1 #ifndef _FS_CEPH_OSD_CLIENT_H 2 #define _FS_CEPH_OSD_CLIENT_H 3 4 #include <linux/completion.h> 5 #include <linux/kref.h> 6 #include <linux/mempool.h> 7 #include <linux/rbtree.h> 8 9 #include <linux/ceph/types.h> 10 #include <linux/ceph/osdmap.h> 11 #include <linux/ceph/messenger.h> 12 #include <linux/ceph/msgpool.h> 13 #include <linux/ceph/auth.h> 14 #include <linux/ceph/pagelist.h> 15 16 struct ceph_msg; 17 struct ceph_snap_context; 18 struct ceph_osd_request; 19 struct ceph_osd_client; 20 21 /* 22 * completion callback for async writepages 23 */ 24 typedef void (*ceph_osdc_callback_t)(struct ceph_osd_request *); 25 26 #define CEPH_HOMELESS_OSD -1 27 28 /* a given osd we're communicating with */ 29 struct ceph_osd { 30 atomic_t o_ref; 31 struct ceph_osd_client *o_osdc; 32 int o_osd; 33 int o_incarnation; 34 struct rb_node o_node; 35 struct ceph_connection o_con; 36 struct rb_root o_requests; 37 struct rb_root o_linger_requests; 38 struct list_head o_osd_lru; 39 struct ceph_auth_handshake o_auth; 40 unsigned long lru_ttl; 41 struct list_head o_keepalive_item; 42 struct mutex lock; 43 }; 44 45 #define CEPH_OSD_SLAB_OPS 2 46 #define CEPH_OSD_MAX_OPS 16 47 48 enum ceph_osd_data_type { 49 CEPH_OSD_DATA_TYPE_NONE = 0, 50 CEPH_OSD_DATA_TYPE_PAGES, 51 CEPH_OSD_DATA_TYPE_PAGELIST, 52 #ifdef CONFIG_BLOCK 53 CEPH_OSD_DATA_TYPE_BIO, 54 #endif /* CONFIG_BLOCK */ 55 }; 56 57 struct ceph_osd_data { 58 enum ceph_osd_data_type type; 59 union { 60 struct { 61 struct page **pages; 62 u64 length; 63 u32 alignment; 64 bool pages_from_pool; 65 bool own_pages; 66 }; 67 struct ceph_pagelist *pagelist; 68 #ifdef CONFIG_BLOCK 69 struct { 70 struct bio *bio; /* list of bios */ 71 size_t bio_length; /* total in list */ 72 }; 73 #endif /* CONFIG_BLOCK */ 74 }; 75 }; 76 77 struct ceph_osd_req_op { 78 u16 op; /* CEPH_OSD_OP_* */ 79 u32 flags; /* CEPH_OSD_OP_FLAG_* */ 80 u32 indata_len; /* request */ 81 u32 outdata_len; /* reply */ 82 s32 rval; 83 84 union { 85 struct ceph_osd_data raw_data_in; 86 struct { 87 u64 offset, length; 88 u64 truncate_size; 89 u32 truncate_seq; 90 struct ceph_osd_data osd_data; 91 } extent; 92 struct { 93 u32 name_len; 94 u32 value_len; 95 __u8 cmp_op; /* CEPH_OSD_CMPXATTR_OP_* */ 96 __u8 cmp_mode; /* CEPH_OSD_CMPXATTR_MODE_* */ 97 struct ceph_osd_data osd_data; 98 } xattr; 99 struct { 100 const char *class_name; 101 const char *method_name; 102 struct ceph_osd_data request_info; 103 struct ceph_osd_data request_data; 104 struct ceph_osd_data response_data; 105 __u8 class_len; 106 __u8 method_len; 107 u32 indata_len; 108 } cls; 109 struct { 110 u64 cookie; 111 __u8 op; /* CEPH_OSD_WATCH_OP_ */ 112 u32 gen; 113 } watch; 114 struct { 115 struct ceph_osd_data request_data; 116 } notify_ack; 117 struct { 118 u64 cookie; 119 struct ceph_osd_data request_data; 120 struct ceph_osd_data response_data; 121 } notify; 122 struct { 123 struct ceph_osd_data response_data; 124 } list_watchers; 125 struct { 126 u64 expected_object_size; 127 u64 expected_write_size; 128 } alloc_hint; 129 }; 130 }; 131 132 struct ceph_osd_request_target { 133 struct ceph_object_id base_oid; 134 struct ceph_object_locator base_oloc; 135 struct ceph_object_id target_oid; 136 struct ceph_object_locator target_oloc; 137 138 struct ceph_pg pgid; 139 u32 pg_num; 140 u32 pg_num_mask; 141 struct ceph_osds acting; 142 struct ceph_osds up; 143 int