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

#ifndef __FS_CEPH_MESSENGER_H
#define __FS_CEPH_MESSENGER_H

#include <linux/blk_types.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 <net/net_namespace.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);

	int (*sign_message) (struct ceph_msg *msg);
	int (*check_message_signature) (struct ceph_msg *msg);
};

/* 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;
	possible_net_t net;

	/*
	 * 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;
};

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 {
	struct list_head		links;	/* 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 {
	size_t			total_resid;	/* across all data items */
	struct list_head	*data_head;	/* = &ceph_msg->data */

	struct ceph_msg_data	*data;		/* current data item */
	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 */
			struct bvec_iter bvec_iter;
		};
#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 */
	union {
		struct ceph_msg_footer footer;		/* footer */
		struct ceph_msg_footer_old old_footer;	/* old format footer */
	};
	struct kvec front;              /* unaligned blobs of message */
	struct ceph_buffer *middle;

	size_t				data_length;
	struct list_head		data;
	struct ceph_msg_data_cursor	cursor;

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

	struct kref kref;
	bool more_to_follow;
	bool needs_out_seq;
	int front_alloc_len;
	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 */

	u64 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 */
	struct ceph_timespec out_temp_keepalive2; /* for writing keepalive2
						     stamp */

	/* 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 timespec last_keepalive_ack; /* keepalive2 ack stamp */

	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);
extern void ceph_messenger_fini(struct ceph_messenger *msgr);

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 bool ceph_con_keepalive_expired(struct ceph_connection *con,
				       unsigned long interval);

extern void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
				size_t length, size_t alignment);
extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
				struct ceph_pagelist *pagelist);
#ifdef CONFIG_BLOCK
extern void ceph_msg_data_add_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 struct ceph_msg *ceph_msg_get(struct ceph_msg *msg);
extern void ceph_msg_put(struct ceph_msg *msg);

