[PATCH 2/3] [kernel] Add support for RX packet classification in a network device

[PATCH 2/3] [kernel] Add support for RX packet classification in a network device

[Santwona.Behera]

[-- Attachment #1: Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: ethtool-kernel-add-rx-pkt-classification-interface.patch --]
[-- Type: text/x-patch, Size: 6790 bytes --]

From: Santwona Behera <santwona.behera@sun.com>
Subject: [PATCH 2/3] [Ethtool] Add RX pkt classification interface


Signed-off-by: Santwona Behera <santwona.behera@sun.com>
---
 include/linux/ethtool.h |   84 +++++++++++++++++++++++++++++++++++++++++++---
 net/core/ethtool.c      |   58 +++++++++++++++++++++++++++-----
 2 files changed, 126 insertions(+), 16 deletions(-)

diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index 27c67a5..0372076 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -287,10 +287,71 @@ enum ethtool_flags {
 	ETH_FLAG_LRO		= (1 << 15),	/* LRO is enabled */
 };
 
-struct ethtool_rxnfc {
-	__u32		cmd;
+struct ethtool_tcpip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u16	psrc;
+	__u16	pdst;
+	__u8    tos;
+};
+
+struct ethtool_ah_espip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u32	spi;
+	__u8    tos;
+};
+
+struct ethtool_rawip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u8	hdata[64];
+};
+
+struct ethtool_ether_spec {
+	__u16	ether_type;
+	__u8	frame_size;
+	__u8	eframe[16];
+};
+
+#define	ETH_RX_NFC_IP4	1
+#define	ETH_RX_NFC_IP6	2
+
+struct ethtool_usrip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u32	l4_4_bytes;
+	__u8    tos;
+	__u8    ip_ver;
+	__u8    proto;
+};
+
+struct ethtool_rx_flow_spec {
 	__u32		flow_type;
-	__u64		data;
+	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 */
+	__u64		ring_cookie;
+	__u32		location;
+};
+
+struct ethtool_rxnfc {
+	__u32				cmd;
+	__u32				flow_type;
+	/* The rx flow hash value or the rule DB size */
+	__u64				data;
+	struct ethtool_rx_flow_spec	fs;
+	__u32				rule_cnt;
+	__u32				rule_locs[0];
 };
 
 #ifdef __KERNEL__
@@ -417,8 +478,8 @@ struct ethtool_ops {
 	/* the following hooks are obsolete */
 	int	(*self_test_count)(struct net_device *);/* use get_sset_count */
 	int	(*get_stats_count)(struct net_device *);/* use get_sset_count */
-	int	(*get_rxhash)(struct net_device *, struct ethtool_rxnfc *);
-	int	(*set_rxhash)(struct net_device *, struct ethtool_rxnfc *);
+	int	(*get_rxnfc)(struct net_device *, struct ethtool_rxnfc *, void *);
+	int	(*set_rxnfc)(struct net_device *, struct ethtool_rxnfc *);
 };
 #endif /* __KERNEL__ */
 
@@ -469,6 +530,12 @@ struct ethtool_ops {
 #define	ETHTOOL_SRXFH		0x0000002a /* Set RX flow hash configuration */
 #define ETHTOOL_GGRO		0x0000002b /* Get GRO enable (ethtool_value) */
 #define ETHTOOL_SGRO		0x0000002c /* Set GRO enable (ethtool_value) */
+#define	ETHTOOL_GRXRINGS	0x0000002d /* Get RX rings available for LB */
+#define	ETHTOOL_GRXCLSRLCNT	0x0000002e /* Get RX class rule count */
+#define	ETHTOOL_GRXCLSRULE	0x0000002f /* Get RX classification rule */
+#define	ETHTOOL_GRXCLSRLALL	0x00000030 /* Get all RX classification rule */
+#define	ETHTOOL_SRXCLSRLDEL	0x00000031 /* Delete RX classification rule */
+#define	ETHTOOL_SRXCLSRLINS	0x00000032 /* Insert RX classification rule */
 
 /* compatibility with older code */
 #define SPARC_ETH_GSET		ETHTOOL_GSET
@@ -565,9 +632,13 @@ struct ethtool_ops {
 #define	UDP_V6_FLOW	0x06
 #define	SCTP_V6_FLOW	0x07
 #define	AH_ESP_V6_FLOW	0x08
+#define	AH_V4_FLOW	0x09
+#define	ESP_V4_FLOW	0x0a
+#define	AH_V6_FLOW	0x0b
+#define	ESP_V6_FLOW	0x0c
+#define	IP_USER_FLOW	0x0d
 