size; 144 int min_size; 145 bool sort_bitwise; 146 147 unsigned int flags; /* CEPH_OSD_FLAG_* */ 148 bool paused; 149 150 int osd; 151 }; 152 153 /* an in-flight request */ 154 struct ceph_osd_request { 155 u64 r_tid; /* unique for this client */ 156 struct rb_node r_node; 157 struct rb_node r_mc_node; /* map check */ 158 struct ceph_osd *r_osd; 159 160 struct ceph_osd_request_target r_t; 161 #define r_base_oid r_t.base_oid 162 #define r_base_oloc r_t.base_oloc 163 #define r_flags r_t.flags 164 165 struct ceph_msg *r_request, *r_reply; 166 u32 r_sent; /* >0 if r_request is sending/sent */ 167 168 /* request osd ops array */ 169 unsigned int r_num_ops; 170 171 int r_result; 172 173 struct ceph_osd_client *r_osdc; 174 struct kref r_kref; 175 bool r_mempool; 176 struct completion r_completion; /* private to osd_client.c */ 177 ceph_osdc_callback_t r_callback; 178 struct list_head r_unsafe_item; 179 180 struct inode *r_inode; /* for use by callbacks */ 181 void *r_priv; /* ditto */ 182 183 /* set by submitter */ 184 u64 r_snapid; /* for reads, CEPH_NOSNAP o/w */ 185 struct ceph_snap_context *r_snapc; /* for writes */ 186 struct timespec r_mtime; /* ditto */ 187 u64 r_data_offset; /* ditto */ 188 bool r_linger; /* don't resend on failure */ 189 190 /* internal */ 191 unsigned long r_stamp; /* jiffies, send or check time */ 192 unsigned long r_start_stamp; /* jiffies */ 193 int r_attempts; 194 struct ceph_eversion r_replay_version; /* aka reassert_version */ 195 u32 r_last_force_resend; 196 u32 r_map_dne_bound; 197 198 struct ceph_osd_req_op r_ops[]; 199 }; 200 201 struct ceph_request_redirect { 202 struct ceph_object_locator oloc; 203 }; 204 205 typedef void (*rados_watchcb2_t)(void *arg, u64 notify_id, u64 cookie, 206 u64 notifier_id, void *data, size_t data_len); 207 typedef void (*rados_watcherrcb_t)(void *arg, u64 cookie, int err); 208 209 struct ceph_osd_linger_request { 210 struct ceph_osd_client *osdc; 211 u64 linger_id; 212 bool committed; 213 bool is_watch; /* watch or notify */ 214 215 struct ceph_osd *osd; 216 struct ceph_osd_request *reg_req; 217 struct ceph_osd_request *ping_req; 218 unsigned long ping_sent; 219 unsigned long watch_valid_thru; 220 struct list_head pending_lworks; 221 222 struct ceph_osd_request_target t; 223 u32 last_force_resend; 224 u32 map_dne_bound; 225 226 struct timespec mtime; 227 228 struct kref kref; 229 struct mutex lock; 230 struct rb_node node; /* osd */ 231 struct rb_node osdc_node; /* osdc */ 232 struct rb_node mc_node; /* map check */ 233 struct list_head scan_item; 234 235 struct completion reg_commit_wait; 236 struct completion notify_finish_wait; 237 int reg_commit_error; 238 int notify_finish_error; 239 int last_error; 240 241 u32 register_gen; 242 u64 notify_id; 243 244 rados_watchcb2_t wcb; 245 rados_watcherrcb_t errcb; 246 void *data; 247 248 struct page ***preply_pages; 249 size_t *preply_len; 250 }; 251 252 struct ceph_watch_item { 253 struct ceph_entity_name name; 254 u64 cookie; 255 struct ceph_entity_addr addr; 256 }; 257 258 #define CEPH_LINGER_ID_START 0xffff000000000000ULL 259 260 struct ceph_osd_client { 261 struct ceph_client *client; 262 263 struct ceph_osdmap *osdmap; /* current map */ 264 struct rw_semaphore lock; 265 266 struct rb_root osds; /* osds */ 267 struct list_head osd_lru; /* idle osds */ 268 spinlock_t osd_lru_lock; 269 struct