1 #ifndef _FS_CEPH_OSD_CLIENT_H
2 #define _FS_CEPH_OSD_CLIENT_H
4 #include <linux/bitrev.h>
5 #include <linux/completion.h>
6 #include <linux/kref.h>
7 #include <linux/mempool.h>
8 #include <linux/rbtree.h>
9 #include <linux/refcount.h>
11 #include <linux/ceph/types.h>
12 #include <linux/ceph/osdmap.h>
13 #include <linux/ceph/messenger.h>
14 #include <linux/ceph/msgpool.h>
15 #include <linux/ceph/auth.h>
16 #include <linux/ceph/pagelist.h>
19 struct ceph_snap_context;
20 struct ceph_osd_request;
21 struct ceph_osd_client;
24 * completion callback for async writepages
26 typedef void (*ceph_osdc_callback_t)(struct ceph_osd_request *);
28 #define CEPH_HOMELESS_OSD -1
30 /* a given osd we're communicating with */
33 struct ceph_osd_client *o_osdc;
36 struct rb_node o_node;
37 struct ceph_connection o_con;
38 struct rb_root o_requests;
39 struct rb_root o_linger_requests;
40 struct rb_root o_backoff_mappings;
41 struct rb_root o_backoffs_by_id;
42 struct list_head o_osd_lru;
43 struct ceph_auth_handshake o_auth;
44 unsigned long lru_ttl;
45 struct list_head o_keepalive_item;
49 #define CEPH_OSD_SLAB_OPS 2
50 #define CEPH_OSD_MAX_OPS 16
52 enum ceph_osd_data_type {
53 CEPH_OSD_DATA_TYPE_NONE = 0,
54 CEPH_OSD_DATA_TYPE_PAGES,
55 CEPH_OSD_DATA_TYPE_PAGELIST,
57 CEPH_OSD_DATA_TYPE_BIO,
58 #endif /* CONFIG_BLOCK */
61 struct ceph_osd_data {
62 enum ceph_osd_data_type type;
71 struct ceph_pagelist *pagelist;
74 struct bio *bio; /* list of bios */
75 size_t bio_length; /* total in list */
77 #endif /* CONFIG_BLOCK */
81 struct ceph_osd_req_op {
82 u16 op; /* CEPH_OSD_OP_* */
83 u32 flags; /* CEPH_OSD_OP_FLAG_* */
84 u32 indata_len; /* request */
85 u32 outdata_len; /* reply */
89 struct ceph_osd_data raw_data_in;
94 struct ceph_osd_data osd_data;
99 __u8 cmp_op; /* CEPH_OSD_CMPXATTR_OP_* */
100 __u8 cmp_mode; /* CEPH_OSD_CMPXATTR_MODE_* */
101 struct ceph_osd_data osd_data;
104 const char *class_name;
105 const char *method_name;
106 struct ceph_osd_data request_info;
107 struct ceph_osd_data request_data;
108 struct ceph_osd_data response_data;
115 __u8 op; /* CEPH_OSD_WATCH_OP_ */
119 struct ceph_osd_data request_data;
123 struct ceph_osd_data request_data;
124 struct ceph_osd_data response_data;
127 struct ceph_osd_data response_data;
130 u64 expected_object_size;
131 u64 expected_write_size;
136 struct ceph_osd_request_target {
137 struct ceph_object_id base_oid;
138 struct ceph_object_locator base_oloc;
139 struct ceph_object_id target_oid;
140 struct ceph_object_locator target_oloc;
142 struct ceph_pg pgid; /* last raw pg we mapped to */
143 struct ceph_spg spgid; /* last actual spg we mapped to */
146 struct ceph_osds acting;
152 unsigned int flags; /* CEPH_OSD_FLAG_* */
156 u32 last_force_resend;
161 /* an in-flight request */
162 struct ceph_osd_request {
163 u64 r_tid; /* unique for this client */
164 struct rb_node r_node;
165 struct rb_node r_mc_node; /* map check */
166 struct ceph_osd *r_osd;
168 struct ceph_osd_request_target r_t;
169 #define r_base_oid r_t.base_oid
170 #define r_base_oloc r_t.base_oloc
171 #define r_flags r_t.flags
173 struct ceph_msg *r_request, *r_reply;
174 u32 r_sent; /* >0 if r_request is sending/sent */
176 /* request osd ops array */
177 unsigned int r_num_ops;
181 struct ceph_osd_client *r_osdc;
184 struct completion r_completion; /* private to osd_client.c */
185 ceph_osdc_callback_t r_callback;
186 struct list_head r_unsafe_item;
188 struct inode *r_inode; /* for use by callbacks */
189 void *r_priv; /* ditto */
191 /* set by submitter */
192 u64 r_snapid; /* for reads, CEPH_NOSNAP o/w */
193 struct ceph_snap_context *r_snapc; /* for writes */
194 struct timespec r_mtime; /* ditto */
195 u64 r_data_offset; /* ditto */
196 bool r_linger; /* don't resend on failure */
197 bool r_abort_on_full; /* return ENOSPC when full */
200 unsigned long r_stamp; /* jiffies, send or check time */
201 unsigned long r_start_stamp; /* jiffies */
205 struct ceph_osd_req_op r_ops[];
208 struct ceph_request_redirect {
