1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef __FS_CEPH_MESSENGER_H 3 #define __FS_CEPH_MESSENGER_H 4 5 #include <linux/bvec.h> 6 #include <linux/kref.h> 7 #include <linux/mutex.h> 8 #include <linux/net.h> 9 #include <linux/radix-tree.h> 10 #include <linux/uio.h> 11 #include <linux/workqueue.h> 12 #include <net/net_namespace.h> 13 14 #include <linux/ceph/types.h> 15 #include <linux/ceph/buffer.h> 16 17 struct ceph_msg; 18 struct ceph_connection; 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 struct ceph_auth_handshake *(*get_authorizer) ( 32 struct ceph_connection *con, 33 int *proto, int force_new); 34 int (*add_authorizer_challenge)(struct ceph_connection *con, 35 void *challenge_buf, 36 int challenge_buf_len); 37 int (*verify_authorizer_reply) (struct ceph_connection *con); 38 int (*invalidate_authorizer)(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 void (*reencode_message) (struct ceph_msg *msg); 52 53 int (*sign_message) (struct ceph_msg *msg); 54 int (*check_message_signature) (struct ceph_msg *msg); 55 }; 56 57 /* use format string %s%d */ 58 #define ENTITY_NAME(n) ceph_entity_type_name((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 64 atomic_t stopping; 65 possible_net_t net; 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 enum ceph_msg_data_type { 76 CEPH_MSG_DATA_NONE, /* message contains no data payload */ 77 CEPH_MSG_DATA_PAGES, /* data source/destination is a page array */ 78 CEPH_MSG_DATA_PAGELIST, /* data source/destination is a pagelist */ 79 #ifdef CONFIG_BLOCK 80 CEPH_MSG_DATA_BIO, /* data source/destination is a bio list */ 81 #endif /* CONFIG_BLOCK */ 82 CEPH_MSG_DATA_BVECS, /* data source/destination is a bio_vec array */ 83 }; 84 85 static __inline__ bool ceph_msg_data_type_valid(enum ceph_msg_data_type type) 86 { 87 switch (type) { 88 case CEPH_MSG_DATA_NONE: 89 case CEPH_MSG_DATA_PAGES: 90 case CEPH_MSG_DATA_PAGELIST: 91 #ifdef CONFIG_BLOCK 92 case CEPH_MSG_DATA_BIO: 93 #endif /* CONFIG_BLOCK */ 94 case CEPH_MSG_DATA_BVECS: 95 return true; 96 default: 97 return false; 98 } 99 } 100 101 #ifdef CONFIG_BLOCK 102 103 struct ceph_bio_iter { 104 struct bio *bio; 105 struct bvec_iter iter; 106 }; 107 108 #define __ceph_bio_iter_advance_step(it, n, STEP) do { \ 109 unsigned int __n = (n), __cur_n; \ 110 \ 111 while (__n) { \ 112 BUG_ON(!(it)->iter.bi_size); \ 113 __cur_n = min((it)->iter.bi_size, __n); \ 114 (void)(STEP); \ 115 bio_advance_iter((it)->bio, &(it)->iter, __cur_n); \ 116 if (!(it)->iter.bi_size && (it)->bio->bi_next) { \ 117 dout("__ceph_bio_iter_advance_step next bio\n"); \ 118 (it)->bio = (it)->bio->bi_next; \ 119 (it)->iter = (it)->bio->bi_iter; \ 120 } \ 121 __n -= __cur_n; \ 122 } \ 123 } while (0) 124 125 /* 126 * Advance @it by @n bytes. 127 */ 128 #define ceph_bio_iter_advance(it, n) \ 129 __ceph_bio_iter_advance_step(it, n, 0) 130 131 /* 132 * Advance @it by @n bytes, executing BVEC_STEP for each bio_vec. 133 */ 134 #define ceph_bio_iter_advance_step(it, n, BVEC_STEP) \ 135 __ceph_bio_iter_advance_step(it, n, ({ \ 136 struct bio_vec bv; \ 137 struct bvec_iter __cur_iter; \ 138 \ 139 __cur_iter = (it)->iter; \ 140 __cur_iter.bi_size = __cur_n; \ 141 __bio_for_each_segment(bv, (it)->bio, __cur_iter, __cur_iter) \ 142 (void)(BVEC_STEP); \ 143 })) 144 145 #endif /* CONFIG_BLOCK */ 146 147 struct ceph_bvec_iter { 148 struct bio_vec *bvecs; 149 struct bvec_iter iter; 150 }; 151 152 #define __ceph_bvec_iter_advance_step(it, n, STEP) do { \ 153 BUG_ON((n) > (it)->iter.bi_size); \ 154 (void)(STEP); \ 155 bvec_iter_advance((it)->bvecs, &(it)->iter, (n)); \ 156 } while (0) 157 158 /* 159 * Advance @it by @n bytes. 