1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_OSD_CLIENT_H
3 #define _FS_CEPH_OSD_CLIENT_H
4
5 #include <linux/bitrev.h>
6 #include <linux/completion.h>
7 #include <linux/kref.h>
8 #include <linux/mempool.h>
9 #include <linux/rbtree.h>
10 #include <linux/refcount.h>
11 #include <linux/ktime.h>
12
13 #include <linux/ceph/types.h>
14 #include <linux/ceph/osdmap.h>
15 #include <linux/ceph/messenger.h>
16 #include <linux/ceph/msgpool.h>
17 #include <linux/ceph/auth.h>
18 #include <linux/ceph/pagelist.h>
19
20 struct ceph_msg;
21 struct ceph_snap_context;
22 struct ceph_osd_request;
23 struct ceph_osd_client;
24
25 /*
26 * completion callback for async writepages
27 */
28 typedef void (*ceph_osdc_callback_t)(struct ceph_osd_request *);
29
30 #define CEPH_HOMELESS_OSD -1
31
32 /* a given osd we're communicating with */
33 struct ceph_osd {
34 refcount_t o_ref;
35 struct ceph_osd_client *o_osdc;
36 int o_osd;
37 int o_incarnation;
38 struct rb_node o_node;
39 struct ceph_connection o_con;
40 struct rb_root o_requests;
41 struct rb_root o_linger_requests;
42 struct rb_root o_backoff_mappings;
43 struct rb_root o_backoffs_by_id;
44 struct list_head o_osd_lru;
45 struct ceph_auth_handshake o_auth;
46 unsigned long lru_ttl;
47 struct list_head o_keepalive_item;
48 struct mutex lock;
49 };
50
51 #define CEPH_OSD_SLAB_OPS 2
52 #define CEPH_OSD_MAX_OPS 16
53
54 enum ceph_osd_data_type {
55 CEPH_OSD_DATA_TYPE_NONE = 0,
56 CEPH_OSD_DATA_TYPE_PAGES,
57 CEPH_OSD_DATA_TYPE_PAGELIST,
58 #ifdef CONFIG_BLOCK
59 CEPH_OSD_DATA_TYPE_BIO,
60 #endif /* CONFIG_BLOCK */
61 CEPH_OSD_DATA_TYPE_BVECS,
62 };
63
64 struct ceph_osd_data {
65 enum ceph_osd_data_type type;
66 union {
67 struct {
68 struct page **pages;
69 u64 length;
70 u32 alignment;
71 bool pages_from_pool;
72 bool own_pages;
73 };
74 struct ceph_pagelist *pagelist;
75 #ifdef CONFIG_BLOCK
76 struct {
77 struct ceph_bio_iter bio_pos;
78 u32 bio_length;
79 };
80 #endif /* CONFIG_BLOCK */
81 struct {
82 struct ceph_bvec_iter bvec_pos;
83 u32 num_bvecs;
84 };
85 };
86 };
87
88 struct ceph_osd_req_op {
89 u16 op; /* CEPH_OSD_OP_* */
90 u32 flags; /* CEPH_OSD_OP_FLAG_* */
91 u32 indata_len; /* request */
92 u32 outdata_len; /* reply */
93 s32 rval;
94
95 union {
96 struct ceph_osd_data raw_data_in;
97 struct {
98 u64 offset, length;
99 u64 truncate_size;
100 u32 truncate_seq;
101 struct ceph_osd_data osd_data;
102 } extent;
103 struct {
104 u32 name_len;
105 u32 value_len;
106 __u8 cmp_op; /* CEPH_OSD_CMPXATTR_OP_* */
107 __u8 cmp_mode; /* CEPH_OSD_CMPXATTR_MODE_* */
108 struct ceph_osd_data osd_data;
109 } xattr;
110 struct {
111 const char *class_name;
112 const char *method_name;
113 struct ceph_osd_data request_info;
114 struct ceph_osd_data request_data;
115 struct ceph_osd_data response_data;
116 __u8 class_len;
117 __u8 method_len;
118 u32 indata_len;
119 } cls;
120 struct {
121 u64 cookie;
122 __u8 op; /* CEPH_OSD_WATCH_OP_ */
123 u32 gen;
124 } watch;
125 struct {
126 struct ceph_osd_data request_data;
127 } notify_ack;
128 struct {
129 u64 cookie;
130 struct ceph_osd_data request_data;
131 struct ceph_osd_data response_data;
132 } notify;
133 struct {
134 struct ceph_osd_data response_data;
135 } list_watchers;
136 struct {
137 u64 expected_object_size;
138 u64 expected_write_size;
139 u32 flags; /* CEPH_OSD_OP_ALLOC_HINT_FLAG_* */
140 } alloc_hint;
141 struct {
142 u64 snapid;
143 u64 src_version;
144 u8 flags;
145 u32 src_fadvise_flags;
146 struct ceph_osd_data osd_data;
147 } copy_from;
148 };
149 };
150
151 struct ceph_osd_request_target {
152 struct ceph_object_id base_oid;
153 struct ceph_object_locator base_oloc;
154 struct ceph_object_id target_oid;
155 struct ceph_object_locator target_oloc;
156
157 struct ceph_pg pgid; /* last raw pg we mapped to */
158 struct ceph_spg spgid; /* last actual spg we mapped to */
159 u32 pg_num;
160 u32 pg_num_mask;
161 struct ceph_osds acting;
162 struct ceph_osds up;
163 int size;
164 int min_size;
165 bool sort_bitwise;
166 bool recovery_deletes;
167
168 unsigned int flags; /* CEPH_OSD_FLAG_* */