 /* L3-L4 network traffic flow hash options */
-#define	RXH_DEV_PORT	(1 << 0)
 #define	RXH_L2DA	(1 << 1)
 #define	RXH_VLAN	(1 << 2)
 #define	RXH_L3_PROTO	(1 << 3)
@@ -577,5 +648,6 @@ struct ethtool_ops {
 #define	RXH_L4_B_2_3	(1 << 7) /* dst port in case of TCP/UDP/SCTP */
 #define	RXH_DISCARD	(1 << 31)
 
+#define	RX_CLS_FLOW_DISC	0xffffffffffffffffULL
 
 #endif /* _LINUX_ETHTOOL_H */
diff --git a/net/core/ethtool.c b/net/core/ethtool.c
index 947710a..244ca56 100644
--- a/net/core/ethtool.c
+++ b/net/core/ethtool.c
@@ -209,34 +209,62 @@ static int ethtool_get_drvinfo(struct net_device *dev, void __user *useraddr)
 	return 0;
 }
 
-static int ethtool_set_rxhash(struct net_device *dev, void __user *useraddr)
+static int ethtool_set_rxnfc(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_rxnfc cmd;
 
-	if (!dev->ethtool_ops->set_rxhash)
+	if (!dev->ethtool_ops->set_rxnfc)
 		return -EOPNOTSUPP;
 
 	if (copy_from_user(&cmd, useraddr, sizeof(cmd)))
 		return -EFAULT;
 
-	return dev->ethtool_ops->set_rxhash(dev, &cmd);
+	return dev->ethtool_ops->set_rxnfc(dev, &cmd);
 }
 
-static int ethtool_get_rxhash(struct net_device *dev, void __user *useraddr)
+static int ethtool_get_rxnfc(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_rxnfc info;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	int ret;
+	void *rule_buf = NULL;
 
-	if (!dev->ethtool_ops->get_rxhash)
+	if (!ops->get_rxnfc)
 		return -EOPNOTSUPP;
 
 	if (copy_from_user(&info, useraddr, sizeof(info)))
 		return -EFAULT;
 
-	dev->ethtool_ops->get_rxhash(dev, &info);
+	if (info.cmd == ETHTOOL_GRXCLSRLALL) {
+		if (info.rule_cnt > 0) {
+			rule_buf = kmalloc(info.rule_cnt * sizeof(u32),
+					   GFP_USER);
+			if (!rule_buf)
+				return -ENOMEM;
+		}
+	}
 
+	ret = ops->get_rxnfc(dev, &info, rule_buf);
+	if (ret < 0)
+		goto err_out;
+
+	ret = -EFAULT;
 	if (copy_to_user(useraddr, &info, sizeof(info)))
-		return -EFAULT;
-	return 0;
+		goto err_out;
+
+	if (rule_buf) {
+		useraddr += offsetof(struct ethtool_rxnfc, rule_locs);
+		if (copy_to_user(useraddr, rule_buf,
+				 info.rule_cnt * sizeof(u32)))
+			goto err_out;
+	}
+	ret = 0;
+
+err_out:
+	if (rule_buf)
+		kfree(rule_buf);
+
+	return ret;
 }
 
 static int ethtool_get_regs(struct net_device *dev, char __user *useraddr)
@@ -901,6 +929,10 @@ int dev_ethtool(struct net *net, struct ifreq *ifr)
 	case ETHTOOL_GFLAGS:
 	case ETHTOOL_GPFLAGS:
 	case ETHTOOL_GRXFH:
+	case ETHTOOL_GRXRINGS:
+	case ETHTOOL_GRXCLSRLCNT:
+	case ETHTOOL_GRXCLSRULE:
+	case ETHTOOL_GRXCLSRLALL:
 		break;
 	default:
 		if (!capable(CAP_NET_ADMIN))
@@ -1052,10 +1084,16 @@ int dev_ethtool(struct net *net, struct ifreq *ifr)
 				       dev->ethtool_ops->set_priv_flags);
 		break;
 	case ETHTOOL_GRXFH:
-		rc = ethtool_get_rxhash(dev, useraddr);
+	case ETHTOOL_GRXRINGS:
+	case ETHTOOL_GRXCLSRLCNT:
+	case ETHTOOL_GRXCLSRULE:
+	case ETHTOOL_GRXCLSRLALL:
+		rc = ethtool_get_rxnfc(dev, useraddr);
 		break;
 	case ETHTOOL_SRXFH:
-		rc = ethtool_set_rxhash(dev, useraddr);
+	case ETHTOOL_SRXCLSRLDEL:
+	case ETHTOOL_SRXCLSRLINS:
+		rc = ethtool_set_rxnfc(dev, useraddr);
 		break;
 	case ETHTOOL_GGRO:
 		rc = ethtool_get_gro(dev, useraddr);
-- 
1.6.0.4

