diff options
Diffstat (limited to 'include/linux/ceph/messenger.h')
| -rw-r--r-- | include/linux/ceph/messenger.h | 261 |
1 files changed, 261 insertions, 0 deletions
diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h new file mode 100644 index 000000000000..5956d62c3057 --- /dev/null +++ b/include/linux/ceph/messenger.h | |||
| @@ -0,0 +1,261 @@ | |||
| 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 | int (*invalidate_authorizer)(struct ceph_connection *con); | ||
| 36 | |||
| 37 | /* protocol version mismatch */ | ||
| 38 | void (*bad_proto) (struct ceph_connection *con); | ||
| 39 | |||
| 40 | /* there was some error on the socket (disconnect, whatever) */ | ||
| 41 | void (*fault) (struct ceph_connection *con); | ||
| 42 | |||
| 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); | ||
| 46 | |||
| 47 | struct ceph_msg * (*alloc_msg) (struct ceph_connection *con, | ||
| 48 | struct ceph_msg_header *hdr, | ||
| 49 | int *skip); | ||
| 50 | }; | ||
| 51 | |||
| 52 | /* use format string %s%d */ | ||
| 53 | #define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num) | ||
| 54 | |||
| 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 */ | ||
| 59 | |||
| 60 | bool nocrc; | ||
| 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 | u32 supported_features; | ||
| 70 | u32 required_features; | ||
| 71 | }; | ||
| 72 | |||
| 73 | /* | ||
| 74 | * a single message. it contains a header (src, dest, message type, etc.), | ||
| 75 | * footer (crc values, mainly), a "front" message body, and possibly a | ||
| 76 | * data payload (stored in some number of pages). | ||
| 77 | */ | ||
| 78 | struct ceph_msg { | ||
| 79 | struct ceph_msg_header hdr; /* header */ | ||
| 80 | struct ceph_msg_footer footer; /* footer */ | ||
| 81 | struct kvec front; /* unaligned blobs of message */ | ||
| 82 | struct ceph_buffer *middle; | ||
| 83 | struct page **pages; /* data payload. NOT OWNER. */ | ||
| 84 | unsigned nr_pages; /* size of page array */ | ||
| 85 | struct ceph_pagelist *pagelist; /* instead of pages */ | ||
| 86 | struct list_head list_head; | ||
| 87 | struct kref kref; | ||
| 88 | struct bio *bio; /* instead of pages/pagelist */ | ||
| 89 | struct bio *bio_iter; /* bio iterator */ | ||
| 90 | int bio_seg; /* current bio segment */ | ||
| 91 | struct ceph_pagelist *trail; /* the trailing part of the data */ | ||
| 92 | bool front_is_vmalloc; | ||
| 93 | bool more_to_follow; | ||
| 94 | bool needs_out_seq; | ||
| 95 | int front_max; | ||
| 96 | |||
| 97 | struct ceph_msgpool *pool; | ||
| 98 | }; | ||
| 99 | |||
| 100 | struct ceph_msg_pos { | ||
| 101 | int page, page_pos; /* which page; offset in page */ | ||
| 102 | int data_pos; /* offset in data payload */ | ||
| 103 | int did_page_crc; /* true if we've calculated crc for current page */ | ||
| 104 | }; | ||
| 105 | |||
| 106 | /* ceph connection fault delay defaults, for exponential backoff */ | ||
| 107 | #define BASE_DELAY_INTERVAL (HZ/2) | ||
| 108 | #define MAX_DELAY_INTERVAL (5 * 60 * HZ) | ||
| 109 | |||
| 110 | /* | ||
| 111 | * ceph_connection state bit flags | ||
| 112 | * | ||
| 113 | * QUEUED and BUSY are used together to ensure that only a single | ||
| 114 | * thread is currently opening, reading or writing data to the socket. | ||
| 115 | */ | ||
