[mirror_ubuntu-artful-kernel.git] / include / linux / ceph / messenger.h

#ifndef __FS_CEPH_MESSENGER_H
#define __FS_CEPH_MESSENGER_H

#include <linux/kref.h>
#include <linux/mutex.h>
#include <linux/net.h>
#include <linux/radix-tree.h>
#include <linux/uio.h>
#include <linux/workqueue.h>

#include <linux/ceph/types.h>
#include <linux/ceph/buffer.h>

struct ceph_msg;
struct ceph_connection;

/*
 * Ceph defines these callbacks for handling connection events.
 */
struct ceph_connection_operations {
	struct ceph_connection *(*get)(struct ceph_connection *);
	void (*put)(struct ceph_connection *);

	/* handle an incoming message. */
	void (*dispatch) (struct ceph_connection *con, struct ceph_msg *m);

	/* authorize an outgoing connection */
	struct ceph_auth_handshake *(*get_authorizer) (
				struct ceph_connection *con,
			       int *proto, int force_new);
	int (*verify_authorizer_reply) (struct ceph_connection *con, int len);
	int (*invalidate_authorizer)(struct ceph_connection *con);

	/* there was some error on the socket (disconnect, whatever) */
	void (*fault) (struct ceph_connection *con);

	/* a remote host as terminated a message exchange session, and messages
	 * we sent (or they tried to send us) may be lost. */
	void (*peer_reset) (struct ceph_connection *con);

	struct ceph_msg * (*alloc_msg) (struct ceph_connection *con,
					struct ceph_msg_header *hdr,
					int *skip);
};

/* use format string %s%d */
#define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)

struct ceph_messenger {
	struct ceph_entity_inst inst;    /* my name+address */
	struct ceph_entity_addr my_enc_addr;

	atomic_t stopping;
	bool nocrc;

	/*
	 * the global_seq counts connections i (attempt to) initiate
	 * in order to disambiguate certain connect race conditions.
	 */
	u32 global_seq;
	spinlock_t global_seq_lock;

	u32 supported_features;
	u32 required_features;
};

enum ceph_msg_data_type {
	CEPH_MSG_DATA_NONE,	/* message contains no data payload */
	CEPH_MSG_DATA_PAGES,	/* data source/destination is a page array */
	CEPH_MSG_DATA_PAGELIST,	/* data source/destination is a pagelist */
#ifdef CONFIG_BLOCK
	CEPH_MSG_DATA_BIO,	/* data source/destination is a bio list */
#endif /* CONFIG_BLOCK */
};

static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type)
{
	switch (type) {
	case CEPH_MSG_DATA_NONE:
	case CEPH_MSG_DATA_PAGES:
	case CEPH_MSG_DATA_PAGELIST:
#ifdef CONFIG_BLOCK
	case CEPH_MSG_DATA_BIO:
#endif /* CONFIG_BLOCK */
		return true;
	default:
		return false;
	}
}

struct ceph_msg_data {
	enum ceph_msg_data_type		type;
	union {
#ifdef CONFIG_BLOCK
		struct {
			struct bio	*bio;
			size_t		bio_length;
		};
#endif /* CONFIG_BLOCK */
		struct {
			struct page	**pages;	/* NOT OWNER. */
			size_t		length;		/* total # bytes */
			unsigned int	alignment;	/* first page */
		};
		struct ceph_pagelist	*pagelist;
	};
};

struct ceph_msg_data_cursor {
	struct ceph_msg_data	*data;		/* data item this describes */
	size_t			resid;		/* bytes not yet consumed */
	bool			last_piece;	/* current is last piece */
	bool			need_crc;	/* crc update needed */
	union {
#ifdef CONFIG_BLOCK
		struct {				/* bio */
			struct bio	*bio;		/* bio from list */
			unsigned int	vector_index;	/* vector from bio */
			unsigned int	vector_offset;	/* bytes from vector */
		};
#endif /* CONFIG_BLOCK */
		struct {				/* pages */
			unsigned int	page_offset;	/* offset in page */
			unsigned short	page_index;	/* index in array */
			unsigned short	page_count;	/* pages in array */
		};
		struct {				/* pagelist */
			struct page	*page;		/* page from list */
			size_t		offset;		/* bytes from list */
		};
	};
};