169 bool used_replica;
170 bool paused;
171
172 u32 epoch;
173 u32 last_force_resend;
174
175 int osd;
176 };
177
178 /* an in-flight request */
179 struct ceph_osd_request {
180 u64 r_tid; /* unique for this client */
181 struct rb_node r_node;
182 struct rb_node r_mc_node; /* map check */
183 struct work_struct r_complete_work;
184 struct ceph_osd *r_osd;
185
186 struct ceph_osd_request_target r_t;
187 #define r_base_oid r_t.base_oid
188 #define r_base_oloc r_t.base_oloc
189 #define r_flags r_t.flags
190
191 struct ceph_msg *r_request, *r_reply;
192 u32 r_sent; /* >0 if r_request is sending/sent */
193
194 /* request osd ops array */
195 unsigned int r_num_ops;
196
197 int r_result;
198
199 struct ceph_osd_client *r_osdc;
200 struct kref r_kref;
201 bool r_mempool;
202 struct completion r_completion; /* private to osd_client.c */
203 ceph_osdc_callback_t r_callback;
204
205 struct inode *r_inode; /* for use by callbacks */
206 struct list_head r_private_item; /* ditto */
207 void *r_priv; /* ditto */
208
209 /* set by submitter */
210 u64 r_snapid; /* for reads, CEPH_NOSNAP o/w */
211 struct ceph_snap_context *r_snapc; /* for writes */
212 struct timespec64 r_mtime; /* ditto */
213 u64 r_data_offset; /* ditto */
214 bool r_linger; /* don't resend on failure */
215
216 /* internal */
217 unsigned long r_stamp; /* jiffies, send or check time */
218 unsigned long r_start_stamp; /* jiffies */
219 ktime_t r_start_latency; /* ktime_t */
220 ktime_t r_end_latency; /* ktime_t */
221 int r_attempts;
222 u32 r_map_dne_bound;
223
224 struct ceph_osd_req_op r_ops[];
225 };
226
227 struct ceph_request_redirect {
228 struct ceph_object_locator oloc;
229 };
230
231 /*
232 * osd request identifier
233 *
234 * caller name + incarnation# + tid to unique identify this request
235 */
236 struct ceph_osd_reqid {
237 struct ceph_entity_name name;
238 __le64 tid;
239 __le32 inc;
240 } __packed;
241
242 struct ceph_blkin_trace_info {
243 __le64 trace_id;
244 __le64 span_id;
245 __le64 parent_span_id;
246 } __packed;
247
248 typedef void (*rados_watchcb2_t)(void *arg, u64 notify_id, u64 cookie,
249 u64 notifier_id, void *data, size_t data_len);
250 typedef void (*rados_watcherrcb_t)(void *arg, u64 cookie, int err);
251
252 struct ceph_osd_linger_request {
253 struct ceph_osd_client *osdc;
254 u64 linger_id;
255 bool committed;
256 bool is_watch; /* watch or notify */
257
258 struct ceph_osd *osd;
259 struct ceph_osd_request *reg_req;
260 struct ceph_osd_request *ping_req;
261 unsigned long ping_sent;
262 unsigned long watch_valid_thru;
263 struct list_head pending_lworks;
264
265 struct ceph_osd_request_target t;
266 u32 map_dne_bound;
267
268 struct timespec64 mtime;
269
270 struct kref kref;
271 struct mutex lock;
272 struct rb_node node; /* osd */
273 struct rb_node osdc_node; /* osdc */
274 struct rb_node mc_node; /* map check */
275 struct list_head scan_item;
276
277 struct completion reg_commit_wait;
278 struct completion notify_finish_wait;
279 int reg_commit_error;
280 int notify_finish_error;
281 int last_error;
282
283 u32 register_gen;
284 u64 notify_id;
285
286 rados_watchcb2_t wcb;
287 rados_watcherrcb_t errcb;
288 void *data;
289
290 struct page ***preply_pages;
291 size_t *preply_len;
292 };
293
294 struct ceph_watch_item {
295 struct ceph_entity_name name;
296 u64 cookie;
297 struct ceph_entity_addr addr;
298 };
299
300 struct ceph_spg_mapping {
301 struct rb_node node;
302 struct ceph_spg spgid;
303
304 struct rb_root backoffs;
305 };
306
307 struct ceph_hobject_id {
308 void *key;
309 size_t key_len;
310 void *oid;
311 size_t oid_len;
312 u64 snapid;
313 u32 hash;
314 u8 is_max;
315 void *nspace;
316 size_t nspace_len;
317 s64 pool;
318
319 /* cache */
320 u32 hash_reverse_bits;
321 };
322
ceph_hoid_build_hash_cache(struct ceph_hobject_id * hoid)323 static inline void ceph_hoid_build_hash_cache(struct ceph_hobject_id *hoid)
324 {
325 hoid->hash_reverse_bits = bitrev32(hoid->hash);
326 }
327
328 /*
329 * PG-wide backoff: [begin, end)
330 * per-object backoff: begin == end
331 */
