b6bec59056d7eb34b8960145baa3ba2d640edde9
[deliverable/linux.git] / fs / ceph / messenger.h
1 #ifndef __FS_CEPH_MESSENGER_H
2 #define __FS_CEPH_MESSENGER_H
3
4 #include <linux/kref.h>
5 #include <linux/mutex.h>
6 #include <linux/net.h>
7 #include <linux/radix-tree.h>
8 #include <linux/uio.h>
9 #include <linux/version.h>
10 #include <linux/workqueue.h>
11
12 #include "types.h"
13 #include "buffer.h"
14
15 struct ceph_msg;
16 struct ceph_connection;
17
18 extern struct workqueue_struct *ceph_msgr_wq; /* receive work queue */
19
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
30 /* authorize an outgoing connection */
31 int (*get_authorizer) (struct ceph_connection *con,
32 void **buf, int *len, int *proto,
33 void **reply_buf, int *reply_len, int force_new);
34 int (*verify_authorizer_reply) (struct ceph_connection *con, int len);
35
36 /* protocol version mismatch */
37 void (*bad_proto) (struct ceph_connection *con);
38
39 /* there was some error on the socket (disconnect, whatever) */
40 void (*fault) (struct ceph_connection *con);
41
42 /* a remote host as terminated a message exchange session, and messages
43 * we sent (or they tried to send us) may be lost. */
44 void (*peer_reset) (struct ceph_connection *con);
45
46 struct ceph_msg * (*alloc_msg) (struct ceph_connection *con,
47 struct ceph_msg_header *hdr,
48 int *skip);
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,
52 int want);
53 };
54
55 extern const char *ceph_name_type_str(int t);
56
57 /* use format string %s%d */
58 #define ENTITY_NAME(n) ceph_name_type_str((n).type), le64_to_cpu((n).num)
59
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 */
64
65 bool nocrc;
66
67 /*
68 * the global_seq counts connections i (attempt to) initiate
69 * in order to disambiguate certain connect race conditions.
70 */
71 u32 global_seq;
72 spinlock_t global_seq_lock;
73 };
74
75 /*
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).
79 */
80 struct ceph_msg {
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 ceph_pagelist *pagelist; /* instead of pages */
88 struct list_head list_head;
89 struct kref kref;
90 bool front_is_vmalloc;
91 bool more_to_follow;
92 int front_max;
93
94 struct ceph_msgpool *pool;
95 };
96
97 struct ceph_msg_pos {
98 int page, page_pos; /* which page; offset in page */
99 int data_pos; /* offset in data payload */
100 int did_page_crc; /* true if we've calculated crc for current page */
101 };
102
103 /* ceph connection fault delay defaults, for exponential backoff */
104 #define BASE_DELAY_INTERVAL (HZ/2)
105 #define MAX_DELAY_INTERVAL (5 * 60 * HZ)
106
107 /*
108 * ceph_connection state bit flags
109 *
110 * QUEUED and BUSY are used together to ensure that only a single
111 * thread is currently opening, reading or writing data to the socket.
112 */
113 #define LOSSYTX 0 /* we can close channel or drop messages on errors */
114 #define CONNECTING 1
115 #define NEGOTIATING 2
116 #define KEEPALIVE_PENDING 3
117 #define WRITE_PENDING 4 /* we have data ready to send */
118 #define QUEUED 5 /* there is work queued on this connection */
119 #define BUSY 6 /* work is being done */
120 #define STANDBY 8 /* no outgoing messages, socket closed. we keep
121 * the ceph_connection around to maintain shared
122 * state with the peer. */
123 #define CLOSED 10 /* we've closed the connection */
124 #define SOCK_CLOSED 11 /* socket state changed to closed */
125 #define REGISTERED 12 /* connection appears in con_tree */
126 #define OPENING 13 /* open connection w/ (possibly new) peer */
127 #define DEAD 14 /* dead, about to kfree */
128
129 /*
130 * A single connection with another host.
131 *
132 * We maintain a queue of outgoing messages, and some session state to
133 * ensure that we can preserve the lossless, ordered delivery of
134 * messages in the case of a TCP disconnect.
