xref: /linux-6.15/include/linux/ceph/osd_client.h (revision f5e4e7fd)
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/auth.h>
13 #include <linux/ceph/pagelist.h>
14 
15 /*
16  * Maximum object name size
17  * (must be at least as big as RBD_MAX_MD_NAME_LEN -- currently 100)
18  */
19 #define MAX_OBJ_NAME_SIZE 100
20 
21 struct ceph_msg;
22 struct ceph_snap_context;
23 struct ceph_osd_request;
24 struct ceph_osd_client;
25 struct ceph_authorizer;
26 
27 /*
28  * completion callback for async writepages
29  */
30 typedef void (*ceph_osdc_callback_t)(struct ceph_osd_request *,
31 				     struct ceph_msg *);
32 typedef void (*ceph_osdc_unsafe_callback_t)(struct ceph_osd_request *, bool);
33 
34 /* a given osd we're communicating with */
35 struct ceph_osd {
36 	atomic_t o_ref;
37 	struct ceph_osd_client *o_osdc;
38 	int o_osd;
39 	int o_incarnation;
40 	struct rb_node o_node;
41 	struct ceph_connection o_con;
42 	struct list_head o_requests;
43 	struct list_head o_linger_requests;
44 	struct list_head o_osd_lru;
45 	struct ceph_auth_handshake o_auth;
46 	unsigned long lru_ttl;
47 	int o_marked_for_keepalive;
48 	struct list_head o_keepalive_item;
49 };
50 
51 
52 #define CEPH_OSD_MAX_OP	2
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 };
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 bio	*bio;		/* list of bios */
77 			size_t		bio_length;	/* total in list */
78 		};
79 #endif /* CONFIG_BLOCK */
80 	};
81 };
82 
83 struct ceph_osd_req_op {
84 	u16 op;           /* CEPH_OSD_OP_* */
85 	u32 payload_len;
86 	union {
87 		struct ceph_osd_data raw_data_in;
88 		struct {
89 			u64 offset, length;
90 			u64 truncate_size;
91 			u32 truncate_seq;
92 			struct ceph_osd_data osd_data;
93 		} extent;
94 		struct {
95 			const char *class_name;
96 			const char *method_name;
97 			struct ceph_osd_data request_info;
98 			struct ceph_osd_data request_data;
99 			struct ceph_osd_data response_data;
100 			__u8 class_len;
101 			__u8 method_len;
102 			__u8 argc;
103 		} cls;
104 		struct {
105 			u64 cookie;
106 			u64 ver;
107 			u32 prot_ver;
108 			u32 timeout;
109 			__u8 flag;
110 		} watch;
111 	};
112 };
113 
114 /* an in-flight request */
115 struct ceph_osd_request {
116 	u64             r_tid;              /* unique for this client */
117 	struct rb_node  r_node;
118 	struct list_head r_req_lru_item;
119 	struct list_head r_osd_item;
120 	struct list_head r_linger_item;
121 	struct list_head r_linger_osd;
122 	struct ceph_osd *r_osd;
123 	struct ceph_pg   r_pgid;
124 	int              r_pg_osds[CEPH_PG_MAX_SIZE];
125 	int              r_num_pg_osds;
126 
127 	struct ceph_msg  *r_request, *r_reply;
128 	int               r_flags;     /* any additional flags for the osd */
129 	u32               r_sent;      /* >0 if r_request is sending/sent */
130 
131 	/* request osd ops array  */
132 	unsigned int		r_num_ops;
133 	struct ceph_osd_req_op	r_ops[CEPH_OSD_MAX_OP];
134 
135 	/* these are updated on each send */
136 	__le32           *r_request_osdmap_epoch;
137 	__le32           *r_request_flags;
138 	__le64           *r_request_pool;
139 	void             *r_request_pgid;
140 	__le32           *r_request_attempts;
141 	struct ceph_eversion *r_request_reassert_version;
142 
143 	int               r_result;
144 	int               r_reply_op_len[CEPH_OSD_MAX_OP];
145 	s32               r_reply_op_result[CEPH_OSD_MAX_OP];
146 	int               r_got_reply;
147 	int		  r_linger;
148 
149 	struct ceph_osd_client *r_osdc;
150 	struct kref       r_kref;
