __KERNEL__
 #include <linux/rculist.h>
+#endif
 
 /* This should work for both 32 and 64 bit userland. */
 struct ethtool_cmd {
maintainer
#8 lore raw

Re: includes build break - was Re: [net-next-2.6 PATCH v5 0/3] Introduce n-tuple ethtool support

From: Oliver Hartkopp <[email protected]>
Date: Sun, 14 Feb 2010 12:56:28 +0100

Quoted 3 lines

The patch below fixes the issue.

Oliver, thanks for your patch.

I'm going to fix this slightly differently by moving the rculist.h include down into the existing __KERNEL__ ifdef protected code block.

Thanks!

#9 lore raw

v5 cover Introduce n-tuple ethtool support

In reply to Jeff Kirsher

Nice. Next step would be to control the RSS-traffic. Doing this we can easily get total separation between "control" and "data" plane. We did a dirty hack to separate out bgp/ssh to one queue/cpu. I think just a RSS_mask would be a good start - each bit controlling if queue should receive RSS traffic or not.

Cheers

				--ro
Quoted 35 lines · Jeff Kirsher

One more round of fixes, based on feedback from Patrick McHardy

1) Change the list count to an unsigned value
2) Fix a memory leak
3) Removed an unnecessary list traversal in the ethtool core
4) Moved all list destruction to a helper function, allowing the driver to control when it clears the list (aside from when free_netdev() kills the cached list).

---

Peter Waskiewicz (3):
ixgbe: Bump driver version up
ixgbe: Add support for the new ethtool n-tuple programming interface
ethtool: Introduce n-tuple filter programming support

 drivers/net/ixgbe/ixgbe.h         |    4 
 drivers/net/ixgbe/ixgbe_82599.c   |  106 ++++++++++--
 drivers/net/ixgbe/ixgbe_ethtool.c |  111 ++++++++++++
 drivers/net/ixgbe/ixgbe_main.c    |   18 ++
 drivers/net/ixgbe/ixgbe_type.h    |    9 +
 include/linux/ethtool.h           |   50 ++++++
 include/linux/netdevice.h         |    3 
 net/core/dev.c                    |    6 +
 net/core/ethtool.c                |  332 +++++++++++++++++++++++++++++++++++++
 9 files changed, 619 insertions(+), 20 deletions(-)
Signature · 6 lines
Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html