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 
19 struct ceph_msg;
20 struct ceph_snap_context;
21 struct ceph_osd_request;
22 struct ceph_osd_client;
23 
24 /*
25  * completion callback for async writepages
26  */
27 typedef 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 */
32 struct 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 
53 enum 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 
63 struct 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 
87 struct 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 
142 struct 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 */
169 struct 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 work_struct r_complete_work;
174 	struct ceph_osd *r_osd;
175 
176 	struct ceph_osd_request_target r_t;
177 #define r_base_oid	r_t.base_oid
178 #define r_base_oloc	r_t.base_oloc
179 #define r_flags		r_t.flags
180 
181 	struct ceph_msg  *r_request, *r_reply;
182 	u32               r_sent;      /* >0 if r_request is sending/sent */
183 
184 	/* request osd ops array  */
185 	unsigned int		r_num_ops;
186 
187 	int               r_result;
188 
189 	struct ceph_osd_client *r_osdc;
190 	struct kref       r_kref;
191 	bool              r_mempool;
192 	struct completion r_completion;       /* private to osd_client.c */
193 	ceph_osdc_callback_t r_callback;
194 	struct list_head  r_unsafe_item;
195 
196 	struct inode *r_inode;         	      /* for use by callbacks */
197 	void *r_priv;			      /* ditto */
198 
199 	/* set by submitter */
200 	u64 r_snapid;                         /* for reads, CEPH_NOSNAP o/w */
201 	struct ceph_snap_context *r_snapc;    /* for writes */
202 	struct timespec64 r_mtime;            /* ditto */
203 	u64 r_data_offset;                    /* ditto */
204 	bool r_linger;                        /* don't resend on failure */
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 
215 struct 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  */
224 struct ceph_osd_reqid {
225 	struct ceph_entity_name name;
226 	__le64 tid;
227 	__le32 inc;
228 } __packed;
229 
230 struct ceph_blkin_trace_info {
231 	__le64 trace_id;
232 	__le64 span_id;
233 	__le64 parent_span_id;
234 } __packed;
235 
236 typedef void (*rados_watchcb2_t)(void *arg, u64 notify_id, u64 cookie,
237 				 u64 notifier_id, void *data, size_t data_len);
238 typedef void (*rados_watcherrcb_t)(void *arg, u64 cookie, int err);
239 
240 struct 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 timespec64 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 
282 struct ceph_watch_item {
283 	struct ceph_entity_name name;
284 	u64 cookie;
285 	struct ceph_entity_addr addr;
286 };
287 
288 struct ceph_spg_mapping {
289 	struct rb_node node;
290 	struct ceph_spg spgid;
291 
292 	struct rb_root backoffs;
293 };
294 
295 struct 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 
ceph_hoid_build_hash_cache(struct ceph_hobject_id * hoid)311 static 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  */
320 struct 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 
332 struct 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 	bool                   abort_on_full; /* abort w/ ENOSPC when full */
351 	int                    abort_err;
352 	struct delayed_work    timeout_work;
353 	struct delayed_work    osds_timeout_work;
354 #ifdef CONFIG_DEBUG_FS
355 	struct dentry 	       *debugfs_file;
356 #endif
357 
358 	mempool_t              *req_mempool;
359 
360 	struct ceph_msgpool	msgpool_op;
361 	struct ceph_msgpool	msgpool_op_reply;
362 
363 	struct workqueue_struct	*notify_wq;
364 	struct workqueue_struct	*completion_wq;
365 };
366 
ceph_osdmap_flag(struct ceph_osd_client * osdc,int flag)367 static inline bool ceph_osdmap_flag(struct ceph_osd_client *osdc, int flag)
368 {
369 	return osdc->osdmap->flags & flag;
370 }
371 
372 extern int ceph_osdc_setup(void);
373 extern void ceph_osdc_cleanup(void);
374 
375 extern int ceph_osdc_init(struct ceph_osd_client *osdc,
376 			  struct ceph_client *client);
377 extern void ceph_osdc_stop(struct ceph_osd_client *osdc);
378 
379 extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc,
380 				   struct ceph_msg *msg);
381 extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc,
382 				 struct ceph_msg *msg);
383 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb);
384 void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err);
385 
386 extern void osd_req_op_init(struct ceph_osd_request *osd_req,
387 			    unsigned int which, u16 opcode, u32 flags);
388 
389 extern void osd_req_op_raw_data_in_pages(struct ceph_osd_request *,
390 					unsigned int which,
391 					struct page **pages, u64 length,
392 					u32 alignment, bool pages_from_pool,
393 					bool own_pages);
394 
395 extern void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
396 					unsigned int which, u16 opcode,
397 					u64 offset, u64 length,
398 					u64 truncate_size, u32 truncate_seq);
399 extern void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
400 					unsigned int which, u64 length);
401 extern void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
402 				       unsigned int which, u64 offset_inc);
403 
404 extern struct ceph_osd_data *osd_req_op_extent_osd_data(
405 					struct ceph_osd_request *osd_req,
406 					unsigned int which);
407 
408 extern void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *,
409 					unsigned int which,
410 					struct page **pages, u64 length,
411 					u32 alignment, bool pages_from_pool,
412 					bool own_pages);
