[thirdparty/linux.git] / include / linux / ceph / messenger.h

/* SPDX-License-Identifier: GPL-2.0 */
#ifndef __FS_CEPH_MESSENGER_H
#define __FS_CEPH_MESSENGER_H

#include <linux/bvec.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 (*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);

	void (*reencode_message) (struct ceph_msg *msg);

	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_header out_hdr;
	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 */
	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
b2441318	1	/* SPDX-License-Identifier: GPL-2.0 */
31b8006e SW	2	#ifndef __FS_CEPH_MESSENGER_H
	3	#define __FS_CEPH_MESSENGER_H
	4
9f082171	5	#include <linux/bvec.h>
c2e552e7	6	#include <linux/kref.h>
31b8006e SW	7	#include <linux/mutex.h>
	8	#include <linux/net.h>
	9	#include <linux/radix-tree.h>
	10	#include <linux/uio.h>
31b8006e	11	#include <linux/workqueue.h>
757856d2	12	#include <net/net_namespace.h>
31b8006e	13
a1ce3928 DH	14	#include <linux/ceph/types.h>
a1ce3928 DH	15	#include <linux/ceph/buffer.h>
31b8006e SW	16
	17	struct ceph_msg;
	18	struct ceph_connection;
	19
31b8006e SW	20	/*
	21	* Ceph defines these callbacks for handling connection events.
	22	*/
	23	struct ceph_connection_operations {
	24	struct ceph_connection (get)(struct ceph_connection *);
	25	void (put)(struct ceph_connection );
	26
	27	/* handle an incoming message. */
	28	void (dispatch) (struct ceph_connection con, struct ceph_msg *m);
	29
4e7a5dcd	30	/* authorize an outgoing connection */
a3530df3 AE	31	struct ceph_auth_handshake (get_authorizer) (
a3530df3 AE	32	struct ceph_connection *con,
8f43fb53	33	int *proto, int force_new);
0dde5848	34	int (verify_authorizer_reply) (struct ceph_connection con);
9bd2e6f8	35	int (invalidate_authorizer)(struct ceph_connection con);
4e7a5dcd	36
31b8006e SW	37	/* there was some error on the socket (disconnect, whatever) */
	38	void (fault) (struct ceph_connection con);
	39
	40	/* a remote host as terminated a message exchange session, and messages
	41	* we sent (or they tried to send us) may be lost. */
	42	void (peer_reset) (struct ceph_connection con);
	43
	44	struct ceph_msg * (alloc_msg) (struct ceph_connection con,
2450418c YS	45	struct ceph_msg_header *hdr,
2450418c YS	46	int *skip);
33d07337	47
98ad5ebd ID	48	void (reencode_message) (struct ceph_msg msg);
98ad5ebd ID	49
79dbd1ba ID	50	int (sign_message) (struct ceph_msg msg);
79dbd1ba ID	51	int (check_message_signature) (struct ceph_msg msg);
31b8006e SW	52	};
31b8006e SW	53
31b8006e	54	/* use format string %s%d */
ca9d93a2	55	#define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
31b8006e SW	56
	57	struct ceph_messenger {
	58	struct ceph_entity_inst inst; /* my name+address */
63f2d211	59	struct ceph_entity_addr my_enc_addr;
31b8006e	60
a2a32584	61	atomic_t stopping;
757856d2	62	possible_net_t net;
31b8006e SW	63
	64	/*
	65	* the global_seq counts connections i (attempt to) initiate
	66	* in order to disambiguate certain connect race conditions.