ceph_osd homeless_osd; 270 atomic64_t last_tid; /* tid of last request */ 271 u64 last_linger_id; 272 struct rb_root linger_requests; /* lingering requests */ 273 struct rb_root map_checks; 274 struct rb_root linger_map_checks; 275 atomic_t num_requests; 276 atomic_t num_homeless; 277 struct delayed_work timeout_work; 278 struct delayed_work osds_timeout_work; 279 #ifdef CONFIG_DEBUG_FS 280 struct dentry *debugfs_file; 281 #endif 282 283 mempool_t *req_mempool; 284 285 struct ceph_msgpool msgpool_op; 286 struct ceph_msgpool msgpool_op_reply; 287 288 struct workqueue_struct *notify_wq; 289 }; 290 291 static inline bool ceph_osdmap_flag(struct ceph_osd_client *osdc, int flag) 292 { 293 return osdc->osdmap->flags & flag; 294 } 295 296 extern int ceph_osdc_setup(void); 297 extern void ceph_osdc_cleanup(void); 298 299 extern int ceph_osdc_init(struct ceph_osd_client *osdc, 300 struct ceph_client *client); 301 extern void ceph_osdc_stop(struct ceph_osd_client *osdc); 302 303 extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc, 304 struct ceph_msg *msg); 305 extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc, 306 struct ceph_msg *msg); 307 308 extern void osd_req_op_init(struct ceph_osd_request *osd_req, 309 unsigned int which, u16 opcode, u32 flags); 310 311 extern void osd_req_op_raw_data_in_pages(struct ceph_osd_request *, 312 unsigned int which, 313 struct page **pages, u64 length, 314 u32 alignment, bool pages_from_pool, 315 bool own_pages); 316 317 extern void osd_req_op_extent_init(struct ceph_osd_request *osd_req, 318 unsigned int which, u16 opcode, 319 u64 offset, u64 length, 320 u64 truncate_size, u32 truncate_seq); 321 extern void osd_req_op_extent_update(struct ceph_osd_request *osd_req, 322 unsigned int which, u64 length); 323 extern void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req, 324 unsigned int which, u64 offset_inc); 325 326 extern struct ceph_osd_data *osd_req_op_extent_osd_data( 327 struct ceph_osd_request *osd_req, 328 unsigned int which); 329 330 extern void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *, 331 unsigned int which, 332 struct page **pages, u64 length, 333 u32 alignment, bool pages_from_pool, 334 bool own_pages); 335 extern void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *, 336 unsigned int which, 337 struct ceph_pagelist *pagelist); 338 #ifdef CONFIG_BLOCK 339 extern void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *, 340 unsigned int which, 341 struct bio *bio, size_t bio_length); 342 #endif /* CONFIG_BLOCK */ 343 344 extern void osd_req_op_cls_request_data_pagelist(struct ceph_osd_request *, 345 unsigned int which, 346 struct ceph_pagelist *pagelist); 347 extern void osd_req_op_cls_request_data_pages(struct ceph_osd_request *, 348 unsigned int which, 349 struct page **pages, u64 length, 350 u32 alignment, bool pages_from_pool, 351 bool own_pages); 352 extern void osd_req_op_cls_response_data_pages(struct ceph_osd_request *, 353 unsigned int which, 354 struct page **pages, u64 length, 355 u32 alignment, bool pages_from_pool, 356 bool own_pages); 357 extern void osd_req_op_cls_init(struct ceph_osd_request *osd_req, 358 unsigned int which, u16 opcode, 359 const char *class, const char *method); 360 extern int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which, 361 u16 opcode, const char *name, const void *value, 362 size_t