1 #ifndef __FS_CEPH_MESSENGER_H
2 #define __FS_CEPH_MESSENGER_H
4 #include <linux/kref.h>
5 #include <linux/mutex.h>
7 #include <linux/radix-tree.h>
9 #include <linux/workqueue.h>
11 #include <linux/ceph/types.h>
12 #include <linux/ceph/buffer.h>
15 struct ceph_connection;
18 * Ceph defines these callbacks for handling connection events.
20 struct ceph_connection_operations {
21 struct ceph_connection *(*get)(struct ceph_connection *);
22 void (*put)(struct ceph_connection *);
24 /* handle an incoming message. */
25 void (*dispatch) (struct ceph_connection *con, struct ceph_msg *m);
27 /* authorize an outgoing connection */
28 struct ceph_auth_handshake *(*get_authorizer) (
29 struct ceph_connection *con,
30 int *proto, int force_new);
31 int (*verify_authorizer_reply) (struct ceph_connection *con, int len);
32 int (*invalidate_authorizer)(struct ceph_connection *con);
34 /* there was some error on the socket (disconnect, whatever) */
35 void (*fault) (struct ceph_connection *con);
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);
41 struct ceph_msg * (*alloc_msg) (struct ceph_connection *con,
42 struct ceph_msg_header *hdr,
46 /* use format string %s%d */
47 #define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
49 struct ceph_messenger {
50 struct ceph_entity_inst inst; /* my name+address */
51 struct ceph_entity_addr my_enc_addr;
57 * the global_seq counts connections i (attempt to) initiate
58 * in order to disambiguate certain connect race conditions.
61 spinlock_t global_seq_lock;
63 u32 supported_features;
64 u32 required_features;
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 */
72 CEPH_MSG_DATA_BIO, /* data source/destination is a bio list */
73 #endif /* CONFIG_BLOCK */
76 static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type)
79 case CEPH_MSG_DATA_NONE:
80 case CEPH_MSG_DATA_PAGES:
81 case CEPH_MSG_DATA_PAGELIST:
83 case CEPH_MSG_DATA_BIO:
84 #endif /* CONFIG_BLOCK */
91 struct ceph_msg_data_cursor {
92 size_t resid; /* bytes not yet consumed */
93 bool last_piece; /* now at last piece of data item */
94 bool need_crc; /* new piece; crc update needed */
98 struct bio *bio; /* bio from list */
99 unsigned int vector_index; /* vector from bio */
100 unsigned int vector_offset; /* bytes from vector */
102 #endif /* CONFIG_BLOCK */
104 unsigned int page_offset; /* offset in page */
105 unsigned short page_index; /* index in array */
106 unsigned short page_count; /* pages in array */
108 struct { /* pagelist */
109 struct page *page; /* page from list */
110 size_t offset; /* bytes from list */
115 struct ceph_msg_data {
116 enum ceph_msg_data_type type;
123 #endif /* CONFIG_BLOCK */
125 struct page **pages; /* NOT OWNER. */
126 size_t length; /* total # bytes */
127 unsigned int alignment; /* first page */
129 struct ceph_pagelist *pagelist;
131 struct ceph_msg_data_cursor cursor; /* pagelist only */
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).
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;
146 struct ceph_msg_data *data; /* data payload */
148 struct ceph_connection *con;
149 struct list_head list_head; /* links for connection lists */
152 bool front_is_vmalloc;
156 unsigned long ack_stamp; /* tx: when we were acked */
158 struct ceph_msgpool *pool;
161 /* ceph connection fault delay defaults, for exponential backoff */
162 #define BASE_DELAY_INTERVAL (HZ/2)
163 #define MAX_DELAY_INTERVAL (5 * 60 * HZ)
166 * A single connection with another host.
168 * We maintain a queue of outgoing messages, and some session state to
169 * ensure that we can preserve the lossless, ordered delivery of
170 * messages in the case of a TCP disconnect.