	67	*/
	68	u32 global_seq;
	69	spinlock_t global_seq_lock;
	70	};
	71
43794509 AE	72	enum ceph_msg_data_type {
	73	CEPH_MSG_DATA_NONE, /* message contains no data payload */
	74	CEPH_MSG_DATA_PAGES, /* data source/destination is a page array */
	75	CEPH_MSG_DATA_PAGELIST, /* data source/destination is a pagelist */
	76	#ifdef CONFIG_BLOCK
	77	CEPH_MSG_DATA_BIO, /* data source/destination is a bio list */
	78	#endif /* CONFIG_BLOCK */
	79	};
	80
	81	static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type)
	82	{
	83	switch (type) {
	84	case CEPH_MSG_DATA_NONE:
	85	case CEPH_MSG_DATA_PAGES:
	86	case CEPH_MSG_DATA_PAGELIST:
	87	#ifdef CONFIG_BLOCK
	88	case CEPH_MSG_DATA_BIO:
	89	#endif /* CONFIG_BLOCK */
	90	return true;
	91	default:
	92	return false;
	93	}
	94	}
	95
36153ec9	96	struct ceph_msg_data {
5240d9f9	97	struct list_head links; /* ceph_msg->data */
36153ec9 AE	98	enum ceph_msg_data_type type;
	99	union {
	100	#ifdef CONFIG_BLOCK
	101	struct {
	102	struct bio *bio;
	103	size_t bio_length;
	104	};
	105	#endif /* CONFIG_BLOCK */
	106	struct {
	107	struct page *pages; / NOT OWNER. */
	108	size_t length; /* total # bytes */
	109	unsigned int alignment; /* first page */
	110	};
	111	struct ceph_pagelist *pagelist;
	112	};
36153ec9 AE	113	};
36153ec9 AE	114
fe38a2b6	115	struct ceph_msg_data_cursor {
ca8b3a69 AE	116	size_t total_resid; /* across all data items */
	117	struct list_head data_head; / = &ceph_msg->data */
	118
	119	struct ceph_msg_data data; / current data item */
8ae4f4f5 AE	120	size_t resid; /* bytes not yet consumed */
	121	bool last_piece; /* current is last piece */
	122	bool need_crc; /* crc update needed */
dd236fcb	123	union {
6aaa4511 AE	124	#ifdef CONFIG_BLOCK
	125	struct { /* bio */
	126	struct bio bio; / bio from list */
f38a5181	127	struct bvec_iter bvec_iter;
6aaa4511 AE	128	};
6aaa4511 AE	129	#endif /* CONFIG_BLOCK */
e766d7b5	130	struct { /* pages */
e766d7b5 AE	131	unsigned int page_offset; /* offset in page */
	132	unsigned short page_index; /* index in array */
	133	unsigned short page_count; /* pages in array */
	134	};
dd236fcb AE	135	struct { /* pagelist */
	136	struct page page; / page from list */
	137	size_t offset; /* bytes from list */
	138	};
	139	};
fe38a2b6 AE	140	};
fe38a2b6 AE	141
31b8006e SW	142	/*
	143	* a single message. it contains a header (src, dest, message type, etc.),
	144	* footer (crc values, mainly), a "front" message body, and possibly a
	145	* data payload (stored in some number of pages).
	146	*/
	147	struct ceph_msg {
	148	struct ceph_msg_header hdr; /* header */
33d07337 YZ	149	union {
	150	struct ceph_msg_footer footer; /* footer */
	151	struct ceph_msg_footer_old old_footer; /* old format footer */
	152	};
31b8006e SW	153	struct kvec front; /* unaligned blobs of message */
31b8006e SW	154	struct ceph_buffer *middle;
38941f80	155
36153ec9	156	size_t data_length;
5240d9f9	157	struct list_head data;
36153ec9	158	struct ceph_msg_data_cursor cursor;
02afca6c AE	159
	160	struct ceph_connection *con;
	161	struct list_head list_head; /* links for connection lists */
	162
	163	struct kref kref;
31b8006e	164	bool more_to_follow;
e84346b7	165	bool needs_out_seq;
3cea4c30	166	int front_alloc_len;
4cf9d544	167	unsigned long ack_stamp; /* tx: when we were acked */
31b8006e SW	168
	169	struct ceph_msgpool *pool;
	170	};
	171
31b8006e SW	172	/* ceph connection fault delay defaults, for exponential backoff */
	173	#define BASE_DELAY_INTERVAL (HZ/2)
	174	#define MAX_DELAY_INTERVAL (5 * 60 * HZ)
	175
31b8006e SW	176	/*
	177	* A single connection with another host.
	178	*
	179	* We maintain a queue of outgoing messages, and some session state to
	180	* ensure that we can preserve the lossless, ordered delivery of
	181	* messages in the case of a TCP disconnect.
