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/version.h>
10 #include <linux/workqueue.h>
16 struct ceph_connection
;
18 extern struct workqueue_struct
*ceph_msgr_wq
; /* receive work queue */
21 * Ceph defines these callbacks for handling connection events.
23 struct ceph_connection_operations
{
24 struct ceph_connection
*(*get
)(struct ceph_connection
*);
25 void (*put
)(struct ceph_connection
*);
27 /* handle an incoming message. */
28 void (*dispatch
) (struct ceph_connection
*con
, struct ceph_msg
*m
);
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 int (*invalidate_authorizer
)(struct ceph_connection
*con
);
37 /* protocol version mismatch */
38 void (*bad_proto
) (struct ceph_connection
*con
);
40 /* there was some error on the socket (disconnect, whatever) */
41 void (*fault
) (struct ceph_connection
*con
);
43 /* a remote host as terminated a message exchange session, and messages
44 * we sent (or they tried to send us) may be lost. */
45 void (*peer_reset
) (struct ceph_connection
*con
);
47 struct ceph_msg
* (*alloc_msg
) (struct ceph_connection
*con
,
48 struct ceph_msg_header
*hdr
,
52 /* use format string %s%d */
53 #define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
55 struct ceph_messenger
{
56 struct ceph_entity_inst inst
; /* my name+address */
57 struct ceph_entity_addr my_enc_addr
;
58 struct page
*zero_page
; /* used in certain error cases */
63 * the global_seq counts connections i (attempt to) initiate
64 * in order to disambiguate certain connect race conditions.
67 spinlock_t global_seq_lock
;
71 * a single message. it contains a header (src, dest, message type, etc.),
72 * footer (crc values, mainly), a "front" message body, and possibly a
73 * data payload (stored in some number of pages).
76 struct ceph_msg_header hdr
; /* header */
77 struct ceph_msg_footer footer
; /* footer */
78 struct kvec front
; /* unaligned blobs of message */
79 struct ceph_buffer
*middle
;
80 struct page
**pages
; /* data payload. NOT OWNER. */
81 unsigned nr_pages
; /* size of page array */
82 struct ceph_pagelist
*pagelist
; /* instead of pages */
83 struct list_head list_head
;
85 bool front_is_vmalloc
;
90 struct ceph_msgpool
*pool
;
94 int page
, page_pos
; /* which page; offset in page */
95 int data_pos
; /* offset in data payload */
96 int did_page_crc
; /* true if we've calculated crc for current page */
99 /* ceph connection fault delay defaults, for exponential backoff */
100 #define BASE_DELAY_INTERVAL (HZ/2)
101 #define MAX_DELAY_INTERVAL (5 * 60 * HZ)
104 * ceph_connection state bit flags
106 * QUEUED and BUSY are used together to ensure that only a single
107 * thread is currently opening, reading or writing data to the socket.
109 #define LOSSYTX 0 /* we can close channel or drop messages on errors */
111 #define NEGOTIATING 2
112 #define KEEPALIVE_PENDING 3
113 #define WRITE_PENDING 4 /* we have data ready to send */
114 #define QUEUED 5 /* there is work queued on this connection */
115 #define BUSY 6 /* work is being done */
116 #define STANDBY 8 /* no outgoing messages, socket closed. we keep
117 * the ceph_connection around to maintain shared
118 * state with the peer. */
119 #define CLOSED 10 /* we've closed the connection */
120 #define SOCK_CLOSED 11 /* socket state changed to closed */
121 #define OPENING 13 /* open connection w/ (possibly new) peer */
122 #define DEAD 14 /* dead, about to kfree */
125 * A single connection with another host.
127 * We maintain a queue of outgoing messages, and some session state to
128 * ensure that we can preserve the lossless, ordered delivery of
129 * messages in the case of a TCP disconnect.
