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