[ovs.git] / datapath / datapath.h

/*
 * Copyright (c) 2009 Nicira Networks.
 * Distributed under the terms of the GNU GPL version 2.
 *
 * Significant portions of this file may be copied from parts of the Linux
 * kernel, by Linus Torvalds and others.
 */

/* Interface exported by openvswitch_mod. */

#ifndef DATAPATH_H
#define DATAPATH_H 1

#include <asm/page.h>
#include <linux/kernel.h>
#include <linux/mutex.h>
#include <linux/netdevice.h>
#include <linux/workqueue.h>
#include <linux/skbuff.h>
#include <linux/version.h>
#include "flow.h"
#include "dp_sysfs.h"

/* Mask for the priority bits in a vlan header.  If we ever merge upstream
 * then this should go into include/linux/if_vlan.h. */
#define VLAN_PCP_MASK 0xe000
#define VLAN_PCP_SHIFT 13

#define DP_MAX_PORTS 1024
#define DP_MAX_GROUPS 16

#define DP_L2_BITS (PAGE_SHIFT - ilog2(sizeof(struct dp_bucket*)))
#define DP_L2_SIZE (1 << DP_L2_BITS)
#define DP_L2_SHIFT 0

#define DP_L1_BITS (PAGE_SHIFT - ilog2(sizeof(struct dp_bucket**)))
#define DP_L1_SIZE (1 << DP_L1_BITS)
#define DP_L1_SHIFT DP_L2_BITS

/* For 4 kB pages, this is 1,048,576 on 32-bit or 262,144 on 64-bit. */
#define DP_MAX_BUCKETS (DP_L1_SIZE * DP_L2_SIZE)

/**
 * struct dp_table - flow table
 * @n_buckets: number of buckets (a power of 2 between %DP_L1_SIZE and
 * %DP_MAX_BUCKETS)
 * @buckets: pointer to @n_buckets/%DP_L1_SIZE pointers to %DP_L1_SIZE pointers
 * to buckets
 * @hash_seed: random number used for flow hashing, to make the hash
 * distribution harder to predict
 * @rcu: RCU callback structure
 *
 * The @buckets array is logically an array of pointers to buckets.  It is
 * broken into two levels to avoid the need to kmalloc() any object larger than
 * a single page or to use vmalloc().  @buckets is always nonnull, as is each
 * @buckets[i], but each @buckets[i][j] is nonnull only if the specified hash
 * bucket is nonempty (for 0 <= i < @n_buckets/%DP_L1_SIZE, 0 <= j <
 * %DP_L1_SIZE).
 */
struct dp_table {
	unsigned int n_buckets;
	struct dp_bucket ***buckets;
	unsigned int hash_seed;
	struct rcu_head rcu;
};

/**
 * struct dp_bucket - single bucket within datapath flow table
 * @rcu: RCU callback structure
 * @n_flows: number of flows in @flows[] array
 * @flows: array of @n_flows pointers to flows
 *
 * The expected number of flows per bucket is 1, but this allows for an
 * arbitrary number of collisions.
 */
struct dp_bucket {
	struct rcu_head rcu;
	unsigned int n_flows;
	struct sw_flow *flows[];
};

#define DP_N_QUEUES 2
#define DP_MAX_QUEUE_LEN 100

struct dp_stats_percpu {
	u64 n_frags;
	u64 n_hit;
	u64 n_missed;
	u64 n_lost;
};

struct dp_port_group {
	struct rcu_head rcu;
	int n_ports;
	u16 ports[];
};

struct datapath {
	struct mutex mutex;
	int dp_idx;

	struct kobject ifobj;

	int drop_frags;

	/* Queued data. */
	struct sk_buff_head queues[DP_N_QUEUES];
	wait_queue_head_t waitqueue;

	/* Flow table. */
	unsigned int n_flows;
	struct dp_table *table;

	/* Port groups. */
	struct dp_port_group *groups[DP_MAX_GROUPS];

	/* Switch ports. */
	unsigned int n_ports;
	struct net_bridge_port *ports[DP_MAX_PORTS];
	struct list_head port_list; /* All ports, including local_port. */

	/* Stats. */
	struct dp_stats_percpu *stats_percpu;
};

struct net_bridge_port {
	u16 port_no;
	struct datapath	*dp;
	struct net_device *dev;
	struct kobject kobj;
	char linkname[IFNAMSIZ];
	struct list_head node;   /* Element in datapath.ports. */
};

extern struct notifier_block dp_device_notifier;
extern int (*dp_ioctl_hook)(struct net_device *dev, struct ifreq *rq, int cmd);

