1 #ifndef __FS_CEPH_MESSENGER_H
2 #define __FS_CEPH_MESSENGER_H
4 #include <linux/mutex.h>
6 #include <linux/radix-tree.h>
8 #include <linux/version.h>
9 #include <linux/workqueue.h>
15 struct ceph_connection;
17 extern struct workqueue_struct *ceph_msgr_wq; /* receive work queue */
20 * Ceph defines these callbacks for handling connection events.
22 struct ceph_connection_operations {
23 struct ceph_connection *(*get)(struct ceph_connection *);
24 void (*put)(struct ceph_connection *);
26 /* handle an incoming message. */
27 void (*dispatch) (struct ceph_connection *con, struct ceph_msg *m);
29 /* authorize an outgoing connection */
30 int (*get_authorizer) (struct ceph_connection *con,
31 void **buf, int *len, int *proto,
32 void **reply_buf, int *reply_len, int force_new);
33 int (*verify_authorizer_reply) (struct ceph_connection *con, int len);
35 /* protocol version mismatch */
36 void (*bad_proto) (struct ceph_connection *con);
38 /* there was some error on the socket (disconnect, whatever) */
39 void (*fault) (struct ceph_connection *con);
41 /* a remote host as terminated a message exchange session, and messages
42 * we sent (or they tried to send us) may be lost. */
43 void (*peer_reset) (struct ceph_connection *con);
45 struct ceph_msg * (*alloc_msg) (struct ceph_connection *con,
46 struct ceph_msg_header *hdr);
47 int (*alloc_middle) (struct ceph_connection *con,
48 struct ceph_msg *msg);
49 /* an incoming message has a data payload; tell me what pages I
50 * should read the data into. */
51 int (*prepare_pages) (struct ceph_connection *con, struct ceph_msg *m,
55 extern const char *ceph_name_type_str(int t);
57 /* use format string %s%d */
58 #define ENTITY_NAME(n) ceph_name_type_str((n).type), le64_to_cpu((n).num)
60 struct ceph_messenger {
61 struct ceph_entity_inst inst; /* my name+address */
62 struct ceph_entity_addr my_enc_addr;
63 struct page *zero_page; /* used in certain error cases */
68 * the global_seq counts connections i (attempt to) initiate
69 * in order to disambiguate certain connect race conditions.
72 spinlock_t global_seq_lock;
76 * a single message. it contains a header (src, dest, message type, etc.),
77 * footer (crc values, mainly), a "front" message body, and possibly a
78 * data payload (stored in some number of pages).
81 struct ceph_msg_header hdr; /* header */
82 struct ceph_msg_footer footer; /* footer */
83 struct kvec front; /* unaligned blobs of message */
84 struct ceph_buffer *middle;
85 struct page **pages; /* data payload. NOT OWNER. */
86 unsigned nr_pages; /* size of page array */
87 struct list_head list_head;
89 bool front_is_vmalloc;
93 struct ceph_msgpool *pool;
97 int page, page_pos; /* which page; offset in page */
98 int data_pos; /* offset in data payload */
99 int did_page_crc; /* true if we've calculated crc for current page */
102 /* ceph connection fault delay defaults, for exponential backoff */
103 #define BASE_DELAY_INTERVAL (HZ/2)
104 #define MAX_DELAY_INTERVAL (5 * 60 * HZ)
107 * ceph_connection state bit flags
109 * QUEUED and BUSY are used together to ensure that only a single
110 * thread is currently opening, reading or writing data to the socket.
112 #define LOSSYTX 0 /* we can close channel or drop messages on errors */
114 #define NEGOTIATING 2
115 #define KEEPALIVE_PENDING 3
116 #define WRITE_PENDING 4 /* we have data ready to send */
117 #define QUEUED 5 /* there is work queued on this connection */
118 #define BUSY 6 /* work is being done */
119 #define STANDBY 8 /* no outgoing messages, socket closed. we keep
120 * the ceph_connection around to maintain shared
121 * state with the peer. */
122 #define CLOSED 10 /* we've closed the connection */
123 #define SOCK_CLOSED 11 /* socket state changed to closed */
124 #define REGISTERED 12 /* connection appears in con_tree */
125 #define OPENING 13 /* open connection w/ (possibly new) peer */
126 #define DEAD 14 /* dead, about to kfree */
129 * A single connection with another host.
131 * We maintain a queue of outgoing messages, and some session state to
132 * ensure that we can preserve the lossless, ordered delivery of
133 * messages in the case of a TCP disconnect.