/*
 * a single message.  it contains a header (src, dest, message type, etc.),
 * footer (crc values, mainly), a "front" message body, and possibly a
 * data payload (stored in some number of pages).
 */
struct ceph_msg {
	struct ceph_msg_header hdr;	/* header */
	struct ceph_msg_footer footer;	/* footer */
	struct kvec front;              /* unaligned blobs of message */
	struct ceph_buffer *middle;

	size_t				data_length;
	struct ceph_msg_data		*data;
	struct ceph_msg_data_cursor	cursor;

	struct ceph_connection *con;
	struct list_head list_head;	/* links for connection lists */

	struct kref kref;
	bool front_is_vmalloc;
	bool more_to_follow;
	bool needs_out_seq;
	int front_max;
	unsigned long ack_stamp;        /* tx: when we were acked */

	struct ceph_msgpool *pool;
};

/* ceph connection fault delay defaults, for exponential backoff */
#define BASE_DELAY_INTERVAL	(HZ/2)
#define MAX_DELAY_INTERVAL	(5 * 60 * HZ)

/*
 * A single connection with another host.
 *
 * We maintain a queue of outgoing messages, and some session state to
 * ensure that we can preserve the lossless, ordered delivery of
 * messages in the case of a TCP disconnect.
 */
struct ceph_connection {
	void *private;

	const struct ceph_connection_operations *ops;

	struct ceph_messenger *msgr;

	atomic_t sock_state;
	struct socket *sock;
	struct ceph_entity_addr peer_addr; /* peer address */
	struct ceph_entity_addr peer_addr_for_me;

	unsigned long flags;
	unsigned long state;
	const char *error_msg;  /* error message, if any */

	struct ceph_entity_name peer_name; /* peer name */

	unsigned peer_features;
	u32 connect_seq;      /* identify the most recent connection
				 attempt for this connection, client */
	u32 peer_global_seq;  /* peer's global seq for this connection */

	int auth_retry;       /* true if we need a newer authorizer */
	void *auth_reply_buf;   /* where to put the authorizer reply */
	int auth_reply_buf_len;

	struct mutex mutex;

	/* out queue */
	struct list_head out_queue;
	struct list_head out_sent;   /* sending or sent but unacked */
	u64 out_seq;		     /* last message queued for send */

	u64 in_seq, in_seq_acked;  /* last message received, acked */

	/* connection negotiation temps */
	char in_banner[CEPH_BANNER_MAX_LEN];
	struct ceph_msg_connect out_connect;
	struct ceph_msg_connect_reply in_reply;
	struct ceph_entity_addr actual_peer_addr;

	/* message out temps */
	struct ceph_msg *out_msg;        /* sending message (== tail of
					    out_sent) */
	bool out_msg_done;

	struct kvec out_kvec[8],         /* sending header/footer data */
		*out_kvec_cur;
	int out_kvec_left;   /* kvec's left in out_kvec */
	int out_skip;        /* skip this many bytes */
	int out_kvec_bytes;  /* total bytes left */
	bool out_kvec_is_msg; /* kvec refers to out_msg */
	int out_more;        /* there is more data after the kvecs */
	__le64 out_temp_ack; /* for writing an ack */