/* Flow table. */
struct dp_table *dp_table_create(unsigned int n_buckets);
void dp_table_destroy(struct dp_table *, int free_flows);
struct sw_flow *dp_table_lookup(struct dp_table *, const struct odp_flow_key *);
int dp_table_insert(struct dp_table *, struct sw_flow *);
int dp_table_delete(struct dp_table *, struct sw_flow *);
int dp_table_expand(struct datapath *);
int dp_table_flush(struct datapath *);
int dp_table_foreach(struct dp_table *table,
		     int (*callback)(struct sw_flow *flow, void *aux),
		     void *aux);

void dp_process_received_packet(struct sk_buff *, struct net_bridge_port *);
int dp_del_port(struct net_bridge_port *);
int dp_output_control(struct datapath *, struct sk_buff *, int, u32 arg);
int dp_min_mtu(const struct datapath *dp);

struct datapath *get_dp(int dp_idx);

static inline const char *dp_name(const struct datapath *dp)
{
	return dp->ports[ODPP_LOCAL]->dev->name;
}

#ifdef CONFIG_XEN
int skb_checksum_setup(struct sk_buff *skb);
#else
static inline int skb_checksum_setup(struct sk_buff *skb)
{
	return 0;
}
#endif

int vswitch_skb_checksum_setup(struct sk_buff *skb);