332 struct ceph_osd_backoff {
333 struct rb_node spg_node;
334 struct rb_node id_node;
335
336 struct ceph_spg spgid;
337 u64 id;
338 struct ceph_hobject_id *begin;
339 struct ceph_hobject_id *end;
340 };
341
342 #define CEPH_LINGER_ID_START 0xffff000000000000ULL
343
344 struct ceph_osd_client {
345 struct ceph_client *client;
346
347 struct ceph_osdmap *osdmap; /* current map */
348 struct rw_semaphore lock;
349
350 struct rb_root osds; /* osds */
351 struct list_head osd_lru; /* idle osds */
352 spinlock_t osd_lru_lock;
353 u32 epoch_barrier;
354 struct ceph_osd homeless_osd;
355 atomic64_t last_tid; /* tid of last request */
356 u64 last_linger_id;
357 struct rb_root linger_requests; /* lingering requests */
358 struct rb_root map_checks;
359 struct rb_root linger_map_checks;
360 atomic_t num_requests;
361 atomic_t num_homeless;
362 int abort_err;
363 struct delayed_work timeout_work;
364 struct delayed_work osds_timeout_work;
365 #ifdef CONFIG_DEBUG_FS
366 struct dentry *debugfs_file;
367 #endif
368
369 mempool_t *req_mempool;
370
371 struct ceph_msgpool msgpool_op;
372 struct ceph_msgpool msgpool_op_reply;
373
374 struct workqueue_struct *notify_wq;
375 struct workqueue_struct *completion_wq;
376 };
377
ceph_osdmap_flag(struct ceph_osd_client * osdc,int flag)378 static inline bool ceph_osdmap_flag(struct ceph_osd_client *osdc, int flag)
379 {
380 return osdc->osdmap->flags & flag;
381 }
382
383 extern int ceph_osdc_setup(void);
384 extern void ceph_osdc_cleanup(void);
385
386 extern int ceph_osdc_init(struct ceph_osd_client *osdc,
387 struct ceph_client *client);
388 extern void ceph_osdc_stop(struct ceph_osd_client *osdc);
389 extern void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc);
390
391 extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc,
392 struct ceph_msg *msg);
393 extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc,
394 struct ceph_msg *msg);
395 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb);
396 void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err);
397 void ceph_osdc_clear_abort_err(struct ceph_osd_client *osdc);
398
399 #define osd_req_op_data(oreq, whch, typ, fld) \
400 ({ \
401 struct ceph_osd_request *__oreq = (oreq); \
402 unsigned int __whch = (whch); \
403 BUG_ON(__whch >= __oreq->r_num_ops); \
404 &__oreq->r_ops[__whch].typ.fld; \
405 })
406
407 struct ceph_osd_req_op *osd_req_op_init(struct ceph_osd_request *osd_req,
408 unsigned int which, u16 opcode, u32 flags);
409
410 extern void osd_req_op_raw_data_in_pages(struct ceph_osd_request *,
411 unsigned int which,
412 struct page **pages, u64 length,
413 u32 alignment, bool pages_from_pool,
414 bool own_pages);
415
416 extern void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
417 unsigned int which, u16 opcode,
418 u64 offset, u64 length,
419 u64 truncate_size, u32 truncate_seq);
420 extern void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
421 unsigned int which, u64 length);
422 extern void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
423 unsigned int which, u64 offset_inc);
424
425 extern struct ceph_osd_data *osd_req_op_extent_osd_data(
426 struct ceph_osd_request *osd_req,
427 unsigned int which);
428
429 extern void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *,
430 unsigned int which,
431 struct page **pages, u64 length,
432 u32 alignment, bool pages_from_pool,
433 bool own_pages);
434 extern void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *,
435 unsigned int which,
436 struct ceph_pagelist *pagelist);
437 #ifdef CONFIG_BLOCK
438 void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
439 unsigned int which,
440 struct ceph_bio_iter *bio_pos,
441 u32 bio_length);
442 #endif /* CONFIG_BLOCK */
443 void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
444 unsigned int which,
445 struct bio_vec *bvecs, u32 num_bvecs,
446 u32 bytes);
447 void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
448 unsigned int which,
449 struct ceph_bvec_iter *bvec_pos);
450
451 extern void osd_req_op_cls_request_data_pagelist(struct ceph_osd_request *,