	/* message in temps */
	struct ceph_msg_header in_hdr;
	struct ceph_msg *in_msg;
	u32 in_front_crc, in_middle_crc, in_data_crc;  /* calculated crc */

	char in_tag;         /* protocol control byte */
	int in_base_pos;     /* bytes read */
	__le64 in_temp_ack;  /* for reading an ack */

	struct delayed_work work;	    /* send|recv work */
	unsigned long       delay;          /* current delay interval */
};


extern const char *ceph_pr_addr(const struct sockaddr_storage *ss);
extern int ceph_parse_ips(const char *c, const char *end,
			  struct ceph_entity_addr *addr,
			  int max_count, int *count);


extern int ceph_msgr_init(void);
extern void ceph_msgr_exit(void);
extern void ceph_msgr_flush(void);

extern void ceph_messenger_init(struct ceph_messenger *msgr,
			struct ceph_entity_addr *myaddr,
			u32 supported_features,
			u32 required_features,
			bool nocrc);

extern void ceph_con_init(struct ceph_connection *con, void *private,
			const struct ceph_connection_operations *ops,
			struct ceph_messenger *msgr);
extern void ceph_con_open(struct ceph_connection *con,
			  __u8 entity_type, __u64 entity_num,
			  struct ceph_entity_addr *addr);
extern bool ceph_con_opened(struct ceph_connection *con);
extern void ceph_con_close(struct ceph_connection *con);
extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);

extern void ceph_msg_revoke(struct ceph_msg *msg);
extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);

extern void ceph_con_keepalive(struct ceph_connection *con);

extern void ceph_msg_data_set_pages(struct ceph_msg *msg, struct page **pages,
				size_t length, size_t alignment);
extern void ceph_msg_data_set_pagelist(struct ceph_msg *msg,
				struct ceph_pagelist *pagelist);
#ifdef CONFIG_BLOCK
extern void ceph_msg_data_set_bio(struct ceph_msg *msg, struct bio *bio,
				size_t length);
#endif /* CONFIG_BLOCK */

extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
				     bool can_fail);
extern void ceph_msg_kfree(struct ceph_msg *m);


static inline struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
{
	kref_get(&msg->kref);
	return msg;
}
extern void ceph_msg_last_put(struct kref *kref);
static inline void ceph_msg_put(struct ceph_msg *msg)
{
	kref_put(&msg->kref, ceph_msg_last_put);
}

extern void ceph_msg_dump(struct ceph_msg *msg);