#endif /* datapath.h */
Commit	Line	Data
a14bc59f BP	1	/*
	2	* Copyright (c) 2009 Nicira Networks.
	3	* Distributed under the terms of the GNU GPL version 2.
	4	*
	5	* Significant portions of this file may be copied from parts of the Linux
	6	* kernel, by Linus Torvalds and others.
	7	*/
	8
064af421 BP	9	/* Interface exported by openvswitch_mod. */
	10
	11	#ifndef DATAPATH_H
	12	#define DATAPATH_H 1
	13
	14	#include <asm/page.h>
	15	#include <linux/kernel.h>
	16	#include <linux/mutex.h>
064af421 BP	17	#include <linux/netdevice.h>
	18	#include <linux/workqueue.h>
	19	#include <linux/skbuff.h>
806e39cf	20	#include <linux/version.h>
064af421	21	#include "flow.h"
2ba9026e	22	#include "dp_sysfs.h"
064af421	23
064af421 BP	24	/* Mask for the priority bits in a vlan header. If we ever merge upstream
	25	* then this should go into include/linux/if_vlan.h. */
	26	#define VLAN_PCP_MASK 0xe000
a4fbb689	27	#define VLAN_PCP_SHIFT 13
064af421	28
5eab9abc	29	#define DP_MAX_PORTS 1024
064af421 BP	30	#define DP_MAX_GROUPS 16
064af421 BP	31
6fa58f7a	32	#define DP_L2_BITS (PAGE_SHIFT - ilog2(sizeof(struct dp_bucket*)))
064af421 BP	33	#define DP_L2_SIZE (1 << DP_L2_BITS)
	34	#define DP_L2_SHIFT 0
	35
6fa58f7a	36	#define DP_L1_BITS (PAGE_SHIFT - ilog2(sizeof(struct dp_bucket**)))
064af421 BP	37	#define DP_L1_SIZE (1 << DP_L1_BITS)
	38	#define DP_L1_SHIFT DP_L2_BITS
	39
6fa58f7a	40	/* For 4 kB pages, this is 1,048,576 on 32-bit or 262,144 on 64-bit. */
064af421 BP	41	#define DP_MAX_BUCKETS (DP_L1_SIZE * DP_L2_SIZE)
064af421 BP	42
6fa58f7a BP	43	/**
	44	* struct dp_table - flow table
	45	* @n_buckets: number of buckets (a power of 2 between %DP_L1_SIZE and
	46	* %DP_MAX_BUCKETS)
	47	* @buckets: pointer to @n_buckets/%DP_L1_SIZE pointers to %DP_L1_SIZE pointers
	48	* to buckets
	49	* @hash_seed: random number used for flow hashing, to make the hash
	50	* distribution harder to predict
	51	* @rcu: RCU callback structure
	52	*
	53	* The @buckets array is logically an array of pointers to buckets. It is
	54	* broken into two levels to avoid the need to kmalloc() any object larger than
	55	* a single page or to use vmalloc(). @buckets is always nonnull, as is each
	56	* @buckets[i], but each @buckets[i][j] is nonnull only if the specified hash
	57	* bucket is nonempty (for 0 <= i < @n_buckets/%DP_L1_SIZE, 0 <= j <
	58	* %DP_L1_SIZE).
	59	*/
064af421 BP	60	struct dp_table {
064af421 BP	61	unsigned int n_buckets;
6fa58f7a BP	62	struct dp_bucket ***buckets;
	63	unsigned int hash_seed;
	64	struct rcu_head rcu;
	65	};
	66
	67	/**
	68	* struct dp_bucket - single bucket within datapath flow table
	69	* @rcu: RCU callback structure
	70	* @n_flows: number of flows in @flows[] array
	71	* @flows: array of @n_flows pointers to flows
	72	*
	73	* The expected number of flows per bucket is 1, but this allows for an
	74	* arbitrary number of collisions.
	75	*/
	76	struct dp_bucket {
064af421	77	struct rcu_head rcu;
6fa58f7a BP	78	unsigned int n_flows;
6fa58f7a BP	79	struct sw_flow *flows[];
064af421 BP	80	};
	81
	82	#define DP_N_QUEUES 2
	83	#define DP_MAX_QUEUE_LEN 100
	84
	85	struct dp_stats_percpu {
	86	u64 n_frags;
	87	u64 n_hit;
	88	u64 n_missed;
	89	u64 n_lost;
	90	};
	91
	92	struct dp_port_group {
	93	struct rcu_head rcu;
	94	int n_ports;
	95	u16 ports[];
	96	};
	97
	98	struct datapath {
	99	struct mutex mutex;
	100	int dp_idx;
	101
064af421	102	struct kobject ifobj;
064af421 BP	103
	104	int drop_frags;
	105
	106	/* Queued data. */
	107	struct sk_buff_head queues[DP_N_QUEUES];
	108	wait_queue_head_t waitqueue;
	109
	110	/* Flow table. */
	111	unsigned int n_flows;
	112	struct dp_table *table;
	113
	114	/* Port groups. */
	115	struct dp_port_group *groups[DP_MAX_GROUPS];
	116
	117	/* Switch ports. */
	118	unsigned int n_ports;
	119	struct net_bridge_port *ports[DP_MAX_PORTS];
	120	struct list_head port_list; /* All ports, including local_port. */
	121
	122	/* Stats. */
	123	struct dp_stats_percpu *stats_percpu;
	124	};
	125
	126	struct net_bridge_port {
	127	u16 port_no;
	128	struct datapath *dp;
	129	struct net_device *dev;
064af421	130	struct kobject kobj;
0515ceb3	131	char linkname[IFNAMSIZ];
064af421 BP	132	struct list_head node; /* Element in datapath.ports. */
	133	};
	134
	135	extern struct notifier_block dp_device_notifier;
	136	extern int (dp_ioctl_hook)(struct net_device dev, struct ifreq *rq, int cmd);
064af421 BP	137
	138	/* Flow table. */
	139	struct dp_table *dp_table_create(unsigned int n_buckets);
	140	void dp_table_destroy(struct dp_table *, int free_flows);
	141	struct sw_flow dp_table_lookup(struct dp_table , const struct odp_flow_key *);
6fa58f7a	142	int dp_table_insert(struct dp_table , struct sw_flow );
064af421 BP	143	int dp_table_delete(struct dp_table , struct sw_flow );
	144	int dp_table_expand(struct datapath *);
	145	int dp_table_flush(struct datapath *);
	146	int dp_table_foreach(struct dp_table *table,
	147	int (callback)(struct sw_flow flow, void *aux),
	148	void *aux);
	149
	150	void dp_process_received_packet(struct sk_buff , struct net_bridge_port );
72ca14c1	151	int dp_del_port(struct net_bridge_port *);
064af421	152	int dp_output_control(struct datapath , struct sk_buff , int, u32 arg);
1dcf111b	153	int dp_min_mtu(const struct datapath *dp);
064af421 BP	154
	155	struct datapath *get_dp(int dp_idx);
	156
	157	static inline const char dp_name(const struct datapath dp)
	158	{
	159	return dp->ports[ODPP_LOCAL]->dev->name;
	160	}
	161
	162	#ifdef CONFIG_XEN
	163	int skb_checksum_setup(struct sk_buff *skb);
	164	#else
	165	static inline int skb_checksum_setup(struct sk_buff *skb)
	166	{
	167	return 0;
	168	}
	169	#endif
	170
b2f460c7 IC	171	int vswitch_skb_checksum_setup(struct sk_buff *skb);
b2f460c7 IC	172
064af421	173	#endif /* datapath.h */