xref: /linux-6.15/include/linux/ceph/messenger.h (revision bb66fc67)
1 #ifndef __FS_CEPH_MESSENGER_H
2 #define __FS_CEPH_MESSENGER_H
3 
4 #include <linux/blk_types.h>
5 #include <linux/kref.h>
6 #include <linux/mutex.h>
7 #include <linux/net.h>
8 #include <linux/radix-tree.h>
9 #include <linux/uio.h>
10 #include <linux/workqueue.h>
11 
12 #include <linux/ceph/types.h>
13 #include <linux/ceph/buffer.h>
14 
15 struct ceph_msg;
16 struct ceph_connection;
17 
18 /*
19  * Ceph defines these callbacks for handling connection events.
20  */
21 struct ceph_connection_operations {
22 	struct ceph_connection *(*get)(struct ceph_connection *);
23 	void (*put)(struct ceph_connection *);
24 
25 	/* handle an incoming message. */
26 	void (*dispatch) (struct ceph_connection *con, struct ceph_msg *m);
27 
28 	/* authorize an outgoing connection */
29 	struct ceph_auth_handshake *(*get_authorizer) (
30 				struct ceph_connection *con,
31 			       int *proto, int force_new);
32 	int (*verify_authorizer_reply) (struct ceph_connection *con, int len);
33 	int (*invalidate_authorizer)(struct ceph_connection *con);
34 
35 	/* there was some error on the socket (disconnect, whatever) */
36 	void (*fault) (struct ceph_connection *con);
37 
38 	/* a remote host as terminated a message exchange session, and messages
39 	 * we sent (or they tried to send us) may be lost. */
40 	void (*peer_reset) (struct ceph_connection *con);
41 
42 	struct ceph_msg * (*alloc_msg) (struct ceph_connection *con,
43 					struct ceph_msg_header *hdr,
44 					int *skip);
45 };
46 
47 /* use format string %s%d */
48 #define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
49 
50 struct ceph_messenger {
51 	struct ceph_entity_inst inst;    /* my name+address */
52 	struct ceph_entity_addr my_enc_addr;
53 
54 	atomic_t stopping;
55 	bool nocrc;
56 
57 	/*
58 	 * the global_seq counts connections i (attempt to) initiate
59 	 * in order to disambiguate certain connect race conditions.
60 	 */
61 	u32 global_seq;
62 	spinlock_t global_seq_lock;
63 
64 	u64 supported_features;
65 	u64 required_features;
66 };
67 
68 enum ceph_msg_data_type {
69 	CEPH_MSG_DATA_NONE,	/* message contains no data payload */
70 	CEPH_MSG_DATA_PAGES,	/* data source/destination is a page array */
71 	CEPH_MSG_DATA_PAGELIST,	/* data source/destination is a pagelist */
72 #ifdef CONFIG_BLOCK
73 	CEPH_MSG_DATA_BIO,	/* data source/destination is a bio list */
74 #endif /* CONFIG_BLOCK */
75 };
76 
77 static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type)
78 {
79 	switch (type) {
80 	case CEPH_MSG_DATA_NONE:
81 	case CEPH_MSG_DATA_PAGES:
82 	case CEPH_MSG_DATA_PAGELIST:
83 #ifdef CONFIG_BLOCK
84 	case CEPH_MSG_DATA_BIO:
85 #endif /* CONFIG_BLOCK */
86 		return true;
87 	default:
88 		return false;
89 	}
90 }
91 
92 struct ceph_msg_data {
93 	struct list_head		links;	/* ceph_msg->data */
94 	enum ceph_msg_data_type		type;
95 	union {
96 #ifdef CONFIG_BLOCK
97 		struct {
98 			struct bio	*bio;
99 			size_t		bio_length;
100 		};
101 #endif /* CONFIG_BLOCK */
102 		struct {
103 			struct page	**pages;	/* NOT OWNER. */
104 			size_t		length;		/* total # bytes */
105 			unsigned int	alignment;	/* first page */
106 		};
107 		struct ceph_pagelist	*pagelist;
108 	};
109 };
110 
111 struct ceph_msg_data_cursor {
112 	size_t			total_resid;	/* across all data items */
113 	struct list_head	*data_head;	/* = &ceph_msg->data */
114 
115 	struct ceph_msg_data	*data;		/* current data item */
116 	size_t			resid;		/* bytes not yet consumed */
117 	bool			last_piece;	/* current is last piece */
118 	bool			need_crc;	/* crc update needed */
119 	union {
120 #ifdef CONFIG_BLOCK
121 		struct {				/* bio */
122 			struct bio	*bio;		/* bio from list */
123 			struct bvec_iter bvec_iter;
124 		};
125 #endif /* CONFIG_BLOCK */
126 		struct {				/* pages */
127 			unsigned int	page_offset;	/* offset in page */
128 			unsigned short	page_index;	/* index in array */
129 			unsigned short	page_count;	/* pages in array */
130 		};
131 		struct {				/* pagelist */
132 			struct page	*page;		/* page from list */
133 			size_t		offset;		/* bytes from list */
134 		};
135 	};
136 };
137 
138 /*
139  * a single message.  it contains a header (src, dest, message type, etc.),
140  * footer (crc values, mainly), a "front" message body, and possibly a
141  * data payload (stored in some number of pages).