160 */ 161 #define ceph_bvec_iter_advance(it, n) \ 162 __ceph_bvec_iter_advance_step(it, n, 0) 163 164 /* 165 * Advance @it by @n bytes, executing BVEC_STEP for each bio_vec. 166 */ 167 #define ceph_bvec_iter_advance_step(it, n, BVEC_STEP) \ 168 __ceph_bvec_iter_advance_step(it, n, ({ \ 169 struct bio_vec bv; \ 170 struct bvec_iter __cur_iter; \ 171 \ 172 __cur_iter = (it)->iter; \ 173 __cur_iter.bi_size = (n); \ 174 for_each_bvec(bv, (it)->bvecs, __cur_iter, __cur_iter) \ 175 (void)(BVEC_STEP); \ 176 })) 177 178 #define ceph_bvec_iter_shorten(it, n) do { \ 179 BUG_ON((n) > (it)->iter.bi_size); \ 180 (it)->iter.bi_size = (n); \ 181 } while (0) 182 183 struct ceph_msg_data { 184 struct list_head links; /* ceph_msg->data */ 185 enum ceph_msg_data_type type; 186 union { 187 #ifdef CONFIG_BLOCK 188 struct { 189 struct ceph_bio_iter bio_pos; 190 u32 bio_length; 191 }; 192 #endif /* CONFIG_BLOCK */ 193 struct ceph_bvec_iter bvec_pos; 194 struct { 195 struct page **pages; /* NOT OWNER. */ 196 size_t length; /* total # bytes */ 197 unsigned int alignment; /* first page */ 198 }; 199 struct ceph_pagelist *pagelist; 200 }; 201 }; 202 203 struct ceph_msg_data_cursor { 204 size_t total_resid; /* across all data items */ 205 struct list_head *data_head; /* = &ceph_msg->data */ 206 207 struct ceph_msg_data *data; /* current data item */ 208 size_t resid; /* bytes not yet consumed */ 209 bool last_piece; /* current is last piece */ 210 bool need_crc; /* crc update needed */ 211 union { 212 #ifdef CONFIG_BLOCK 213 struct ceph_bio_iter bio_iter; 214 #endif /* CONFIG_BLOCK */ 215 struct bvec_iter bvec_iter; 216 struct { /* pages */ 217 unsigned int page_offset; /* offset in page */ 218 unsigned short page_index; /* index in array */ 219 unsigned short page_count; /* pages in array */ 220 }; 221 struct { /* pagelist */ 222 struct page *page; /* page from list */ 223 size_t offset; /* bytes from list */ 224 }; 225 }; 226 }; 227 228 /* 229 * a single message. it contains a header (src, dest, message type, etc.), 230 * footer (crc values, mainly), a "front" message body, and possibly a 231 * data payload (stored in some number of pages). 232 */ 233 struct ceph_msg { 234 struct ceph_msg_header hdr; /* header */ 235 union { 236 struct ceph_msg_footer footer; /* footer */ 237 struct ceph_msg_footer_old old_footer; /* old format footer */ 238 }; 239 struct kvec front; /* unaligned blobs of message */ 240 struct ceph_buffer *middle; 241 242 size_t data_length; 243 struct list_head data; 244 struct ceph_msg_data_cursor cursor; 245 246 struct ceph_connection *con; 247 struct list_head list_head; /* links for connection lists */ 248 249 struct kref kref; 250 bool more_to_follow; 251 bool needs_out_seq; 252 int front_alloc_len; 253 unsigned long ack_stamp; /* tx: when we were acked */ 254 255 struct ceph_msgpool *pool; 256 }; 257 258 /* ceph connection fault delay defaults, for exponential backoff */ 259 #define BASE_DELAY_INTERVAL (HZ/2) 260 #define MAX_DELAY_INTERVAL (5 * 60 * HZ) 261 262 /* 263 * A single connection with another host. 264 * 265 * We maintain a queue of outgoing messages, and some session state to 266 * ensure that we can preserve the lossless, ordered delivery of 267 * messages in the case of a TCP disconnect. 268 */ 269 struct ceph_connection { 270 void *private; 271 272 const struct ceph_connection_operations *ops; 273 274 struct ceph_messenger *msgr; 275 276 atomic_t sock_state; 277 struct socket *sock; 278 struct ceph_entity_addr peer_addr; /* peer address */ 279 struct ceph_entity_addr peer_addr_for_me; 280 281 unsigned long flags; 282 unsigned long state; 283 const char *error_msg; /* error message, if any */ 284 285 struct ceph_entity_name peer_name; /* peer name */ 286 287 u64 peer_features; 288 u32 connect_seq; /* identify the most recent connection 289 attempt for this connection, client */ 290 