include/linux/ceph/messenger.h

   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%lld */
  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 #ifdef CONFIG_BLOCK
  86
  87 struct ceph_bio_iter {
  88         struct bio *bio;
  89         struct bvec_iter iter;
  90 };
  91
  92 #define __ceph_bio_iter_advance_step(it, n, STEP) do {                        \
  93         unsigned int __n = (n), __cur_n;                                      \
  94                                                                               \
  95         while (__n) {                                                         \
  96                 BUG_ON(!(it)->iter.bi_size);                                  \
  97                 __cur_n = min((it)->iter.bi_size, __n);                       \
  98                 (void)(STEP);                                                 \
  99                 bio_advance_iter((it)->bio, &(it)->iter, __cur_n);            \
 100                 if (!(it)->iter.bi_size && (it)->bio->bi_next) {              \
 101                         dout("__ceph_bio_iter_advance_step next bio\n");      \
 102                         (it)->bio = (it)->bio->bi_next;                       \
 103                         (it)->iter = (it)->bio->bi_iter;                      \
 104                 }                                                             \
 105                 __n -= __cur_n;                                               \
 106         }                                                                     \
 107 } while (0)
 108
 109 /*
 110  * Advance @it by @n bytes.
 111  */
 112 #define ceph_bio_iter_advance(it, n)                                          \
 113         __ceph_bio_iter_advance_step(it, n, 0)
 114
 115 /*
 116  * Advance @it by @n bytes, executing BVEC_STEP for each bio_vec.
 117  */
 118 #define ceph_bio_iter_advance_step(it, n, BVEC_STEP)                          \
 119         __ceph_bio_iter_advance_step(it, n, ({                                \
 120                 struct bio_vec bv;                                            \
 121                 struct bvec_iter __cur_iter;                                  \
 122                                                                               \
 123                 __cur_iter = (it)->iter;                                      \
 124                 __cur_iter.bi_size = __cur_n;                                 \
 125                 __bio_for_each_segment(bv, (it)->bio, __cur_iter, __cur_iter) \
 126                         (void)(BVEC_STEP);                                    \
 127         }))
 128
 129 #endif /* CONFIG_BLOCK */
 130
 131 struct ceph_bvec_iter {
 132         struct bio_vec *bvecs;
 133         struct bvec_iter iter;
 134 };
 135
 136 #define __ceph_bvec_iter_advance_step(it, n, STEP) do {                       \
 137         BUG_ON((n) > (it)->iter.bi_size);                                     \
 138         (void)(STEP);                                                         \
 139         bvec_iter_advance((it)->bvecs, &(it)->iter, (n));                     \
 140 } while (0)
 141
 142 /*
 143  * Advance @it by @n bytes.
 144  */
 145 #define ceph_bvec_iter_advance(it, n)                                         \
 146         __ceph_bvec_iter_advance_step(it, n, 0)
 147
 148 /*
 149  * Advance @it by @n bytes, executing BVEC_STEP for each bio_vec.
 150  */
 151 #define ceph_bvec_iter_advance_step(it, n, BVEC_STEP)                         \
 152         __ceph_bvec_iter_advance_step(it, n, ({                               \
 153                 struct bio_vec bv;                                            \
 154                 struct bvec_iter __cur_iter;                                  \
 155                                                                               \
 156                 __cur_iter = (it)->iter;                                      \
 157                 __cur_iter.bi_size = (n);                                     \
 158                 for_each_bvec(bv, (it)->bvecs, __cur_iter, __cur_iter)        \
 159                         (void)(BVEC_STEP);                                    \
 160         }))
 161
 162 #define ceph_bvec_iter_shorten(it, n) do {                                    \
 163         BUG_ON((n) > (it)->iter.bi_size);                                     \
 164         (it)->iter.bi_size = (n);                                             \
 165 } while (0)
 166
 167 struct ceph_msg_data {
 168         enum ceph_msg_data_type         type;
 169         union {
 170 #ifdef CONFIG_BLOCK
 171                 struct {
 172                         struct ceph_bio_iter    bio_pos;
 173                         u32                     bio_length;
 174                 };
 175 #endif /* CONFIG_BLOCK */
 176                 struct ceph_bvec_iter   bvec_pos;
 177                 struct {
 178                         struct page     **pages;
 179                         size_t          length;         /* total # bytes */
 180                         unsigned int    alignment;      /* first page */
 181                         bool            own_pages;
 182                 };
 183                 struct ceph_pagelist    *pagelist;
 184         };
 185 };
 186
 187 struct ceph_msg_data_cursor {
 188         size_t                  total_resid;    /* across all data items */
 189
 190         struct ceph_msg_data    *data;          /* current data item */
 191         size_t                  resid;          /* bytes not yet consumed */
 192         bool                    last_piece;     /* current is last piece */
 193         bool                    need_crc;       /* crc update needed */
 194         union {
 195 #ifdef CONFIG_BLOCK
 196                 struct ceph_bio_iter    bio_iter;
 197 #endif /* CONFIG_BLOCK */
 198                 struct bvec_iter        bvec_iter;
 199                 struct {                                /* pages */
 200                         unsigned int    page_offset;    /* offset in page */
 201                         unsigned short  page_index;     /* index in array */
 202                         unsigned short  page_count;     /* pages in array */
 203                 };
 204                 struct {                                /* pagelist */
 205                         struct page     *page;          /* page from list */
 206                         size_t          offset;         /* bytes from list */
 207                 };
 208         };
 209 };
 210
 211 /*
 212  * a single message.  it contains a header (src, dest, message type, etc.),
 213  * footer (crc values, mainly), a "front" message body, and possibly a
 214  * data payload (stored in some number of pages).
