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