[PATCH 2/3] [kernel] Add support for RX packet classification in a network device

[Santwona.Behera]

[-- Attachment #1: Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: ethtool-kernel-add-rx-pkt-classification-interface.patch --]
[-- Type: text/x-patch, Size: 7293 bytes --]

From: Santwona Behera <santwona.behera@sun.com>
Subject: [PATCH 2/3] [Ethtool] Add RX pkt classification interface


Signed-off-by: Santwona Behera <santwona.behera@sun.com>
---
 include/linux/ethtool.h |   90 +++++++++++++++++++++++++++++++++++++++++++---
 net/core/ethtool.c      |   58 +++++++++++++++++++++++++-----
 2 files changed, 132 insertions(+), 16 deletions(-)

diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index 27c67a5..cb0e1c5 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -7,6 +7,7 @@
  * Portions Copyright 2002 Intel (eli.kupermann@intel.com,
  *                                christopher.leech@intel.com,
  *                                scott.feldman@intel.com)
+ * Portions Copyright (C) Sun Microsystems 2008
  */
 
 #ifndef _LINUX_ETHTOOL_H
@@ -287,10 +288,76 @@ enum ethtool_flags {
 	ETH_FLAG_LRO		= (1 << 15),	/* LRO is enabled */
 };
 
-struct ethtool_rxnfc {
-	__u32		cmd;
+/* The following structures are for supporting RX network flow
+ * classification configuration. Note, all multibyte fields, e.g.,
+ * ip4src, ip4dst, psrc, pdst, spi, etc. are expected to be in network
+ * byte order.
+ */
+struct ethtool_tcpip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u16	psrc;
+	__u16	pdst;
+	__u8    tos;
+};
+
+struct ethtool_ah_espip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u32	spi;
+	__u8    tos;
+};
+
+struct ethtool_rawip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u8	hdata[64];
+};
+
+struct ethtool_ether_spec {
+	__u16	ether_type;
+	__u8	frame_size;
+	__u8	eframe[16];
+};
+
+#define	ETH_RX_NFC_IP4	1
+#define	ETH_RX_NFC_IP6	2
+
+struct ethtool_usrip4_spec {
+	__u32	ip4src;
+	__u32	ip4dst;
+	__u32	l4_4_bytes;
+	__u8    tos;
+	__u8    ip_ver;
+	__u8    proto;
+};
+
+struct ethtool_rx_flow_spec {
 	__u32		flow_type;
-	__u64		data;
+	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 */
+	__u64		ring_cookie;
+	__u32		location;
+};
+
+struct ethtool_rxnfc {
+	__u32				cmd;
+	__u32				flow_type;
+	/* The rx flow hash value or the rule DB size */
+	__u64				data;
+	struct ethtool_rx_flow_spec	fs;
+	__u32				rule_cnt;
+	__u32				rule_locs[0];
 };
 
 #ifdef __KERNEL__
@@ -417,8 +484,8 @@ struct ethtool_ops {
 	/* the following hooks are obsolete */
 	int	(*self_test_count)(struct net_device *);/* use get_sset_count */
 	int	(*get_stats_count)(struct net_device *);/* use get_sset_count */
-	int	(*get_rxhash)(struct net_device *, struct ethtool_rxnfc *);
-	int	(*set_rxhash)(struct net_device *, struct ethtool_rxnfc *);
+	int	(*get_rxnfc)(struct net_device *, struct ethtool_rxnfc *, void *);
+	int	(*set_rxnfc)(struct net_device *, struct ethtool_rxnfc *);
 };
 #endif /* __KERNEL__ */
 
@@ -469,6 +536,12 @@ struct ethtool_ops {
 #define	ETHTOOL_SRXFH		0x0000002a /* Set RX flow hash configuration */
 #define ETHTOOL_GGRO		0x0000002b /* Get GRO enable (ethtool_value) */
 #define ETHTOOL_SGRO		0x0000002c /* Set GRO enable (ethtool_value) */
+#define	ETHTOOL_GRXRINGS	0x0000002d /* Get RX rings available for LB */
+#define	ETHTOOL_GRXCLSRLCNT	0x0000002e /* Get RX class rule count */
+#define	ETHTOOL_GRXCLSRULE	0x0000002f /* Get RX classification rule */
+#define	ETHTOOL_GRXCLSRLALL	0x00000030 /* Get all RX classification rule */
+#define	ETHTOOL_SRXCLSRLDEL	0x00000031 /* Delete RX classification rule */
+#define	ETHTOOL_SRXCLSRLINS	0x00000032 /* Insert RX classification rule */
 