 215  */
 216 struct ceph_msg {
 217         struct ceph_msg_header hdr;     /* header */
 218         union {
 219                 struct ceph_msg_footer footer;          /* footer */
 220                 struct ceph_msg_footer_old old_footer;  /* old format footer */
 221         };
 222         struct kvec front;              /* unaligned blobs of message */
 223         struct ceph_buffer *middle;
 224
 225         size_t                          data_length;
 226         struct ceph_msg_data            *data;
 227         int                             num_data_items;
 228         int                             max_data_items;
 229         struct ceph_msg_data_cursor     cursor;
 230
 231         struct ceph_connection *con;
 232         struct list_head list_head;     /* links for connection lists */
 233
 234         struct kref kref;
 235         bool more_to_follow;
 236         bool needs_out_seq;
 237         int front_alloc_len;
 238
 239         struct ceph_msgpool *pool;
 240 };
 241
 242 /*
 243  * connection states
 244  */
 245 #define CEPH_CON_S_CLOSED               1
 246 #define CEPH_CON_S_PREOPEN              2
 247 #define CEPH_CON_S_V1_BANNER            3
 248 #define CEPH_CON_S_V1_CONNECT_MSG       4
 249 #define CEPH_CON_S_OPEN                 5
 250 #define CEPH_CON_S_STANDBY              6
 251
 252 /*
 253  * ceph_connection flag bits
 254  */
 255 #define CEPH_CON_F_LOSSYTX              0  /* we can close channel or drop
 256                                               messages on errors */
 257 #define CEPH_CON_F_KEEPALIVE_PENDING    1  /* we need to send a keepalive */
 258 #define CEPH_CON_F_WRITE_PENDING        2  /* we have data ready to send */
 259 #define CEPH_CON_F_SOCK_CLOSED          3  /* socket state changed to closed */
 260 #define CEPH_CON_F_BACKOFF              4  /* need to retry queuing delayed
 261                                               work */
 262
 263 /* ceph connection fault delay defaults, for exponential backoff */
 264 #define BASE_DELAY_INTERVAL     (HZ / 4)
 265 #define MAX_DELAY_INTERVAL      (15 * HZ)
 266
 267 /*
 268  * A single connection with another host.
 269  *
 270  * We maintain a queue of outgoing messages, and some session state to
 271  * ensure that we can preserve the lossless, ordered delivery of
 272  * messages in the case of a TCP disconnect.