| 116 | #define LOSSYTX 0 /* we can close channel or drop messages on errors */ | ||
| 117 | #define CONNECTING 1 | ||
| 118 | #define NEGOTIATING 2 | ||
| 119 | #define KEEPALIVE_PENDING 3 | ||
| 120 | #define WRITE_PENDING 4 /* we have data ready to send */ | ||
| 121 | #define QUEUED 5 /* there is work queued on this connection */ | ||
| 122 | #define BUSY 6 /* work is being done */ | ||
| 123 | #define STANDBY 8 /* no outgoing messages, socket closed. we keep | ||
| 124 | * the ceph_connection around to maintain shared | ||
| 125 | * state with the peer. */ | ||
| 126 | #define CLOSED 10 /* we've closed the connection */ | ||
| 127 | #define SOCK_CLOSED 11 /* socket state changed to closed */ | ||
| 128 | #define OPENING 13 /* open connection w/ (possibly new) peer */ | ||
| 129 | #define DEAD 14 /* dead, about to kfree */ | ||
| 130 | |||
| 131 | /* | ||
| 132 | * A single connection with another host. | ||
| 133 | * | ||
| 134 | * We maintain a queue of outgoing messages, and some session state to | ||
| 135 | * ensure that we can preserve the lossless, ordered delivery of | ||
| 136 | * messages in the case of a TCP disconnect. | ||
| 137 | */ | ||
| 138 | struct ceph_connection { | ||
| 139 | void *private; | ||
| 140 | atomic_t nref; | ||
| 141 | |||
| 142 | const struct ceph_connection_operations *ops; | ||
| 143 | |||
| 144 | struct ceph_messenger *msgr; | ||
| 145 | struct socket *sock; | ||
| 146 | unsigned long state; /* connection state (see flags above) */ | ||
| 147 | const char *error_msg; /* error message, if any */ | ||
| 148 | |||
| 149 | struct ceph_entity_addr peer_addr; /* peer address */ | ||
| 150 | struct ceph_entity_name peer_name; /* peer name */ | ||
| 151 | struct ceph_entity_addr peer_addr_for_me; | ||
| 152 | unsigned peer_features; | ||
| 153 | u32 connect_seq; /* identify the most recent connection | ||
| 154 | attempt for this connection, client */ | ||
| 155 | u32 peer_global_seq; /* peer's global seq for this connection */ | ||
| 156 | |||
| 157 | int auth_retry; /* true if we need a newer authorizer */ | ||
| 158 | void *auth_reply_buf; /* where to put the authorizer reply */ | ||
| 159 | int auth_reply_buf_len; | ||
| 160 | |||
| 161 | struct mutex mutex; | ||
| 162 | |||
| 163 | /* out queue */ | ||
| 164 | struct list_head out_queue; | ||
| 165 | struct list_head out_sent; /* sending or sent but unacked */ | ||
| 166 | u64 out_seq; /* last message queued for send */ | ||
| 167 | bool out_keepalive_pending; | ||
| 168 | |||
| 169 | u64 in_seq, in_seq_acked; /* last message received, acked */ | ||
| 170 | |||
| 171 | /* connection negotiation temps */ | ||
| 172 | char in_banner[CEPH_BANNER_MAX_LEN]; | ||
| 173 | union { | ||
| 174 | struct { /* outgoing connection */ | ||
| 175 | struct ceph_msg_connect out_connect; | ||
| 176 | struct ceph_msg_connect_reply in_reply; | ||
| 177 | }; | ||
| 178 | struct { /* incoming */ | ||
| 179 | struct ceph_msg_connect in_connect; | ||
| 180 | struct ceph_msg_connect_reply out_reply; | ||
| 181 | }; | ||
| 182 | }; | ||
| 183 | struct ceph_entity_addr actual_peer_addr; | ||
| 184 | |||
| 185 | /* message out temps */ | ||
| 186 | struct ceph_msg *out_msg; /* sending message (== tail of | ||
| 187 | out_sent) */ | ||
| 188 | bool out_msg_done; | ||