#endif
Commit	Line	Data
31b8006e SW	1	#ifndef __FS_CEPH_MESSENGER_H
	2	#define __FS_CEPH_MESSENGER_H
	3
c2e552e7	4	#include <linux/kref.h>
31b8006e SW	5	#include <linux/mutex.h>
	6	#include <linux/net.h>
	7	#include <linux/radix-tree.h>
	8	#include <linux/uio.h>
31b8006e SW	9	#include <linux/workqueue.h>
31b8006e SW	10
a1ce3928 DH	11	#include <linux/ceph/types.h>
a1ce3928 DH	12	#include <linux/ceph/buffer.h>
31b8006e SW	13
	14	struct ceph_msg;
	15	struct ceph_connection;
	16
31b8006e SW	17	/*
	18	* Ceph defines these callbacks for handling connection events.
	19	*/
	20	struct ceph_connection_operations {
	21	struct ceph_connection (get)(struct ceph_connection *);
	22	void (put)(struct ceph_connection );
	23
	24	/* handle an incoming message. */
	25	void (dispatch) (struct ceph_connection con, struct ceph_msg *m);
	26
4e7a5dcd	27	/* authorize an outgoing connection */
a3530df3 AE	28	struct ceph_auth_handshake (get_authorizer) (
a3530df3 AE	29	struct ceph_connection *con,
8f43fb53	30	int *proto, int force_new);
4e7a5dcd	31	int (verify_authorizer_reply) (struct ceph_connection con, int len);
9bd2e6f8	32	int (invalidate_authorizer)(struct ceph_connection con);
4e7a5dcd	33
31b8006e SW	34	/* there was some error on the socket (disconnect, whatever) */
	35	void (fault) (struct ceph_connection con);
	36
	37	/* a remote host as terminated a message exchange session, and messages
	38	* we sent (or they tried to send us) may be lost. */
	39	void (peer_reset) (struct ceph_connection con);
	40
	41	struct ceph_msg * (alloc_msg) (struct ceph_connection con,
2450418c YS	42	struct ceph_msg_header *hdr,
2450418c YS	43	int *skip);
31b8006e SW	44	};
31b8006e SW	45
31b8006e	46	/* use format string %s%d */
ca9d93a2	47	#define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
31b8006e SW	48
	49	struct ceph_messenger {
	50	struct ceph_entity_inst inst; /* my name+address */
63f2d211	51	struct ceph_entity_addr my_enc_addr;
31b8006e	52
a2a32584	53	atomic_t stopping;
31b8006e SW	54	bool nocrc;
	55
	56	/*
	57	* the global_seq counts connections i (attempt to) initiate
	58	* in order to disambiguate certain connect race conditions.
	59	*/
	60	u32 global_seq;
	61	spinlock_t global_seq_lock;
3d14c5d2 YS	62
	63	u32 supported_features;
	64	u32 required_features;
31b8006e SW	65	};
31b8006e SW	66
43794509 AE	67	enum ceph_msg_data_type {
	68	CEPH_MSG_DATA_NONE, /* message contains no data payload */
	69	CEPH_MSG_DATA_PAGES, /* data source/destination is a page array */
	70	CEPH_MSG_DATA_PAGELIST, /* data source/destination is a pagelist */
	71	#ifdef CONFIG_BLOCK
	72	CEPH_MSG_DATA_BIO, /* data source/destination is a bio list */
	73	#endif /* CONFIG_BLOCK */
	74	};
	75
	76	static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type)
	77	{
	78	switch (type) {
	79	case CEPH_MSG_DATA_NONE:
	80	case CEPH_MSG_DATA_PAGES:
	81	case CEPH_MSG_DATA_PAGELIST:
	82	#ifdef CONFIG_BLOCK
	83	case CEPH_MSG_DATA_BIO:
	84	#endif /* CONFIG_BLOCK */
	85	return true;
	86	default:
	87	return false;
	88	}
	89	}
	90
36153ec9 AE	91	struct ceph_msg_data {
	92	enum ceph_msg_data_type type;
	93	union {
	94	#ifdef CONFIG_BLOCK
	95	struct {
	96	struct bio *bio;
	97	size_t bio_length;
	98	};
	99	#endif /* CONFIG_BLOCK */
	100	struct {
	101	struct page *pages; / NOT OWNER. */
	102	size_t length; /* total # bytes */
	103	unsigned int alignment; /* first page */
	104	};
	105	struct ceph_pagelist *pagelist;
	106	};
36153ec9 AE	107	};
36153ec9 AE	108
fe38a2b6	109	struct ceph_msg_data_cursor {
8ae4f4f5 AE	110	struct ceph_msg_data data; / data item this describes */
	111	size_t resid; /* bytes not yet consumed */