209 struct ceph_object_locator oloc;
213 * osd request identifier
215 * caller name + incarnation# + tid to unique identify this request
217 struct ceph_osd_reqid {
218 struct ceph_entity_name name;
223 struct ceph_blkin_trace_info {
226 __le64 parent_span_id;
229 typedef void (*rados_watchcb2_t)(void *arg, u64 notify_id, u64 cookie,
230 u64 notifier_id, void *data, size_t data_len);
231 typedef void (*rados_watcherrcb_t)(void *arg, u64 cookie, int err);
233 struct ceph_osd_linger_request {
234 struct ceph_osd_client *osdc;
237 bool is_watch; /* watch or notify */
239 struct ceph_osd *osd;
240 struct ceph_osd_request *reg_req;
241 struct ceph_osd_request *ping_req;
242 unsigned long ping_sent;
243 unsigned long watch_valid_thru;
244 struct list_head pending_lworks;
246 struct ceph_osd_request_target t;
249 struct timespec mtime;
253 struct rb_node node; /* osd */
254 struct rb_node osdc_node; /* osdc */
255 struct rb_node mc_node; /* map check */
256 struct list_head scan_item;
258 struct completion reg_commit_wait;
259 struct completion notify_finish_wait;
260 int reg_commit_error;
261 int notify_finish_error;
267 rados_watchcb2_t wcb;
268 rados_watcherrcb_t errcb;
271 struct page ***preply_pages;
275 struct ceph_watch_item {
276 struct ceph_entity_name name;
278 struct ceph_entity_addr addr;
281 struct ceph_spg_mapping {
283 struct ceph_spg spgid;
285 struct rb_root backoffs;
288 struct ceph_hobject_id {
301 u32 hash_reverse_bits;
304 static inline void ceph_hoid_build_hash_cache(struct ceph_hobject_id *hoid)
306 hoid->hash_reverse_bits = bitrev32(hoid->hash);
310 * PG-wide backoff: [begin, end)
311 * per-object backoff: begin == end
313 struct ceph_osd_backoff {
314 struct rb_node spg_node;
315 struct rb_node id_node;
317 struct ceph_spg spgid;
319 struct ceph_hobject_id *begin;
320 struct ceph_hobject_id *end;
323 #define CEPH_LINGER_ID_START 0xffff000000000000ULL
325 struct ceph_osd_client {
326 struct ceph_client *client;
328 struct ceph_osdmap *osdmap; /* current map */
329 struct rw_semaphore lock;
331 struct rb_root osds; /* osds */
332 struct list_head osd_lru; /* idle osds */
333 spinlock_t osd_lru_lock;
335 struct ceph_osd homeless_osd;
336 atomic64_t last_tid; /* tid of last request */
338 struct rb_root linger_requests; /* lingering requests */
339 struct rb_root map_checks;
340 struct rb_root linger_map_checks;
341 atomic_t num_requests;
342 atomic_t num_homeless;
343 struct delayed_work timeout_work;
344 struct delayed_work osds_timeout_work;
345 #ifdef CONFIG_DEBUG_FS
346 struct dentry *debugfs_file;
349 mempool_t *req_mempool;
351 struct ceph_msgpool msgpool_op;
352 struct ceph_msgpool msgpool_op_reply;
354 struct workqueue_struct *notify_wq;
357 static inline bool ceph_osdmap_flag(struct ceph_osd_client *osdc, int flag)
359 return osdc->osdmap->flags & flag;
362 extern int ceph_osdc_setup(void);
363 extern void ceph_osdc_cleanup(void);
365 extern int ceph_osdc_init(struct ceph_osd_client *osdc,
366 struct ceph_client *client);
367 extern void ceph_osdc_stop(struct ceph_osd_client *osdc);
369 extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc,
370 struct ceph_msg *msg);
371 extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc,
372 struct ceph_msg *msg);
373 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb);
375 extern void osd_req_op_init(struct ceph_osd_request *osd_req,
376 unsigned int which, u16 opcode, u32 flags);
378 extern void osd_req_op_raw_data_in_pages(struct ceph_osd_request *,
380 struct page **pages, u64 length,
381 u32 alignment, bool pages_from_pool,
384 extern void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
385 unsigned int which, u16 opcode,
386 u64 offset, u64 length,
387 u64 truncate_size, u32 truncate_seq);
388 extern void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
389 unsigned int which, u64 length);
390 extern void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
391 unsigned int which, u64 offset_inc);
393 extern struct ceph_osd_data *osd_req_op_extent_osd_data(