| 189 | struct ceph_msg_pos out_msg_pos; | ||
| 190 | |||
| 191 | struct kvec out_kvec[8], /* sending header/footer data */ | ||
| 192 | *out_kvec_cur; | ||
| 193 | int out_kvec_left; /* kvec's left in out_kvec */ | ||
| 194 | int out_skip; /* skip this many bytes */ | ||
| 195 | int out_kvec_bytes; /* total bytes left */ | ||
| 196 | bool out_kvec_is_msg; /* kvec refers to out_msg */ | ||
| 197 | int out_more; /* there is more data after the kvecs */ | ||
| 198 | __le64 out_temp_ack; /* for writing an ack */ | ||
| 199 | |||
| 200 | /* message in temps */ | ||
| 201 | struct ceph_msg_header in_hdr; | ||
| 202 | struct ceph_msg *in_msg; | ||
| 203 | struct ceph_msg_pos in_msg_pos; | ||
| 204 | u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */ | ||
| 205 | |||
| 206 | char in_tag; /* protocol control byte */ | ||
| 207 | int in_base_pos; /* bytes read */ | ||
| 208 | __le64 in_temp_ack; /* for reading an ack */ | ||
| 209 | |||
| 210 | struct delayed_work work; /* send|recv work */ | ||
| 211 | unsigned long delay; /* current delay interval */ | ||
| 212 | }; | ||
| 213 | |||
| 214 | |||
| 215 | extern const char *ceph_pr_addr(const struct sockaddr_storage *ss); | ||
| 216 | extern int ceph_parse_ips(const char *c, const char *end, | ||
| 217 | struct ceph_entity_addr *addr, | ||
| 218 | int max_count, int *count); | ||
| 219 | |||
| 220 | |||
| 221 | extern int ceph_msgr_init(void); | ||
| 222 | extern void ceph_msgr_exit(void); | ||
| 223 | extern void ceph_msgr_flush(void); | ||
| 224 | |||
| 225 | extern struct ceph_messenger *ceph_messenger_create( | ||
| 226 | struct ceph_entity_addr *myaddr, | ||
| 227 | u32 features, u32 required); | ||
| 228 | extern void ceph_messenger_destroy(struct ceph_messenger *); | ||
| 229 | |||
| 230 | extern void ceph_con_init(struct ceph_messenger *msgr, | ||
| 231 | struct ceph_connection *con); | ||
| 232 | extern void ceph_con_open(struct ceph_connection *con, | ||
| 233 | struct ceph_entity_addr *addr); | ||
| 234 | extern bool ceph_con_opened(struct ceph_connection *con); | ||
| 235 | extern void ceph_con_close(struct ceph_connection *con); | ||
| 236 | extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg); | ||
| 237 | extern void ceph_con_revoke(struct ceph_connection *con, struct ceph_msg *msg); | ||
| 238 | extern void ceph_con_revoke_message(struct ceph_connection *con, | ||
| 239 | struct ceph_msg *msg); | ||
| 240 | extern void ceph_con_keepalive(struct ceph_connection *con); | ||
| 241 | extern struct ceph_connection *ceph_con_get(struct ceph_connection *con); | ||
| 242 | extern void ceph_con_put(struct ceph_connection *con); | ||
| 243 | |||
| 244 | extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags); | ||
| 245 | extern void ceph_msg_kfree(struct ceph_msg *m); | ||
| 246 | |||
| 247 | |||
| 248 | static inline struct ceph_msg *ceph_msg_get(struct ceph_msg *msg) | ||
| 249 | { | ||
| 250 | kref_get(&msg->kref); | ||
| 251 | return msg; | ||
| 252 | } | ||
| 253 | extern void ceph_msg_last_put(struct kref *kref); | ||
| 254 | static inline void ceph_msg_put(struct ceph_msg *msg) | ||
| 255 | { | ||
| 256 | kref_put(&msg->kref, ceph_msg_last_put); | ||
| 257 | } | ||
| 258 | |||
| 259 | extern void ceph_msg_dump(struct ceph_msg *msg); | ||
| 260 | |||
| 261 | #endif | ||