size, u8 cmp_op, u8 cmp_mode); 363 extern void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req, 364 unsigned int which, 365 u64 expected_object_size, 366 u64 expected_write_size); 367 368 extern struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc, 369 struct ceph_snap_context *snapc, 370 unsigned int num_ops, 371 bool use_mempool, 372 gfp_t gfp_flags); 373 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp); 374 375 extern struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *, 376 struct ceph_file_layout *layout, 377 struct ceph_vino vino, 378 u64 offset, u64 *len, 379 unsigned int which, int num_ops, 380 int opcode, int flags, 381 struct ceph_snap_context *snapc, 382 u32 truncate_seq, u64 truncate_size, 383 bool use_mempool); 384 385 extern void ceph_osdc_get_request(struct ceph_osd_request *req); 386 extern void ceph_osdc_put_request(struct ceph_osd_request *req); 387 388 extern int ceph_osdc_start_request(struct ceph_osd_client *osdc, 389 struct ceph_osd_request *req, 390 bool nofail); 391 extern void ceph_osdc_cancel_request(struct ceph_osd_request *req); 392 extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc, 393 struct ceph_osd_request *req); 394 extern void ceph_osdc_sync(struct ceph_osd_client *osdc); 395 396 extern void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc); 397 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc); 398 399 int ceph_osdc_call(struct ceph_osd_client *osdc, 400 struct ceph_object_id *oid, 401 struct ceph_object_locator *oloc, 402 const char *class, const char *method, 403 unsigned int flags, 404 struct page *req_page, size_t req_len, 405 struct page *resp_page, size_t *resp_len); 406 407 extern int ceph_osdc_readpages(struct ceph_osd_client *osdc, 408 struct ceph_vino vino, 409 struct ceph_file_layout *layout, 410 u64 off, u64 *plen, 411 u32 truncate_seq, u64 truncate_size, 412 struct page **pages, int nr_pages, 413 int page_align); 414 415 extern int ceph_osdc_writepages(struct ceph_osd_client *osdc, 416 struct ceph_vino vino, 417 struct ceph_file_layout *layout, 418 struct ceph_snap_context *sc, 419 u64 off, u64 len, 420 u32 truncate_seq, u64 truncate_size, 421 struct timespec *mtime, 422 struct page **pages, int nr_pages); 423 424 /* watch/notify */ 425 struct ceph_osd_linger_request * 426 ceph_osdc_watch(struct ceph_osd_client *osdc, 427 struct ceph_object_id *oid, 428 struct ceph_object_locator *oloc, 429 rados_watchcb2_t wcb, 430 rados_watcherrcb_t errcb, 431 void *data); 432 int ceph_osdc_unwatch(struct ceph_osd_client *osdc, 433 struct ceph_osd_linger_request *lreq); 434 435 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc, 436 struct ceph_object_id *oid, 437 struct ceph_object_locator *oloc, 438 u64 notify_id, 439 u64 cookie, 440 void *payload, 441 size_t payload_len); 442 int ceph_osdc_notify(struct ceph_osd_client *osdc, 443 struct ceph_object_id *oid, 444 struct ceph_object_locator *oloc, 445 void *payload, 446 size_t payload_len, 447 u32 timeout, 448 struct page ***preply_pages, 449 size_t *preply_len); 450 int ceph_osdc_watch_check(struct ceph_osd_client *osdc, 451 struct ceph_osd_linger_request *lreq); 452 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc, 453 struct ceph_object_id *oid, 454 struct ceph_object_locator *oloc, 455 struct ceph_watch_item **watchers, 456 u32 *num_watchers); 457 #endif 458 459