135 */
136 struct ceph_connection {
137 void *private;
138 atomic_t nref;
139
140 const struct ceph_connection_operations *ops;
141
142 struct ceph_messenger *msgr;
143 struct socket *sock;
144 unsigned long state; /* connection state (see flags above) */
145 const char *error_msg; /* error message, if any */
146
147 struct ceph_entity_addr peer_addr; /* peer address */
148 struct ceph_entity_name peer_name; /* peer name */
149 struct ceph_entity_addr peer_addr_for_me;
150 u32 connect_seq; /* identify the most recent connection
151 attempt for this connection, client */
152 u32 peer_global_seq; /* peer's global seq for this connection */
153
154 int auth_retry; /* true if we need a newer authorizer */
155 void *auth_reply_buf; /* where to put the authorizer reply */
156 int auth_reply_buf_len;
157
158 struct mutex mutex;
159
160 /* out queue */
161 struct list_head out_queue;
162 struct list_head out_sent; /* sending or sent but unacked */
163 u64 out_seq; /* last message queued for send */
164 u64 out_seq_sent; /* last message sent */
165 bool out_keepalive_pending;
166
167 u64 in_seq, in_seq_acked; /* last message received, acked */
168
169 /* connection negotiation temps */
170 char in_banner[CEPH_BANNER_MAX_LEN];
171 union {
172 struct { /* outgoing connection */
173 struct ceph_msg_connect out_connect;
174 struct ceph_msg_connect_reply in_reply;
175 };
176 struct { /* incoming */
177 struct ceph_msg_connect in_connect;
178 struct ceph_msg_connect_reply out_reply;
179 };
180 };
181 struct ceph_entity_addr actual_peer_addr;
182
183 /* message out temps */
184 struct ceph_msg *out_msg; /* sending message (== tail of
185 out_sent) */
186 bool out_msg_done;
187 struct ceph_msg_pos out_msg_pos;
188
189 struct kvec out_kvec[8], /* sending header/footer data */
190 *out_kvec_cur;
191 int out_kvec_left; /* kvec's left in out_kvec */
192 int out_skip; /* skip this many bytes */
193 int out_kvec_bytes; /* total bytes left */
194 bool out_kvec_is_msg; /* kvec refers to out_msg */
195 int out_more; /* there is more data after the kvecs */
196 __le64 out_temp_ack; /* for writing an ack */
197
198 /* message in temps */
199 struct ceph_msg_header in_hdr;
200 struct ceph_msg *in_msg;
201 struct ceph_msg_pos in_msg_pos;
202 u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
203
204 char in_tag; /* protocol control byte */
205 int in_base_pos; /* bytes read */
206 __le64 in_temp_ack; /* for reading an ack */
207
208 struct delayed_work work; /* send|recv work */
209 unsigned long delay; /* current delay interval */
210 };
211
212
213 extern const char *pr_addr(const struct sockaddr_storage *ss);
214 extern int ceph_parse_ips(const char *c, const char *end,
215 struct ceph_entity_addr *addr,
216 int max_count, int *count);
217
218
219 extern int ceph_msgr_init(void);
220 extern void ceph_msgr_exit(void);
221
222 extern struct ceph_messenger *ceph_messenger_create(
223 struct ceph_entity_addr *myaddr);
224 extern void ceph_messenger_destroy(struct ceph_messenger *);
225
226 extern void ceph_con_init(struct ceph_messenger *msgr,
227 struct ceph_connection *con);
228 extern void ceph_con_open(struct ceph_connection *con,
229 struct ceph_entity_addr *addr);
230 extern void ceph_con_close(struct ceph_connection *con);
231 extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);
232 extern void ceph_con_revoke(struct ceph_connection *con, struct ceph_msg *msg);
233 extern void ceph_con_revoke_pages(struct ceph_connection *con,
234 struct page **pages);
235 extern void ceph_con_keepalive(struct ceph_connection *con);
236 extern struct ceph_connection *ceph_con_get(struct ceph_connection *con);
237 extern void ceph_con_put(struct ceph_connection *con);
238
239 extern struct ceph_msg *ceph_msg_new(int type, int front_len,
240 int page_len, int page_off,
241 struct page **pages);
242 extern void ceph_msg_kfree(struct ceph_msg *m);
243
244
245 static inline struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
246 {
247 kref_get(&msg->kref);
248 return msg;
249 }
250 extern void ceph_msg_last_put(struct kref *kref);
251 static inline void ceph_msg_put(struct ceph_msg *msg)
252 {
253 kref_put(&msg->kref, ceph_msg_last_put);
254 }
255
256 extern void ceph_msg_dump(struct ceph_msg *msg);
257
258 #endif
This page took 0.07654 seconds and 4 git commands to generate.