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
f38a5181	4	#include <linux/blk_types.h>
c2e552e7	5	#include <linux/kref.h>
31b8006e SW	6	#include <linux/mutex.h>
	7	#include <linux/net.h>
	8	#include <linux/radix-tree.h>
	9	#include <linux/uio.h>
31b8006e	10	#include <linux/workqueue.h>
757856d2	11	#include <net/net_namespace.h>
31b8006e	12
a1ce3928 DH	13	#include <linux/ceph/types.h>
a1ce3928 DH	14	#include <linux/ceph/buffer.h>
31b8006e SW	15
	16	struct ceph_msg;
	17	struct ceph_connection;
	18
31b8006e SW	19	/*
	20	* Ceph defines these callbacks for handling connection events.
	21	*/
	22	struct ceph_connection_operations {
	23	struct ceph_connection (get)(struct ceph_connection *);
	24	void (put)(struct ceph_connection );
	25
	26	/* handle an incoming message. */
	27	void (dispatch) (struct ceph_connection con, struct ceph_msg *m);
	28
4e7a5dcd	29	/* authorize an outgoing connection */
a3530df3 AE	30	struct ceph_auth_handshake (get_authorizer) (
a3530df3 AE	31	struct ceph_connection *con,
8f43fb53	32	int *proto, int force_new);
4e7a5dcd	33	int (verify_authorizer_reply) (struct ceph_connection con, int len);
9bd2e6f8	34	int (invalidate_authorizer)(struct ceph_connection con);
4e7a5dcd	35
31b8006e SW	36	/* there was some error on the socket (disconnect, whatever) */
	37	void (fault) (struct ceph_connection con);
	38
	39	/* a remote host as terminated a message exchange session, and messages
	40	* we sent (or they tried to send us) may be lost. */
	41	void (peer_reset) (struct ceph_connection con);
	42
	43	struct ceph_msg * (alloc_msg) (struct ceph_connection con,
2450418c YS	44	struct ceph_msg_header *hdr,
2450418c YS	45	int *skip);
33d07337	46
79dbd1ba ID	47	int (sign_message) (struct ceph_msg msg);
79dbd1ba ID	48	int (check_message_signature) (struct ceph_msg msg);
31b8006e SW	49	};
31b8006e SW	50
31b8006e	51	/* use format string %s%d */
ca9d93a2	52	#define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
31b8006e SW	53
	54	struct ceph_messenger {
	55	struct ceph_entity_inst inst; /* my name+address */
63f2d211	56	struct ceph_entity_addr my_enc_addr;
31b8006e	57
a2a32584	58	atomic_t stopping;
757856d2	59	possible_net_t net;
31b8006e SW	60
	61	/*
	62	* the global_seq counts connections i (attempt to) initiate
	63	* in order to disambiguate certain connect race conditions.
	64	*/
	65	u32 global_seq;
	66	spinlock_t global_seq_lock;
	67	};
	68
43794509 AE	69	enum ceph_msg_data_type {
	70	CEPH_MSG_DATA_NONE, /* message contains no data payload */
	71	CEPH_MSG_DATA_PAGES, /* data source/destination is a page array */
	72	CEPH_MSG_DATA_PAGELIST, /* data source/destination is a pagelist */
	73	#ifdef CONFIG_BLOCK
	74	CEPH_MSG_DATA_BIO, /* data source/destination is a bio list */
	75	#endif /* CONFIG_BLOCK */
	76	};
	77
	78	static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type)
	79	{
	80	switch (type) {
	81	case CEPH_MSG_DATA_NONE:
	82	case CEPH_MSG_DATA_PAGES:
	83	case CEPH_MSG_DATA_PAGELIST:
	84	#ifdef CONFIG_BLOCK
	85	case CEPH_MSG_DATA_BIO:
	86	#endif /* CONFIG_BLOCK */
	87	return true;
	88	default:
	89	return false;
	90	}
	91	}
	92
36153ec9	93	struct ceph_msg_data {
5240d9f9	94	struct list_head links; /* ceph_msg->data */
36153ec9 AE	95	enum ceph_msg_data_type type;
	96	union {
	97	#ifdef CONFIG_BLOCK
	98	struct {
	99	struct bio *bio;
	100	size_t bio_length;
	101	};
	102	#endif /* CONFIG_BLOCK */
	103	struct {
	104	struct page *pages; / NOT OWNER. */
	105	size_t length; /* total # bytes */
	106	unsigned int alignment; /* first page */
	107	};
	108	struct ceph_pagelist *pagelist;
	109	};
36153ec9 AE	110	};
36153ec9 AE	111
fe38a2b6	112	struct ceph_msg_data_cursor {
ca8b3a69 AE	113	size_t total_resid; /* across all data items */
	114	struct list_head data_head; / = &ceph_msg->data */