 273  */
 274 struct ceph_connection {
 275         void *private;
 276
 277         const struct ceph_connection_operations *ops;
 278
 279         struct ceph_messenger *msgr;
 280
 281         int state;  /* CEPH_CON_S_* */
 282         atomic_t sock_state;
 283         struct socket *sock;
 284         struct ceph_entity_addr peer_addr; /* peer address */
 285         struct ceph_entity_addr peer_addr_for_me;
 286
 287         unsigned long flags;  /* CEPH_CON_F_* */
 288         const char *error_msg;  /* error message, if any */
 289
 290         struct ceph_entity_name peer_name; /* peer name */
 291
 292         u64 peer_features;
 293         u32 connect_seq;      /* identify the most recent connection
 294                                  attempt for this connection, client */
 295         u32 peer_global_seq;  /* peer's global seq for this connection */
 296
 297         struct ceph_auth_handshake *auth;
 298         int auth_retry;       /* true if we need a newer authorizer */
 299
 300         struct mutex mutex;
 301
 302         /* out queue */
 303         struct list_head out_queue;
 304         struct list_head out_sent;   /* sending or sent but unacked */
 305         u64 out_seq;                 /* last message queued for send */
 306
 307         u64 in_seq, in_seq_acked;  /* last message received, acked */
 308
 309         /* connection negotiation temps */
 310         char in_banner[CEPH_BANNER_MAX_LEN];
 311         struct ceph_msg_connect out_connect;
 312         struct ceph_msg_connect_reply in_reply;
 313         struct ceph_entity_addr actual_peer_addr;
 314
 315         /* message out temps */
 316         struct ceph_msg_header out_hdr;
 317         struct ceph_msg *out_msg;        /* sending message (== tail of
 318                                             out_sent) */
 319         bool out_msg_done;
 320
 321         struct kvec out_kvec[8],         /* sending header/footer data */
 322                 *out_kvec_cur;
 323         int out_kvec_left;   /* kvec's left in out_kvec */
 324         int out_skip;        /* skip this many bytes */
 325         int out_kvec_bytes;  /* total bytes left */
 326         int out_more;        /* there is more data after the kvecs */
 327         __le64 out_temp_ack; /* for writing an ack */
 328         struct ceph_timespec out_temp_keepalive2; /* for writing keepalive2
 329                                                      stamp */
 330
 331         /* message in temps */
 332         struct ceph_msg_header in_hdr;
 333         struct ceph_msg *in_msg;
 334         u32 in_front_crc, in_middle_crc, in_data_crc;  /* calculated crc */
 335
 336         char in_tag;         /* protocol control byte */
 337         int in_base_pos;     /* bytes read */
 338         __le64 in_temp_ack;  /* for reading an ack */
 339
 340         struct timespec64 last_keepalive_ack; /* keepalive2 ack stamp */
 341
 342         struct delayed_work work;           /* send|recv work */
 343         unsigned long       delay;          /* current delay interval */
 344 };
 345
 346
 347 extern const char *ceph_pr_addr(const struct ceph_entity_addr *addr);
 348
 349 extern int ceph_parse_ips(const char *c, const char *end,
 350                           struct ceph_entity_addr *addr,
 351                           int max_count, int *count);
 352
 353
 354 extern int ceph_msgr_init(void);
 355 extern void ceph_msgr_exit(void);
 356 extern void ceph_msgr_flush(void);
 357
 358 extern void ceph_messenger_init(struct ceph_messenger *msgr,
 359                                 struct ceph_entity_addr *myaddr);
 360 extern void ceph_messenger_fini(struct ceph_messenger *msgr);
 361 extern void ceph_messenger_reset_nonce(struct ceph_messenger *msgr);
 362
 363 extern void ceph_con_init(struct ceph_connection *con, void *private,
 364                         const struct ceph_connection_operations *ops,
 365                         struct ceph_messenger *msgr);
 366 extern void ceph_con_open(struct ceph_connection *con,
 367                           __u8 entity_type, __u64 entity_num,
 368                           struct ceph_entity_addr *addr);
 369 extern bool ceph_con_opened(struct ceph_connection *con);
 370 extern void ceph_con_close(struct ceph_connection *con);
 371 extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);
 372
 373 extern void ceph_msg_revoke(struct ceph_msg *msg);
 374 extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
 375
 376 extern void ceph_con_keepalive(struct ceph_connection *con);
 377 extern bool ceph_con_keepalive_expired(struct ceph_connection *con,
 378                                        unsigned long interval);
 379
 380 void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
 381                              size_t length, size_t alignment, bool own_pages);
 382 extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
 383                                 struct ceph_pagelist *pagelist);
 384 #ifdef CONFIG_BLOCK
 385 void ceph_msg_data_add_bio(struct ceph_msg *msg, struct ceph_bio_iter *bio_pos,
 386                            u32 length);
 387 #endif /* CONFIG_BLOCK */
 388 void ceph_msg_data_add_bvecs(struct ceph_msg *msg,
 389                              struct ceph_bvec_iter *bvec_pos);
 390
 391 struct ceph_msg *ceph_msg_new2(int type, int front_len, int max_data_items,
 392                                gfp_t flags, bool can_fail);
 393 extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
 394                                      bool can_fail);
 395
 396 extern struct ceph_msg *ceph_msg_get(struct ceph_msg *msg);
 397 extern void ceph_msg_put(struct ceph_msg *msg);
 398
 399 extern void ceph_msg_dump(struct ceph_msg *msg);
 400
 401 #endif