 /* compatibility with older code */
 #define SPARC_ETH_GSET		ETHTOOL_GSET
@@ -565,9 +638,13 @@ struct ethtool_ops {
 #define	UDP_V6_FLOW	0x06
 #define	SCTP_V6_FLOW	0x07
 #define	AH_ESP_V6_FLOW	0x08
+#define	AH_V4_FLOW	0x09
+#define	ESP_V4_FLOW	0x0a
+#define	AH_V6_FLOW	0x0b
+#define	ESP_V6_FLOW	0x0c
+#define	IP_USER_FLOW	0x0d
 
 /* L3-L4 network traffic flow hash options */
-#define	RXH_DEV_PORT	(1 << 0)
 #define	RXH_L2DA	(1 << 1)
 #define	RXH_VLAN	(1 << 2)
 #define	RXH_L3_PROTO	(1 << 3)
@@ -577,5 +654,6 @@ struct ethtool_ops {
 #define	RXH_L4_B_2_3	(1 << 7) /* dst port in case of TCP/UDP/SCTP */
 #define	RXH_DISCARD	(1 << 31)
 
+#define	RX_CLS_FLOW_DISC	0xffffffffffffffffULL
 
 #endif /* _LINUX_ETHTOOL_H */
diff --git a/net/core/ethtool.c b/net/core/ethtool.c
index 947710a..244ca56 100644
--- a/net/core/ethtool.c
+++ b/net/core/ethtool.c
@@ -209,34 +209,62 @@ static int ethtool_get_drvinfo(struct net_device *dev, void __user *useraddr)
 	return 0;
 }
 
-static int ethtool_set_rxhash(struct net_device *dev, void __user *useraddr)
+static int ethtool_set_rxnfc(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_rxnfc cmd;
 
-	if (!dev->ethtool_ops->set_rxhash)
+	if (!dev->ethtool_ops->set_rxnfc)
 		return -EOPNOTSUPP;
 
 	if (copy_from_user(&cmd, useraddr, sizeof(cmd)))
 		return -EFAULT;
 
-	return dev->ethtool_ops->set_rxhash(dev, &cmd);
+	return dev->ethtool_ops->set_rxnfc(dev, &cmd);
 }
 
-static int ethtool_get_rxhash(struct net_device *dev, void __user *useraddr)
+static int ethtool_get_rxnfc(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_rxnfc info;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	int ret;
+	void *rule_buf = NULL;
 
-	if (!dev->ethtool_ops->get_rxhash)
+	if (!ops->get_rxnfc)
 		return -EOPNOTSUPP;
 
 	if (copy_from_user(&info, useraddr, sizeof(info)))
 		return -EFAULT;
 
-	dev->ethtool_ops->get_rxhash(dev, &info);
+	if (info.cmd == ETHTOOL_GRXCLSRLALL) {
+		if (info.rule_cnt > 0) {
+			rule_buf = kmalloc(info.rule_cnt * sizeof(u32),
+					   GFP_USER);
+			if (!rule_buf)
+				return -ENOMEM;
+		}
+	}
 
+	ret = ops->get_rxnfc(dev, &info, rule_buf);
+	if (ret < 0)
+		goto err_out;
+
+	ret = -EFAULT;
 	if (copy_to_user(useraddr, &info, sizeof(info)))
-		return -EFAULT;
-	return 0;
+		goto err_out;
+
+	if (rule_buf) {
+		useraddr += offsetof(struct ethtool_rxnfc, rule_locs);
+		if (copy_to_user(useraddr, rule_buf,
+				 info.rule_cnt * sizeof(u32)))
+			goto err_out;
+	}
+	ret = 0;
+
+err_out:
+	if (rule_buf)
+		kfree(rule_buf);
+
+	return ret;
 }
 