	112	bool last_piece; /* current is last piece */
	113	bool need_crc; /* crc update needed */
dd236fcb	114	union {
6aaa4511 AE	115	#ifdef CONFIG_BLOCK
	116	struct { /* bio */
	117	struct bio bio; / bio from list */
	118	unsigned int vector_index; /* vector from bio */
	119	unsigned int vector_offset; /* bytes from vector */
	120	};
	121	#endif /* CONFIG_BLOCK */
e766d7b5	122	struct { /* pages */
e766d7b5 AE	123	unsigned int page_offset; /* offset in page */
	124	unsigned short page_index; /* index in array */
	125	unsigned short page_count; /* pages in array */
	126	};
dd236fcb AE	127	struct { /* pagelist */
	128	struct page page; / page from list */
	129	size_t offset; /* bytes from list */
	130	};
	131	};
fe38a2b6 AE	132	};
fe38a2b6 AE	133
31b8006e SW	134	/*
	135	* a single message. it contains a header (src, dest, message type, etc.),
	136	* footer (crc values, mainly), a "front" message body, and possibly a
	137	* data payload (stored in some number of pages).
	138	*/
	139	struct ceph_msg {
	140	struct ceph_msg_header hdr; /* header */
	141	struct ceph_msg_footer footer; /* footer */
	142	struct kvec front; /* unaligned blobs of message */
	143	struct ceph_buffer *middle;
38941f80	144
36153ec9 AE	145	size_t data_length;
	146	struct ceph_msg_data *data;
	147	struct ceph_msg_data_cursor cursor;
02afca6c AE	148
	149	struct ceph_connection *con;
	150	struct list_head list_head; /* links for connection lists */
	151
	152	struct kref kref;
31b8006e SW	153	bool front_is_vmalloc;
31b8006e SW	154	bool more_to_follow;
e84346b7	155	bool needs_out_seq;
31b8006e	156	int front_max;
4cf9d544	157	unsigned long ack_stamp; /* tx: when we were acked */
31b8006e SW	158
	159	struct ceph_msgpool *pool;
	160	};
	161
31b8006e SW	162	/* ceph connection fault delay defaults, for exponential backoff */
	163	#define BASE_DELAY_INTERVAL (HZ/2)
	164	#define MAX_DELAY_INTERVAL (5 * 60 * HZ)
	165
31b8006e SW	166	/*
	167	* A single connection with another host.
	168	*
	169	* We maintain a queue of outgoing messages, and some session state to
	170	* ensure that we can preserve the lossless, ordered delivery of
	171	* messages in the case of a TCP disconnect.
	172	*/
	173	struct ceph_connection {
	174	void *private;
31b8006e SW	175
	176	const struct ceph_connection_operations *ops;
	177
	178	struct ceph_messenger *msgr;
ce2c8903 AE	179
ce2c8903 AE	180	atomic_t sock_state;
31b8006e	181	struct socket *sock;
ce2c8903 AE	182	struct ceph_entity_addr peer_addr; /* peer address */
	183	struct ceph_entity_addr peer_addr_for_me;
	184
928443cd AE	185	unsigned long flags;
928443cd AE	186	unsigned long state;
31b8006e SW	187	const char error_msg; / error message, if any */
31b8006e SW	188
31b8006e	189	struct ceph_entity_name peer_name; /* peer name */
ce2c8903	190
aba558e2	191	unsigned peer_features;
31b8006e SW	192	u32 connect_seq; /* identify the most recent connection
	193	attempt for this connection, client */
	194	u32 peer_global_seq; /* peer's global seq for this connection */
	195
4e7a5dcd SW	196	int auth_retry; /* true if we need a newer authorizer */
	197	void auth_reply_buf; / where to put the authorizer reply */
	198	int auth_reply_buf_len;
	199
ec302645 SW	200	struct mutex mutex;
ec302645 SW	201
31b8006e	202	/* out queue */
31b8006e SW	203	struct list_head out_queue;
	204	struct list_head out_sent; /* sending or sent but unacked */
	205	u64 out_seq; /* last message queued for send */
31b8006e SW	206
	207	u64 in_seq, in_seq_acked; /* last message received, acked */
	208
	209	/* connection negotiation temps */
	210	char in_banner[CEPH_BANNER_MAX_LEN];
a16cb1f7 SW	211	struct ceph_msg_connect out_connect;