452 unsigned int which,
453 struct ceph_pagelist *pagelist);
454 extern void osd_req_op_cls_request_data_pages(struct ceph_osd_request *,
455 unsigned int which,
456 struct page **pages, u64 length,
457 u32 alignment, bool pages_from_pool,
458 bool own_pages);
459 void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
460 unsigned int which,
461 struct bio_vec *bvecs, u32 num_bvecs,
462 u32 bytes);
463 extern void osd_req_op_cls_response_data_pages(struct ceph_osd_request *,
464 unsigned int which,
465 struct page **pages, u64 length,
466 u32 alignment, bool pages_from_pool,
467 bool own_pages);
468 int osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
469 const char *class, const char *method);
470 extern int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
471 u16 opcode, const char *name, const void *value,
472 size_t size, u8 cmp_op, u8 cmp_mode);
473 extern void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
474 unsigned int which,
475 u64 expected_object_size,
476 u64 expected_write_size,
477 u32 flags);
478
479 extern struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
480 struct ceph_snap_context *snapc,
481 unsigned int num_ops,
482 bool use_mempool,
483 gfp_t gfp_flags);
484 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp);
485
486 extern struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *,
487 struct ceph_file_layout *layout,
488 struct ceph_vino vino,
489 u64 offset, u64 *len,
490 unsigned int which, int num_ops,
491 int opcode, int flags,
492 struct ceph_snap_context *snapc,
493 u32 truncate_seq, u64 truncate_size,
494 bool use_mempool);
495
496 extern void ceph_osdc_get_request(struct ceph_osd_request *req);
497 extern void ceph_osdc_put_request(struct ceph_osd_request *req);
498
499 extern int ceph_osdc_start_request(struct ceph_osd_client *osdc,
500 struct ceph_osd_request *req,
501 bool nofail);
502 extern void ceph_osdc_cancel_request(struct ceph_osd_request *req);
503 extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
504 struct ceph_osd_request *req);
505 extern void ceph_osdc_sync(struct ceph_osd_client *osdc);
506
507 extern void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc);
508 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc);
509
510 int ceph_osdc_call(struct ceph_osd_client *osdc,
511 struct ceph_object_id *oid,
512 struct ceph_object_locator *oloc,
513 const char *class, const char *method,
514 unsigned int flags,
515 struct page *req_page, size_t req_len,
516 struct page **resp_pages, size_t *resp_len);
517
518 int ceph_osdc_copy_from(struct ceph_osd_client *osdc,
519 u64 src_snapid, u64 src_version,
520 struct ceph_object_id *src_oid,
521 struct ceph_object_locator *src_oloc,
522 u32 src_fadvise_flags,
523 struct ceph_object_id *dst_oid,
524 struct ceph_object_locator *dst_oloc,
525 u32 dst_fadvise_flags,
526 u32 truncate_seq, u64 truncate_size,
527 u8 copy_from_flags);
528
529 /* watch/notify */
530 struct ceph_osd_linger_request *
531 ceph_osdc_watch(struct ceph_osd_client *osdc,
532 struct ceph_object_id *oid,
533 struct ceph_object_locator *oloc,
534 rados_watchcb2_t wcb,
535 rados_watcherrcb_t errcb,
536 void *data);
537 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
538 struct ceph_osd_linger_request *lreq);
539
540 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
541 struct ceph_object_id *oid,
542 struct ceph_object_locator *oloc,
543 u64 notify_id,
544 u64 cookie,
545 void *payload,
546 u32 payload_len);
547 int ceph_osdc_notify(struct ceph_osd_client *osdc,
548 struct ceph_object_id *oid,
549 struct ceph_object_locator *oloc,
550 void *payload,
551 u32 payload_len,
552 u32 timeout,
553 struct page ***preply_pages,
554 size_t *preply_len);
555 int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
556 struct ceph_osd_linger_request *lreq);
557 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
558 struct ceph_object_id *oid,
559 struct ceph_object_locator *oloc,
560 struct ceph_watch_item **watchers,
561 u32 *num_watchers);
562 #endif
563
564