	115
	116	struct ceph_msg_data data; / current data item */
8ae4f4f5 AE	117	size_t resid; /* bytes not yet consumed */
	118	bool last_piece; /* current is last piece */
	119	bool need_crc; /* crc update needed */
dd236fcb	120	union {
6aaa4511 AE	121	#ifdef CONFIG_BLOCK
	122	struct { /* bio */
	123	struct bio bio; / bio from list */
f38a5181	124	struct bvec_iter bvec_iter;
6aaa4511 AE	125	};
6aaa4511 AE	126	#endif /* CONFIG_BLOCK */
e766d7b5	127	struct { /* pages */
e766d7b5 AE	128	unsigned int page_offset; /* offset in page */
	129	unsigned short page_index; /* index in array */
	130	unsigned short page_count; /* pages in array */
	131	};
dd236fcb AE	132	struct { /* pagelist */
	133	struct page page; / page from list */
	134	size_t offset; /* bytes from list */
	135	};
	136	};
fe38a2b6 AE	137	};
fe38a2b6 AE	138
31b8006e SW	139	/*
	140	* a single message. it contains a header (src, dest, message type, etc.),
	141	* footer (crc values, mainly), a "front" message body, and possibly a
	142	* data payload (stored in some number of pages).
	143	*/
	144	struct ceph_msg {
	145	struct ceph_msg_header hdr; /* header */
33d07337 YZ	146	union {
	147	struct ceph_msg_footer footer; /* footer */
	148	struct ceph_msg_footer_old old_footer; /* old format footer */
	149	};
31b8006e SW	150	struct kvec front; /* unaligned blobs of message */
31b8006e SW	151	struct ceph_buffer *middle;
38941f80	152
36153ec9	153	size_t data_length;
5240d9f9	154	struct list_head data;
36153ec9	155	struct ceph_msg_data_cursor cursor;
02afca6c AE	156
	157	struct ceph_connection *con;
	158	struct list_head list_head; /* links for connection lists */
	159
	160	struct kref kref;
31b8006e	161	bool more_to_follow;
e84346b7	162	bool needs_out_seq;
3cea4c30	163	int front_alloc_len;
4cf9d544	164	unsigned long ack_stamp; /* tx: when we were acked */
31b8006e SW	165
	166	struct ceph_msgpool *pool;
	167	};
	168
31b8006e SW	169	/* ceph connection fault delay defaults, for exponential backoff */
	170	#define BASE_DELAY_INTERVAL (HZ/2)
	171	#define MAX_DELAY_INTERVAL (5 * 60 * HZ)
	172
31b8006e SW	173	/*
	174	* A single connection with another host.
	175	*
	176	* We maintain a queue of outgoing messages, and some session state to
	177	* ensure that we can preserve the lossless, ordered delivery of
	178	* messages in the case of a TCP disconnect.
	179	*/
	180	struct ceph_connection {
	181	void *private;
31b8006e SW	182
	183	const struct ceph_connection_operations *ops;
	184
	185	struct ceph_messenger *msgr;
ce2c8903 AE	186
ce2c8903 AE	187	atomic_t sock_state;
31b8006e	188	struct socket *sock;
ce2c8903 AE	189	struct ceph_entity_addr peer_addr; /* peer address */
	190	struct ceph_entity_addr peer_addr_for_me;
	191
928443cd AE	192	unsigned long flags;
928443cd AE	193	unsigned long state;
31b8006e SW	194	const char error_msg; / error message, if any */
31b8006e SW	195
31b8006e	196	struct ceph_entity_name peer_name; /* peer name */
ce2c8903	197
12b4629a	198	u64 peer_features;
31b8006e SW	199	u32 connect_seq; /* identify the most recent connection
	200	attempt for this connection, client */
	201	u32 peer_global_seq; /* peer's global seq for this connection */
	202
4e7a5dcd SW	203	int auth_retry; /* true if we need a newer authorizer */
	204	void auth_reply_buf; / where to put the authorizer reply */
	205	int auth_reply_buf_len;
	206
ec302645 SW	207	struct mutex mutex;
ec302645 SW	208
31b8006e	209	/* out queue */
31b8006e SW	210	struct list_head out_queue;
	211	struct list_head out_sent; /* sending or sent but unacked */
	212	u64 out_seq; /* last message queued for send */
31b8006e SW	213
	214	u64 in_seq, in_seq_acked; /* last message received, acked */
	215
	216	/* connection negotiation temps */
	217	char in_banner[CEPH_BANNER_MAX_LEN];