135 struct ceph_connection {
139 const struct ceph_connection_operations *ops;
141 struct ceph_messenger *msgr;
143 unsigned long state; /* connection state (see flags above) */
144 const char *error_msg; /* error message, if any */
146 struct ceph_entity_addr peer_addr; /* peer address */
147 struct ceph_entity_name peer_name; /* peer name */
148 struct ceph_entity_addr peer_addr_for_me;
149 u32 connect_seq; /* identify the most recent connection
150 attempt for this connection, client */
151 u32 peer_global_seq; /* peer's global seq for this connection */
153 int auth_retry; /* true if we need a newer authorizer */
154 void *auth_reply_buf; /* where to put the authorizer reply */
155 int auth_reply_buf_len;
158 struct mutex out_mutex;
159 struct list_head out_queue;
160 struct list_head out_sent; /* sending or sent but unacked */
161 u64 out_seq; /* last message queued for send */
162 u64 out_seq_sent; /* last message sent */
163 bool out_keepalive_pending;
165 u64 in_seq, in_seq_acked; /* last message received, acked */
167 /* connection negotiation temps */
168 char in_banner[CEPH_BANNER_MAX_LEN];
170 struct { /* outgoing connection */
171 struct ceph_msg_connect out_connect;
172 struct ceph_msg_connect_reply in_reply;
174 struct { /* incoming */
175 struct ceph_msg_connect in_connect;
176 struct ceph_msg_connect_reply out_reply;
179 struct ceph_entity_addr actual_peer_addr;
181 /* message out temps */
182 struct ceph_msg *out_msg; /* sending message (== tail of
184 struct ceph_msg_pos out_msg_pos;
186 struct kvec out_kvec[8], /* sending header/footer data */
188 int out_kvec_left; /* kvec's left in out_kvec */
189 int out_skip; /* skip this many bytes */
190 int out_kvec_bytes; /* total bytes left */
191 bool out_kvec_is_msg; /* kvec refers to out_msg */
192 int out_more; /* there is more data after the kvecs */
193 __le64 out_temp_ack; /* for writing an ack */
195 /* message in temps */
196 struct ceph_msg_header in_hdr;
197 struct ceph_msg *in_msg;
198 struct ceph_msg_pos in_msg_pos;
199 u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
201 char in_tag; /* protocol control byte */
202 int in_base_pos; /* bytes read */
203 __le64 in_temp_ack; /* for reading an ack */
205 struct delayed_work work; /* send|recv work */
206 unsigned long delay; /* current delay interval */
210 extern const char *pr_addr(const struct sockaddr_storage *ss);
211 extern int ceph_parse_ips(const char *c, const char *end,
212 struct ceph_entity_addr *addr,
213 int max_count, int *count);
216 extern int ceph_msgr_init(void);
217 extern void ceph_msgr_exit(void);
219 extern struct ceph_messenger *ceph_messenger_create(
220 struct ceph_entity_addr *myaddr);
221 extern void ceph_messenger_destroy(struct ceph_messenger *);
223 extern void ceph_con_init(struct ceph_messenger *msgr,
224 struct ceph_connection *con);
225 extern void ceph_con_open(struct ceph_connection *con,
226 struct ceph_entity_addr *addr);
227 extern void ceph_con_close(struct ceph_connection *con);
228 extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);
229 extern void ceph_con_revoke(struct ceph_connection *con, struct ceph_msg *msg);
230 extern void ceph_con_keepalive(struct ceph_connection *con);
231 extern struct ceph_connection *ceph_con_get(struct ceph_connection *con);
232 extern void ceph_con_put(struct ceph_connection *con);
234 extern struct ceph_msg *ceph_msg_new(int type, int front_len,
235 int page_len, int page_off,
236 struct page **pages);
237 extern void ceph_msg_kfree(struct ceph_msg *m);
239 extern struct ceph_msg *ceph_alloc_msg(struct ceph_connection *con,
240 struct ceph_msg_header *hdr);
241 extern int ceph_alloc_middle(struct ceph_connection *con, struct ceph_msg *msg);
244 static inline struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
246 dout("ceph_msg_get %p %d -> %d\n", msg, atomic_read(&msg->nref),
247 atomic_read(&msg->nref)+1);
248 atomic_inc(&msg->nref);
251 extern void ceph_msg_put(struct ceph_msg *msg);