	182	*/
	183	struct ceph_connection {
	184	void *private;
31b8006e SW	185
	186	const struct ceph_connection_operations *ops;
	187
	188	struct ceph_messenger *msgr;
ce2c8903 AE	189
ce2c8903 AE	190	atomic_t sock_state;
31b8006e	191	struct socket *sock;
ce2c8903 AE	192	struct ceph_entity_addr peer_addr; /* peer address */
	193	struct ceph_entity_addr peer_addr_for_me;
	194
928443cd AE	195	unsigned long flags;
928443cd AE	196	unsigned long state;
31b8006e SW	197	const char error_msg; / error message, if any */
31b8006e SW	198
31b8006e	199	struct ceph_entity_name peer_name; /* peer name */
ce2c8903	200
12b4629a	201	u64 peer_features;
31b8006e SW	202	u32 connect_seq; /* identify the most recent connection
	203	attempt for this connection, client */
	204	u32 peer_global_seq; /* peer's global seq for this connection */
	205
4e7a5dcd SW	206	int auth_retry; /* true if we need a newer authorizer */
	207	void auth_reply_buf; / where to put the authorizer reply */
	208	int auth_reply_buf_len;
	209
ec302645 SW	210	struct mutex mutex;
ec302645 SW	211
31b8006e	212	/* out queue */
31b8006e SW	213	struct list_head out_queue;
	214	struct list_head out_sent; /* sending or sent but unacked */
	215	u64 out_seq; /* last message queued for send */
31b8006e SW	216
	217	u64 in_seq, in_seq_acked; /* last message received, acked */
	218
	219	/* connection negotiation temps */
	220	char in_banner[CEPH_BANNER_MAX_LEN];
a16cb1f7 SW	221	struct ceph_msg_connect out_connect;
a16cb1f7 SW	222	struct ceph_msg_connect_reply in_reply;
31b8006e SW	223	struct ceph_entity_addr actual_peer_addr;
	224
	225	/* message out temps */
67645d76	226	struct ceph_msg_header out_hdr;
31b8006e SW	227	struct ceph_msg out_msg; / sending message (== tail of
31b8006e SW	228	out_sent) */
c86a2930	229	bool out_msg_done;
31b8006e SW	230
	231	struct kvec out_kvec[8], /* sending header/footer data */
	232	*out_kvec_cur;
	233	int out_kvec_left; /* kvec's left in out_kvec */
	234	int out_skip; /* skip this many bytes */
	235	int out_kvec_bytes; /* total bytes left */
31b8006e SW	236	int out_more; /* there is more data after the kvecs */
31b8006e SW	237	__le64 out_temp_ack; /* for writing an ack */
7f61f545 ID	238	struct ceph_timespec out_temp_keepalive2; /* for writing keepalive2
7f61f545 ID	239	stamp */
31b8006e SW	240
	241	/* message in temps */
	242	struct ceph_msg_header in_hdr;
	243	struct ceph_msg *in_msg;
31b8006e SW	244	u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
	245
	246	char in_tag; /* protocol control byte */
	247	int in_base_pos; /* bytes read */
	248	__le64 in_temp_ack; /* for reading an ack */
	249
7f61f545	250	struct timespec last_keepalive_ack; /* keepalive2 ack stamp */
8b9558aa	251
31b8006e SW	252	struct delayed_work work; /* send\|recv work */
	253	unsigned long delay; /* current delay interval */
	254	};
	255
	256
3d14c5d2	257	extern const char ceph_pr_addr(const struct sockaddr_storage ss);
31b8006e SW	258	extern int ceph_parse_ips(const char c, const char end,
	259	struct ceph_entity_addr *addr,
	260	int max_count, int *count);
	261
	262
	263	extern int ceph_msgr_init(void);
	264	extern void ceph_msgr_exit(void);
a922d38f	265	extern void ceph_msgr_flush(void);
31b8006e	266
15d9882c	267	extern void ceph_messenger_init(struct ceph_messenger *msgr,
859bff51	268	struct ceph_entity_addr *myaddr);
757856d2	269	extern void ceph_messenger_fini(struct ceph_messenger *msgr);
31b8006e	270
1bfd89f4 AE	271	extern void ceph_con_init(struct ceph_connection con, void private,
1bfd89f4 AE	272	const struct ceph_connection_operations *ops,
b7a9e5dd	273	struct ceph_messenger *msgr);
31b8006e	274	extern void ceph_con_open(struct ceph_connection *con,
b7a9e5dd	275	__u8 entity_type, __u64 entity_num,
31b8006e	276	struct ceph_entity_addr *addr);
87b315a5	277	extern bool ceph_con_opened(struct ceph_connection *con);
31b8006e SW	278	extern void ceph_con_close(struct ceph_connection *con);
31b8006e SW	279	extern void ceph_con_send(struct ceph_connection con, struct ceph_msg msg);
6740a845 AE	280
6740a845 AE	281	extern void ceph_msg_revoke(struct ceph_msg *msg);
8921d114 AE	282	extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
8921d114 AE	283
31b8006e	284	extern void ceph_con_keepalive(struct ceph_connection *con);
8b9558aa YZ	285	extern bool ceph_con_keepalive_expired(struct ceph_connection *con,
8b9558aa YZ	286	unsigned long interval);
31b8006e	287
90af3602	288	extern void ceph_msg_data_add_pages(struct ceph_msg msg, struct page *pages,
f1baeb2b	289	size_t length, size_t alignment);
90af3602	290	extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
27fa8385	291	struct ceph_pagelist *pagelist);
ea96571f	292	#ifdef CONFIG_BLOCK
90af3602	293	extern void ceph_msg_data_add_bio(struct ceph_msg msg, struct bio bio,
a1930804	294	size_t length);
ea96571f	295	#endif /* CONFIG_BLOCK */
02afca6c	296
b61c2763 SW	297	extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
b61c2763 SW	298	bool can_fail);
31b8006e	299
0215e44b ID	300	extern struct ceph_msg ceph_msg_get(struct ceph_msg msg);
0215e44b ID	301	extern void ceph_msg_put(struct ceph_msg *msg);
31b8006e	302
9ec7cab1 SW	303	extern void ceph_msg_dump(struct ceph_msg *msg);
9ec7cab1 SW	304
31b8006e	305	#endif