a16cb1f7 SW	218	struct ceph_msg_connect out_connect;
a16cb1f7 SW	219	struct ceph_msg_connect_reply in_reply;
31b8006e SW	220	struct ceph_entity_addr actual_peer_addr;
	221
	222	/* message out temps */
	223	struct ceph_msg out_msg; / sending message (== tail of
	224	out_sent) */
c86a2930	225	bool out_msg_done;
31b8006e SW	226
	227	struct kvec out_kvec[8], /* sending header/footer data */
	228	*out_kvec_cur;
	229	int out_kvec_left; /* kvec's left in out_kvec */
	230	int out_skip; /* skip this many bytes */
	231	int out_kvec_bytes; /* total bytes left */
	232	bool out_kvec_is_msg; /* kvec refers to out_msg */
	233	int out_more; /* there is more data after the kvecs */
	234	__le64 out_temp_ack; /* for writing an ack */
7f61f545 ID	235	struct ceph_timespec out_temp_keepalive2; /* for writing keepalive2
7f61f545 ID	236	stamp */
31b8006e SW	237
	238	/* message in temps */
	239	struct ceph_msg_header in_hdr;
	240	struct ceph_msg *in_msg;
31b8006e SW	241	u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
	242
	243	char in_tag; /* protocol control byte */
	244	int in_base_pos; /* bytes read */
	245	__le64 in_temp_ack; /* for reading an ack */
	246
7f61f545	247	struct timespec last_keepalive_ack; /* keepalive2 ack stamp */
8b9558aa	248
31b8006e SW	249	struct delayed_work work; /* send\|recv work */
	250	unsigned long delay; /* current delay interval */
	251	};
	252
	253
3d14c5d2	254	extern const char ceph_pr_addr(const struct sockaddr_storage ss);
31b8006e SW	255	extern int ceph_parse_ips(const char c, const char end,
	256	struct ceph_entity_addr *addr,
	257	int max_count, int *count);
	258
	259
	260	extern int ceph_msgr_init(void);
	261	extern void ceph_msgr_exit(void);
a922d38f	262	extern void ceph_msgr_flush(void);
31b8006e	263
15d9882c	264	extern void ceph_messenger_init(struct ceph_messenger *msgr,
859bff51	265	struct ceph_entity_addr *myaddr);
757856d2	266	extern void ceph_messenger_fini(struct ceph_messenger *msgr);
31b8006e	267
1bfd89f4 AE	268	extern void ceph_con_init(struct ceph_connection con, void private,
1bfd89f4 AE	269	const struct ceph_connection_operations *ops,
b7a9e5dd	270	struct ceph_messenger *msgr);
31b8006e	271	extern void ceph_con_open(struct ceph_connection *con,
b7a9e5dd	272	__u8 entity_type, __u64 entity_num,
31b8006e	273	struct ceph_entity_addr *addr);
87b315a5	274	extern bool ceph_con_opened(struct ceph_connection *con);
31b8006e SW	275	extern void ceph_con_close(struct ceph_connection *con);
31b8006e SW	276	extern void ceph_con_send(struct ceph_connection con, struct ceph_msg msg);
6740a845 AE	277
6740a845 AE	278	extern void ceph_msg_revoke(struct ceph_msg *msg);
8921d114 AE	279	extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
8921d114 AE	280
31b8006e	281	extern void ceph_con_keepalive(struct ceph_connection *con);
8b9558aa YZ	282	extern bool ceph_con_keepalive_expired(struct ceph_connection *con,
8b9558aa YZ	283	unsigned long interval);
31b8006e	284
90af3602	285	extern void ceph_msg_data_add_pages(struct ceph_msg msg, struct page *pages,
f1baeb2b	286	size_t length, size_t alignment);
90af3602	287	extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
27fa8385	288	struct ceph_pagelist *pagelist);
ea96571f	289	#ifdef CONFIG_BLOCK
90af3602	290	extern void ceph_msg_data_add_bio(struct ceph_msg msg, struct bio bio,
a1930804	291	size_t length);
ea96571f	292	#endif /* CONFIG_BLOCK */
02afca6c	293
b61c2763 SW	294	extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
b61c2763 SW	295	bool can_fail);
31b8006e	296
0215e44b ID	297	extern struct ceph_msg ceph_msg_get(struct ceph_msg msg);
0215e44b ID	298	extern void ceph_msg_put(struct ceph_msg *msg);
31b8006e	299
9ec7cab1 SW	300	extern void ceph_msg_dump(struct ceph_msg *msg);
9ec7cab1 SW	301
31b8006e	302	#endif