172 struct ceph_connection {
175 const struct ceph_connection_operations *ops;
177 struct ceph_messenger *msgr;
181 struct ceph_entity_addr peer_addr; /* peer address */
182 struct ceph_entity_addr peer_addr_for_me;
186 const char *error_msg; /* error message, if any */
188 struct ceph_entity_name peer_name; /* peer name */
190 unsigned peer_features;
191 u32 connect_seq; /* identify the most recent connection
192 attempt for this connection, client */
193 u32 peer_global_seq; /* peer's global seq for this connection */
195 int auth_retry; /* true if we need a newer authorizer */
196 void *auth_reply_buf; /* where to put the authorizer reply */
197 int auth_reply_buf_len;
202 struct list_head out_queue;
203 struct list_head out_sent; /* sending or sent but unacked */
204 u64 out_seq; /* last message queued for send */
206 u64 in_seq, in_seq_acked; /* last message received, acked */
208 /* connection negotiation temps */
209 char in_banner[CEPH_BANNER_MAX_LEN];
210 struct ceph_msg_connect out_connect;
211 struct ceph_msg_connect_reply in_reply;
212 struct ceph_entity_addr actual_peer_addr;
214 /* message out temps */
215 struct ceph_msg *out_msg; /* sending message (== tail of
219 struct kvec out_kvec[8], /* sending header/footer data */
221 int out_kvec_left; /* kvec's left in out_kvec */
222 int out_skip; /* skip this many bytes */
223 int out_kvec_bytes; /* total bytes left */
224 bool out_kvec_is_msg; /* kvec refers to out_msg */
225 int out_more; /* there is more data after the kvecs */
226 __le64 out_temp_ack; /* for writing an ack */
228 /* message in temps */
229 struct ceph_msg_header in_hdr;
230 struct ceph_msg *in_msg;
231 u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
233 char in_tag; /* protocol control byte */
234 int in_base_pos; /* bytes read */
235 __le64 in_temp_ack; /* for reading an ack */
237 struct delayed_work work; /* send|recv work */
238 unsigned long delay; /* current delay interval */
242 extern const char *ceph_pr_addr(const struct sockaddr_storage *ss);
243 extern int ceph_parse_ips(const char *c, const char *end,
244 struct ceph_entity_addr *addr,
245 int max_count, int *count);
248 extern int ceph_msgr_init(void);
249 extern void ceph_msgr_exit(void);
250 extern void ceph_msgr_flush(void);
252 extern void ceph_messenger_init(struct ceph_messenger *msgr,
253 struct ceph_entity_addr *myaddr,
254 u32 supported_features,
255 u32 required_features,
258 extern void ceph_con_init(struct ceph_connection *con, void *private,
259 const struct ceph_connection_operations *ops,
260 struct ceph_messenger *msgr);
261 extern void ceph_con_open(struct ceph_connection *con,
262 __u8 entity_type, __u64 entity_num,
263 struct ceph_entity_addr *addr);
264 extern bool ceph_con_opened(struct ceph_connection *con);
265 extern void ceph_con_close(struct ceph_connection *con);
266 extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);
268 extern void ceph_msg_revoke(struct ceph_msg *msg);
269 extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
271 extern void ceph_con_keepalive(struct ceph_connection *con);
273 extern void ceph_msg_data_set_pages(struct ceph_msg *msg, struct page **pages,
274 size_t length, size_t alignment);
275 extern void ceph_msg_data_set_pagelist(struct ceph_msg *msg,
276 struct ceph_pagelist *pagelist);
278 extern void ceph_msg_data_set_bio(struct ceph_msg *msg, struct bio *bio,
280 #endif /* CONFIG_BLOCK */
282 extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
284 extern void ceph_msg_kfree(struct ceph_msg *m);
287 static inline struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
289 kref_get(&msg->kref);
292 extern void ceph_msg_last_put(struct kref *kref);
293 static inline void ceph_msg_put(struct ceph_msg *msg)
295 kref_put(&msg->kref, ceph_msg_last_put);
298 extern void ceph_msg_dump(struct ceph_msg *msg);