394 struct ceph_osd_request *osd_req,
397 extern void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *,
399 struct page **pages, u64 length,
400 u32 alignment, bool pages_from_pool,
402 extern void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *,
404 struct ceph_pagelist *pagelist);
406 extern void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *,
408 struct bio *bio, size_t bio_length);
409 #endif /* CONFIG_BLOCK */
411 extern void osd_req_op_cls_request_data_pagelist(struct ceph_osd_request *,
413 struct ceph_pagelist *pagelist);
414 extern void osd_req_op_cls_request_data_pages(struct ceph_osd_request *,
416 struct page **pages, u64 length,
417 u32 alignment, bool pages_from_pool,
419 extern void osd_req_op_cls_response_data_pages(struct ceph_osd_request *,
421 struct page **pages, u64 length,
422 u32 alignment, bool pages_from_pool,
424 extern void osd_req_op_cls_init(struct ceph_osd_request *osd_req,
425 unsigned int which, u16 opcode,
426 const char *class, const char *method);
427 extern int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
428 u16 opcode, const char *name, const void *value,
429 size_t size, u8 cmp_op, u8 cmp_mode);
430 extern void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
432 u64 expected_object_size,
433 u64 expected_write_size);
435 extern struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
436 struct ceph_snap_context *snapc,
437 unsigned int num_ops,
440 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp);
442 extern struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *,
443 struct ceph_file_layout *layout,
444 struct ceph_vino vino,
445 u64 offset, u64 *len,
446 unsigned int which, int num_ops,
447 int opcode, int flags,
448 struct ceph_snap_context *snapc,
449 u32 truncate_seq, u64 truncate_size,
452 extern void ceph_osdc_get_request(struct ceph_osd_request *req);
453 extern void ceph_osdc_put_request(struct ceph_osd_request *req);
455 extern int ceph_osdc_start_request(struct ceph_osd_client *osdc,
456 struct ceph_osd_request *req,
458 extern void ceph_osdc_cancel_request(struct ceph_osd_request *req);
459 extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
460 struct ceph_osd_request *req);
461 extern void ceph_osdc_sync(struct ceph_osd_client *osdc);
463 extern void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc);
464 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc);
466 int ceph_osdc_call(struct ceph_osd_client *osdc,
467 struct ceph_object_id *oid,
468 struct ceph_object_locator *oloc,
469 const char *class, const char *method,
471 struct page *req_page, size_t req_len,
472 struct page *resp_page, size_t *resp_len);
474 extern int ceph_osdc_readpages(struct ceph_osd_client *osdc,
475 struct ceph_vino vino,
476 struct ceph_file_layout *layout,
478 u32 truncate_seq, u64 truncate_size,
479 struct page **pages, int nr_pages,
482 extern int ceph_osdc_writepages(struct ceph_osd_client *osdc,
483 struct ceph_vino vino,
484 struct ceph_file_layout *layout,
485 struct ceph_snap_context *sc,
487 u32 truncate_seq, u64 truncate_size,
488 struct timespec *mtime,
489 struct page **pages, int nr_pages);
492 struct ceph_osd_linger_request *
493 ceph_osdc_watch(struct ceph_osd_client *osdc,
494 struct ceph_object_id *oid,
495 struct ceph_object_locator *oloc,
496 rados_watchcb2_t wcb,
497 rados_watcherrcb_t errcb,
499 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
500 struct ceph_osd_linger_request *lreq);
502 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
503 struct ceph_object_id *oid,
504 struct ceph_object_locator *oloc,
509 int ceph_osdc_notify(struct ceph_osd_client *osdc,
510 struct ceph_object_id *oid,
511 struct ceph_object_locator *oloc,
515 struct page ***preply_pages,
517 int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
518 struct ceph_osd_linger_request *lreq);
519 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
520 struct ceph_object_id *oid,
521 struct ceph_object_locator *oloc,
522 struct ceph_watch_item **watchers,