142  */
143 struct ceph_msg {
144 	struct ceph_msg_header hdr;	/* header */
145 	struct ceph_msg_footer footer;	/* footer */
146 	struct kvec front;              /* unaligned blobs of message */
147 	struct ceph_buffer *middle;
148 
149 	size_t				data_length;
150 	struct list_head		data;
151 	struct ceph_msg_data_cursor	cursor;
152 
153 	struct ceph_connection *con;
154 	struct list_head list_head;	/* links for connection lists */
155 
156 	struct kref kref;
157 	bool more_to_follow;
158 	bool needs_out_seq;
159 	int front_alloc_len;
160 	unsigned long ack_stamp;        /* tx: when we were acked */
161 
162 	struct ceph_msgpool *pool;
163 };
164 
165 /* ceph connection fault delay defaults, for exponential backoff */
166 #define BASE_DELAY_INTERVAL	(HZ/2)
167 #define MAX_DELAY_INTERVAL	(5 * 60 * HZ)
168 
169 /*
170  * A single connection with another host.
171  *
172  * We maintain a queue of outgoing messages, and some session state to
173  * ensure that we can preserve the lossless, ordered delivery of
174  * messages in the case of a TCP disconnect.
175  */
176 struct ceph_connection {
177 	void *private;
178 
179 	const struct ceph_connection_operations *ops;
180 
181 	struct ceph_messenger *msgr;
182 
183 	atomic_t sock_state;
184 	struct socket *sock;
185 	struct ceph_entity_addr peer_addr; /* peer address */
186 	struct ceph_entity_addr peer_addr_for_me;
187 
188 	unsigned long flags;
189 	unsigned long state;
190 	const char *error_msg;  /* error message, if any */
191 
192 	struct ceph_entity_name peer_name; /* peer name */
193 
194 	u64 peer_features;
195 	u32 connect_seq;      /* identify the most recent connection
196 				 attempt for this connection, client */
197 	u32 peer_global_seq;  /* peer's global seq for this connection */
198 
199 	int auth_retry;       /* true if we need a newer authorizer */
200 	void *auth_reply_buf;   /* where to put the authorizer reply */
201 	int auth_reply_buf_len;
202 
203 	struct mutex mutex;
204 
205 	/* out queue */
206 	struct list_head out_queue;
207 	struct list_head out_sent;   /* sending or sent but unacked */
208 	u64 out_seq;		     /* last message queued for send */
209 
210 	u64 in_seq, in_seq_acked;  /* last message received, acked */
211 
212 	/* connection negotiation temps */
213 	char in_banner[CEPH_BANNER_MAX_LEN];
214 	struct ceph_msg_connect out_connect;
215 	struct ceph_msg_connect_reply in_reply;
216 	struct ceph_entity_addr actual_peer_addr;
217 
218 	/* message out temps */
219 	struct ceph_msg *out_msg;        /* sending message (== tail of
220 					    out_sent) */
221 	bool out_msg_done;
222 
223 	struct kvec out_kvec[8],         /* sending header/footer data */
224 		*out_kvec_cur;
225 	int out_kvec_left;   /* kvec's left in out_kvec */
226 	int out_skip;        /* skip this many bytes */
227 	int out_kvec_bytes;  /* total bytes left */
228 	bool out_kvec_is_msg; /* kvec refers to out_msg */
229 	int out_more;        /* there is more data after the kvecs */
230 	__le64 out_temp_ack; /* for writing an ack */
231 
232 	/* message in temps */
233 	struct ceph_msg_header in_hdr;
234 	struct ceph_msg *in_msg;
235 	u32 in_front_crc, in_middle_crc, in_data_crc;  /* calculated crc */
236 
237 	char in_tag;         /* protocol control byte */
238 	int in_base_pos;     /* bytes read */
239 	__le64 in_temp_ack;  /* for reading an ack */
240 
241 	struct delayed_work work;	    /* send|recv work */
242 	unsigned long       delay;          /* current delay interval */
243 };
244 
245 
246 extern const char *ceph_pr_addr(const struct sockaddr_storage *ss);
247 extern int ceph_parse_ips(const char *c, const char *end,
248 			  struct ceph_entity_addr *addr,
249 			  int max_count, int *count);
250 
251 
252 extern int ceph_msgr_init(void);
253 extern void ceph_msgr_exit(void);
254 extern void ceph_msgr_flush(void);
255 
256 extern void ceph_messenger_init(struct ceph_messenger *msgr,
257 			struct ceph_entity_addr *myaddr,
258 			u64 supported_features,
259 			u64 required_features,
260 			bool nocrc);
261 
262 extern void ceph_con_init(struct ceph_connection *con, void *private,
263 			const struct ceph_connection_operations *ops,
264 			struct ceph_messenger *msgr);
265 extern void ceph_con_open(struct ceph_connection *con,
266 			  __u8 entity_type, __u64 entity_num,
267 			  struct ceph_entity_addr *addr);
268 extern bool ceph_con_opened(struct ceph_connection *con);
269 extern void ceph_con_close(struct ceph_connection *con);
270 extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);
271 
272 extern void ceph_msg_revoke(struct ceph_msg *msg);
273 extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
274 
275 extern void ceph_con_keepalive(struct ceph_connection *con);
276 
277 extern void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
278 				size_t length, size_t alignment);
279 extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
280 				struct ceph_pagelist *pagelist);
281 #ifdef CONFIG_BLOCK
282 extern void ceph_msg_data_add_bio(struct ceph_msg *msg, struct bio *bio,
283 				size_t length);
284 #endif /* CONFIG_BLOCK */
285 
286 extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
287 				     bool can_fail);
288 extern void ceph_msg_kfree(struct ceph_msg *m);
289 
290 
291 static inline struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
292 {
293 	kref_get(&msg->kref);
294 	return msg;
295 }
296 extern void ceph_msg_last_put(struct kref *kref);
297 static inline void ceph_msg_put(struct ceph_msg *msg)
298 {
299 	kref_put(&msg->kref, ceph_msg_last_put);
300 }
301 
302 extern void ceph_msg_dump(struct ceph_msg *msg);
303 
304 #endif
305