a16cb1f7 SW	212	struct ceph_msg_connect_reply in_reply;
31b8006e SW	213	struct ceph_entity_addr actual_peer_addr;
	214
	215	/* message out temps */
	216	struct ceph_msg out_msg; / sending message (== tail of
	217	out_sent) */
c86a2930	218	bool out_msg_done;
31b8006e SW	219
	220	struct kvec out_kvec[8], /* sending header/footer data */
	221	*out_kvec_cur;
	222	int out_kvec_left; /* kvec's left in out_kvec */
	223	int out_skip; /* skip this many bytes */
	224	int out_kvec_bytes; /* total bytes left */
	225	bool out_kvec_is_msg; /* kvec refers to out_msg */
	226	int out_more; /* there is more data after the kvecs */
	227	__le64 out_temp_ack; /* for writing an ack */
	228
	229	/* message in temps */
	230	struct ceph_msg_header in_hdr;
	231	struct ceph_msg *in_msg;
31b8006e SW	232	u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
	233
	234	char in_tag; /* protocol control byte */
	235	int in_base_pos; /* bytes read */
	236	__le64 in_temp_ack; /* for reading an ack */
	237
	238	struct delayed_work work; /* send\|recv work */
	239	unsigned long delay; /* current delay interval */
	240	};
	241
	242
3d14c5d2	243	extern const char ceph_pr_addr(const struct sockaddr_storage ss);
31b8006e SW	244	extern int ceph_parse_ips(const char c, const char end,
	245	struct ceph_entity_addr *addr,
	246	int max_count, int *count);
	247
	248
	249	extern int ceph_msgr_init(void);
	250	extern void ceph_msgr_exit(void);
a922d38f	251	extern void ceph_msgr_flush(void);
31b8006e	252
15d9882c AE	253	extern void ceph_messenger_init(struct ceph_messenger *msgr,
	254	struct ceph_entity_addr *myaddr,
	255	u32 supported_features,
	256	u32 required_features,
	257	bool nocrc);
31b8006e	258
1bfd89f4 AE	259	extern void ceph_con_init(struct ceph_connection con, void private,
1bfd89f4 AE	260	const struct ceph_connection_operations *ops,
b7a9e5dd	261	struct ceph_messenger *msgr);
31b8006e	262	extern void ceph_con_open(struct ceph_connection *con,
b7a9e5dd	263	__u8 entity_type, __u64 entity_num,
31b8006e	264	struct ceph_entity_addr *addr);
87b315a5	265	extern bool ceph_con_opened(struct ceph_connection *con);
31b8006e SW	266	extern void ceph_con_close(struct ceph_connection *con);
31b8006e SW	267	extern void ceph_con_send(struct ceph_connection con, struct ceph_msg msg);
6740a845 AE	268
6740a845 AE	269	extern void ceph_msg_revoke(struct ceph_msg *msg);
8921d114 AE	270	extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
8921d114 AE	271
31b8006e	272	extern void ceph_con_keepalive(struct ceph_connection *con);
31b8006e	273
02afca6c	274	extern void ceph_msg_data_set_pages(struct ceph_msg msg, struct page *pages,
f1baeb2b	275	size_t length, size_t alignment);
27fa8385 AE	276	extern void ceph_msg_data_set_pagelist(struct ceph_msg *msg,
27fa8385 AE	277	struct ceph_pagelist *pagelist);
ea96571f	278	#ifdef CONFIG_BLOCK
a1930804 AE	279	extern void ceph_msg_data_set_bio(struct ceph_msg msg, struct bio bio,
a1930804 AE	280	size_t length);
ea96571f	281	#endif /* CONFIG_BLOCK */
02afca6c	282
b61c2763 SW	283	extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
b61c2763 SW	284	bool can_fail);
31b8006e SW	285	extern void ceph_msg_kfree(struct ceph_msg *m);
31b8006e SW	286
31b8006e SW	287
	288	static inline struct ceph_msg ceph_msg_get(struct ceph_msg msg)
	289	{
c2e552e7	290	kref_get(&msg->kref);
31b8006e SW	291	return msg;
31b8006e SW	292	}
c2e552e7 SW	293	extern void ceph_msg_last_put(struct kref *kref);
	294	static inline void ceph_msg_put(struct ceph_msg *msg)
	295	{
	296	kref_put(&msg->kref, ceph_msg_last_put);
	297	}
31b8006e	298
9ec7cab1 SW	299	extern void ceph_msg_dump(struct ceph_msg *msg);
9ec7cab1 SW	300
31b8006e	301	#endif