u32 peer_global_seq; /* peer's global seq for this connection */ 291 292 struct ceph_auth_handshake *auth; 293 int auth_retry; /* true if we need a newer authorizer */ 294 295 struct mutex mutex; 296 297 /* out queue */ 298 struct list_head out_queue; 299 struct list_head out_sent; /* sending or sent but unacked */ 300 u64 out_seq; /* last message queued for send */ 301 302 u64 in_seq, in_seq_acked; /* last message received, acked */ 303 304 /* connection negotiation temps */ 305 char in_banner[CEPH_BANNER_MAX_LEN]; 306 struct ceph_msg_connect out_connect; 307 struct ceph_msg_connect_reply in_reply; 308 struct ceph_entity_addr actual_peer_addr; 309 310 /* message out temps */ 311 struct ceph_msg_header out_hdr; 312 struct ceph_msg *out_msg; /* sending message (== tail of 313 out_sent) */ 314 bool out_msg_done; 315 316 struct kvec out_kvec[8], /* sending header/footer data */ 317 *out_kvec_cur; 318 int out_kvec_left; /* kvec's left in out_kvec */ 319 int out_skip; /* skip this many bytes */ 320 int out_kvec_bytes; /* total bytes left */ 321 int out_more; /* there is more data after the kvecs */ 322 __le64 out_temp_ack; /* for writing an ack */ 323 struct ceph_timespec out_temp_keepalive2; /* for writing keepalive2 324 stamp */ 325 326 /* message in temps */ 327 struct ceph_msg_header in_hdr; 328 struct ceph_msg *in_msg; 329 u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */ 330 331 char in_tag; /* protocol control byte */ 332 int in_base_pos; /* bytes read */ 333 __le64 in_temp_ack; /* for reading an ack */ 334 335 struct timespec64 last_keepalive_ack; /* keepalive2 ack stamp */ 336 337 struct delayed_work work; /* send|recv work */ 338 unsigned long delay; /* current delay interval */ 339 }; 340 341 342 extern const char *ceph_pr_addr(const struct sockaddr_storage *ss); 343 extern int ceph_parse_ips(const char *c, const char *end, 344 struct ceph_entity_addr *addr, 345 int max_count, int *count); 346 347 348 extern int ceph_msgr_init(void); 349 extern void ceph_msgr_exit(void); 350 extern void ceph_msgr_flush(void); 351 352 extern void ceph_messenger_init(struct ceph_messenger *msgr, 353 struct ceph_entity_addr *myaddr); 354 extern void ceph_messenger_fini(struct ceph_messenger *msgr); 355 356 extern void ceph_con_init(struct ceph_connection *con, void *private, 357 const struct ceph_connection_operations *ops, 358 struct ceph_messenger *msgr); 359 extern void ceph_con_open(struct ceph_connection *con, 360 __u8 entity_type, __u64 entity_num, 361 struct ceph_entity_addr *addr); 362 extern bool ceph_con_opened(struct ceph_connection *con); 363 extern void ceph_con_close(struct ceph_connection *con); 364 extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg); 365 366 extern void ceph_msg_revoke(struct ceph_msg *msg); 367 extern void ceph_msg_revoke_incoming(struct ceph_msg *msg); 368 369 extern void ceph_con_keepalive(struct ceph_connection *con); 370 extern bool ceph_con_keepalive_expired(struct ceph_connection *con, 371 unsigned long interval); 372 373 extern void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages, 374 size_t length, size_t alignment); 375 extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg, 376 struct ceph_pagelist *pagelist); 377 #ifdef CONFIG_BLOCK 378 void ceph_msg_data_add_bio(struct ceph_msg *msg, struct ceph_bio_iter *bio_pos, 379 u32 length); 380 #endif /* CONFIG_BLOCK */ 381 void ceph_msg_data_add_bvecs(struct ceph_msg *msg, 382 struct ceph_bvec_iter *bvec_pos); 383 384 extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags, 385 bool can_fail); 386 387 extern struct ceph_msg *ceph_msg_get(struct ceph_msg *msg); 388 extern void ceph_msg_put(struct ceph_msg *msg); 389 390 extern void ceph_msg_dump(struct ceph_msg *msg); 391 392 #endif 393