131 struct ceph_connection
{
135 const struct ceph_connection_operations
*ops
;
137 struct ceph_messenger
*msgr
;
139 unsigned long state
; /* connection state (see flags above) */
140 const char *error_msg
; /* error message, if any */
142 struct ceph_entity_addr peer_addr
; /* peer address */
143 struct ceph_entity_name peer_name
; /* peer name */
144 struct ceph_entity_addr peer_addr_for_me
;
145 u32 connect_seq
; /* identify the most recent connection
146 attempt for this connection, client */
147 u32 peer_global_seq
; /* peer's global seq for this connection */
149 int auth_retry
; /* true if we need a newer authorizer */
150 void *auth_reply_buf
; /* where to put the authorizer reply */
151 int auth_reply_buf_len
;
156 struct list_head out_queue
;
157 struct list_head out_sent
; /* sending or sent but unacked */
158 u64 out_seq
; /* last message queued for send */
159 bool out_keepalive_pending
;
161 u64 in_seq
, in_seq_acked
; /* last message received, acked */
163 /* connection negotiation temps */
164 char in_banner
[CEPH_BANNER_MAX_LEN
];
166 struct { /* outgoing connection */
167 struct ceph_msg_connect out_connect
;
168 struct ceph_msg_connect_reply in_reply
;
170 struct { /* incoming */
171 struct ceph_msg_connect in_connect
;
172 struct ceph_msg_connect_reply out_reply
;
175 struct ceph_entity_addr actual_peer_addr
;
177 /* message out temps */
178 struct ceph_msg
*out_msg
; /* sending message (== tail of
181 struct ceph_msg_pos out_msg_pos
;
183 struct kvec out_kvec
[8], /* sending header/footer data */
185 int out_kvec_left
; /* kvec's left in out_kvec */
186 int out_skip
; /* skip this many bytes */
187 int out_kvec_bytes
; /* total bytes left */
188 bool out_kvec_is_msg
; /* kvec refers to out_msg */
189 int out_more
; /* there is more data after the kvecs */
190 __le64 out_temp_ack
; /* for writing an ack */
192 /* message in temps */
193 struct ceph_msg_header in_hdr
;
194 struct ceph_msg
*in_msg
;
195 struct ceph_msg_pos in_msg_pos
;
196 u32 in_front_crc
, in_middle_crc
, in_data_crc
; /* calculated crc */
198 char in_tag
; /* protocol control byte */
199 int in_base_pos
; /* bytes read */
200 __le64 in_temp_ack
; /* for reading an ack */
202 struct delayed_work work
; /* send|recv work */
203 unsigned long delay
; /* current delay interval */
207 extern const char *pr_addr(const struct sockaddr_storage
*ss
);
208 extern int ceph_parse_ips(const char *c
, const char *end
,
209 struct ceph_entity_addr
*addr
,
210 int max_count
, int *count
);
213 extern int ceph_msgr_init(void);
214 extern void ceph_msgr_exit(void);
216 extern struct ceph_messenger
*ceph_messenger_create(
217 struct ceph_entity_addr
*myaddr
);
218 extern void ceph_messenger_destroy(struct ceph_messenger
*);
220 extern void ceph_con_init(struct ceph_messenger
*msgr
,
221 struct ceph_connection
*con
);
222 extern void ceph_con_open(struct ceph_connection
*con
,
223 struct ceph_entity_addr
*addr
);
224 extern bool ceph_con_opened(struct ceph_connection
*con
);
225 extern void ceph_con_close(struct ceph_connection
*con
);
226 extern void ceph_con_send(struct ceph_connection
*con
, struct ceph_msg
*msg
);
227 extern void ceph_con_revoke(struct ceph_connection
*con
, struct ceph_msg
*msg
);
228 extern void ceph_con_revoke_message(struct ceph_connection
*con
,
229 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 extern void ceph_msg_kfree(struct ceph_msg
*m
);
238 static inline struct ceph_msg
*ceph_msg_get(struct ceph_msg
*msg
)
240 kref_get(&msg
->kref
);
243 extern void ceph_msg_last_put(struct kref
*kref
);
244 static inline void ceph_msg_put(struct ceph_msg
*msg
)
246 kref_put(&msg
->kref
, ceph_msg_last_put
);
249 extern void ceph_msg_dump(struct ceph_msg
*msg
);