413 extern void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *,
414 					unsigned int which,
415 					struct ceph_pagelist *pagelist);
416 #ifdef CONFIG_BLOCK
417 void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
418 				    unsigned int which,
419 				    struct ceph_bio_iter *bio_pos,
420 				    u32 bio_length);
421 #endif /* CONFIG_BLOCK */
422 void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
423 				      unsigned int which,
424 				      struct bio_vec *bvecs, u32 num_bvecs,
425 				      u32 bytes);
426 void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
427 					 unsigned int which,
428 					 struct ceph_bvec_iter *bvec_pos);
429 
430 extern void osd_req_op_cls_request_data_pagelist(struct ceph_osd_request *,
431 					unsigned int which,
432 					struct ceph_pagelist *pagelist);
433 extern void osd_req_op_cls_request_data_pages(struct ceph_osd_request *,
434 					unsigned int which,
435 					struct page **pages, u64 length,
436 					u32 alignment, bool pages_from_pool,
437 					bool own_pages);
438 void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
439 				       unsigned int which,
440 				       struct bio_vec *bvecs, u32 num_bvecs,
441 				       u32 bytes);
442 extern void osd_req_op_cls_response_data_pages(struct ceph_osd_request *,
443 					unsigned int which,
444 					struct page **pages, u64 length,
445 					u32 alignment, bool pages_from_pool,
446 					bool own_pages);
447 extern int osd_req_op_cls_init(struct ceph_osd_request *osd_req,
448 					unsigned int which, u16 opcode,
449 					const char *class, const char *method);
450 extern int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
451 				 u16 opcode, const char *name, const void *value,
452 				 size_t size, u8 cmp_op, u8 cmp_mode);
453 extern void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
454 				       unsigned int which,
455 				       u64 expected_object_size,
456 				       u64 expected_write_size);
457 
458 extern struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
459 					       struct ceph_snap_context *snapc,
460 					       unsigned int num_ops,
461 					       bool use_mempool,
462 					       gfp_t gfp_flags);
463 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp);
464 
465 extern struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *,
466 				      struct ceph_file_layout *layout,
467 				      struct ceph_vino vino,
468 				      u64 offset, u64 *len,
469 				      unsigned int which, int num_ops,
470 				      int opcode, int flags,
471 				      struct ceph_snap_context *snapc,
472 				      u32 truncate_seq, u64 truncate_size,
473 				      bool use_mempool);
474 
475 extern void ceph_osdc_get_request(struct ceph_osd_request *req);
476 extern void ceph_osdc_put_request(struct ceph_osd_request *req);
477 
478 extern int ceph_osdc_start_request(struct ceph_osd_client *osdc,
479 				   struct ceph_osd_request *req,
480 				   bool nofail);
481 extern void ceph_osdc_cancel_request(struct ceph_osd_request *req);
482 extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
483 				  struct ceph_osd_request *req);
484 extern void ceph_osdc_sync(struct ceph_osd_client *osdc);
485 
486 extern void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc);
487 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc);
488 
489 int ceph_osdc_call(struct ceph_osd_client *osdc,
490 		   struct ceph_object_id *oid,
491 		   struct ceph_object_locator *oloc,
492 		   const char *class, const char *method,
493 		   unsigned int flags,
494 		   struct page *req_page, size_t req_len,
495 		   struct page *resp_page, size_t *resp_len);
496 
497 extern int ceph_osdc_readpages(struct ceph_osd_client *osdc,
498 			       struct ceph_vino vino,
499 			       struct ceph_file_layout *layout,
500 			       u64 off, u64 *plen,
501 			       u32 truncate_seq, u64 truncate_size,
502 			       struct page **pages, int nr_pages,
503 			       int page_align);
504 
505 extern int ceph_osdc_writepages(struct ceph_osd_client *osdc,
506 				struct ceph_vino vino,
507 				struct ceph_file_layout *layout,
508 				struct ceph_snap_context *sc,
509 				u64 off, u64 len,
510 				u32 truncate_seq, u64 truncate_size,
511 				struct timespec64 *mtime,
512 				struct page **pages, int nr_pages);
513 
514 /* watch/notify */
515 struct ceph_osd_linger_request *
516 ceph_osdc_watch(struct ceph_osd_client *osdc,
517 		struct ceph_object_id *oid,
518 		struct ceph_object_locator *oloc,
519 		rados_watchcb2_t wcb,
520 		rados_watcherrcb_t errcb,
521 		void *data);
522 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
523 		      struct ceph_osd_linger_request *lreq);
524 
525 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
526 			 struct ceph_object_id *oid,
527 			 struct ceph_object_locator *oloc,
528 			 u64 notify_id,
529 			 u64 cookie,
530 			 void *payload,
531 			 u32 payload_len);
532 int ceph_osdc_notify(struct ceph_osd_client *osdc,
533 		     struct ceph_object_id *oid,
534 		     struct ceph_object_locator *oloc,
535 		     void *payload,
536 		     u32 payload_len,
537 		     u32 timeout,
538 		     struct page ***preply_pages,
539 		     size_t *preply_len);
540 int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
541 			  struct ceph_osd_linger_request *lreq);
542 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
543 			    struct ceph_object_id *oid,
544 			    struct ceph_object_locator *oloc,
545 			    struct ceph_watch_item **watchers,
546 			    u32 *num_watchers);
547 #endif
548 
549