 static int ethtool_get_regs(struct net_device *dev, char __user *useraddr)
@@ -901,6 +929,10 @@ int dev_ethtool(struct net *net, struct ifreq *ifr)
 	case ETHTOOL_GFLAGS:
 	case ETHTOOL_GPFLAGS:
 	case ETHTOOL_GRXFH:
+	case ETHTOOL_GRXRINGS:
+	case ETHTOOL_GRXCLSRLCNT:
+	case ETHTOOL_GRXCLSRULE:
+	case ETHTOOL_GRXCLSRLALL:
 		break;
 	default:
 		if (!capable(CAP_NET_ADMIN))
@@ -1052,10 +1084,16 @@ int dev_ethtool(struct net *net, struct ifreq *ifr)
 				       dev->ethtool_ops->set_priv_flags);
 		break;
 	case ETHTOOL_GRXFH:
-		rc = ethtool_get_rxhash(dev, useraddr);
+	case ETHTOOL_GRXRINGS:
+	case ETHTOOL_GRXCLSRLCNT:
+	case ETHTOOL_GRXCLSRULE:
+	case ETHTOOL_GRXCLSRLALL:
+		rc = ethtool_get_rxnfc(dev, useraddr);
 		break;
 	case ETHTOOL_SRXFH:
-		rc = ethtool_set_rxhash(dev, useraddr);
+	case ETHTOOL_SRXCLSRLDEL:
+	case ETHTOOL_SRXCLSRLINS:
+		rc = ethtool_set_rxnfc(dev, useraddr);
 		break;
 	case ETHTOOL_GGRO:
 		rc = ethtool_get_gro(dev, useraddr);
-- 
1.5.2

Re: [PATCH 2/3] [kernel] Add support for RX packet classification in a network device

[David Miller]

From: Santwona.Behera@Sun.COM
Date: Mon, 16 Feb 2009 09:17:23 -0800

> -struct ethtool_rxnfc {
> -	__u32		cmd;
> +/* The following structures are for supporting RX network flow
> + * classification configuration. Note, all multibyte fields, e.g.,
> + * ip4src, ip4dst, psrc, pdst, spi, etc. are expected to be in network
> + * byte order.
> + */
> +struct ethtool_tcpip4_spec {
> +	__u32	ip4src;
> +	__u32	ip4dst;
> +	__u16	psrc;
> +	__u16	pdst;
> +	__u8    tos;
> +};

We have specific types for this, "__be32" et al.

Please use them.

And once you do that properly, checking the code with sparse via "make
C=1" will warn for you if these data items are not accessed as
big-endian items.

Re: [PATCH 2/3] [kernel] Add support for RX packet classification in a network device

[Harvey Harrison]

On Tue, 2009-02-17 at 20:28 -0800, David Miller wrote:
> From: Santwona.Behera@Sun.COM
> Date: Mon, 16 Feb 2009 09:17:23 -0800
> 
> > -struct ethtool_rxnfc {
> > -	__u32		cmd;
> > +/* The following structures are for supporting RX network flow
> > + * classification configuration. Note, all multibyte fields, e.g.,
> > + * ip4src, ip4dst, psrc, pdst, spi, etc. are expected to be in network
> > + * byte order.
> > + */
> > +struct ethtool_tcpip4_spec {
> > +	__u32	ip4src;
> > +	__u32	ip4dst;
> > +	__u16	psrc;
> > +	__u16	pdst;
> > +	__u8    tos;
> > +};
> 
> We have specific types for this, "__be32" et al.
> 
> Please use them.
> 
> And once you do that properly, checking the code with sparse via "make
> C=1" will warn for you if these data items are not accessed as
> big-endian items.

make C=1 CF=-D__CHECK_ENDIAN__

The endian checks aren't on yet by default due to the noise level, but
after the work over the past few kernel releases it may be getting
close.

I'll look again and see what it looks like for the next devel cycle.

Harvey

Re: [PATCH 2/3] [kernel] Add support for RX packet classification in a network device

[David Miller]

From: David Miller <davem@davemloft.net>
Date: Tue, 17 Feb 2009 20:28:38 -0800 (PST)

> From: Santwona.Behera@Sun.COM
> Date: Mon, 16 Feb 2009 09:17:23 -0800
> 
> > -struct ethtool_rxnfc {
> > -	__u32		cmd;
> > +/* The following structures are for supporting RX network flow
> > + * classification configuration. Note, all multibyte fields, e.g.,
> > + * ip4src, ip4dst, psrc, pdst, spi, etc. are expected to be in network
> > + * byte order.
> > + */
> > +struct ethtool_tcpip4_spec {
> > +	__u32	ip4src;
> > +	__u32	ip4dst;
> > +	__u16	psrc;
> > +	__u16	pdst;
> > +	__u8    tos;
> > +};
> 
> We have specific types for this, "__be32" et al.

Nevermind, you obviously don't have time to get to this quickly
so I'll take care of it while checking this stuff in...