151 	bool              r_mempool;
152 	struct completion r_completion, r_safe_completion;
153 	ceph_osdc_callback_t r_callback;
154 	ceph_osdc_unsafe_callback_t r_unsafe_callback;
155 	struct ceph_eversion r_reassert_version;
156 	struct list_head  r_unsafe_item;
157 
158 	struct inode *r_inode;         	      /* for use by callbacks */
159 	void *r_priv;			      /* ditto */
160 
161 	char              r_oid[MAX_OBJ_NAME_SIZE];          /* object name */
162 	int               r_oid_len;
163 	u64               r_snapid;
164 	unsigned long     r_stamp;            /* send OR check time */
165 
166 	struct ceph_file_layout r_file_layout;
167 	struct ceph_snap_context *r_snapc;    /* snap context for writes */
168 };
169 
170 struct ceph_osd_event {
171 	u64 cookie;
172 	int one_shot;
173 	struct ceph_osd_client *osdc;
174 	void (*cb)(u64, u64, u8, void *);
175 	void *data;
176 	struct rb_node node;
177 	struct list_head osd_node;
178 	struct kref kref;
179 };
180 
181 struct ceph_osd_event_work {
182 	struct work_struct work;
183 	struct ceph_osd_event *event;
184         u64 ver;
185         u64 notify_id;
186         u8 opcode;
187 };
188 
189 struct ceph_osd_client {
190 	struct ceph_client     *client;
191 
192 	struct ceph_osdmap     *osdmap;       /* current map */
193 	struct rw_semaphore    map_sem;
194 	struct completion      map_waiters;
195 	u64                    last_requested_map;
196 
197 	struct mutex           request_mutex;
198 	struct rb_root         osds;          /* osds */
199 	struct list_head       osd_lru;       /* idle osds */
200 	u64                    timeout_tid;   /* tid of timeout triggering rq */
201 	u64                    last_tid;      /* tid of last request */
202 	struct rb_root         requests;      /* pending requests */
203 	struct list_head       req_lru;	      /* in-flight lru */
204 	struct list_head       req_unsent;    /* unsent/need-resend queue */
205 	struct list_head       req_notarget;  /* map to no osd */
206 	struct list_head       req_linger;    /* lingering requests */
207 	int                    num_requests;
208 	struct delayed_work    timeout_work;
209 	struct delayed_work    osds_timeout_work;
210 #ifdef CONFIG_DEBUG_FS
211 	struct dentry 	       *debugfs_file;
212 #endif
213 
214 	mempool_t              *req_mempool;
215 
216 	struct ceph_msgpool	msgpool_op;
217 	struct ceph_msgpool	msgpool_op_reply;
218 
219 	spinlock_t		event_lock;
220 	struct rb_root		event_tree;
221 	u64			event_count;
222 
223 	struct workqueue_struct	*notify_wq;
224 };
225 
226 extern int ceph_osdc_setup(void);
227 extern void ceph_osdc_cleanup(void);
228 
229 extern int ceph_osdc_init(struct ceph_osd_client *osdc,
230 			  struct ceph_client *client);
231 extern void ceph_osdc_stop(struct ceph_osd_client *osdc);
232 
233 extern void ceph_osdc_handle_reply(struct ceph_osd_client *osdc,
234 				   struct ceph_msg *msg);
235 extern void ceph_osdc_handle_map(struct ceph_osd_client *osdc,
236 				 struct ceph_msg *msg);
237 
238 extern void osd_req_op_init(struct ceph_osd_request *osd_req,
239 					unsigned int which, u16 opcode);
240 
241 extern void osd_req_op_raw_data_in_pages(struct ceph_osd_request *,
242 					unsigned int which,
243 					struct page **pages, u64 length,
244 					u32 alignment, bool pages_from_pool,
245 					bool own_pages);
246 
247 extern void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
248 					unsigned int which, u16 opcode,
249 					u64 offset, u64 length,
250 					u64 truncate_size, u32 truncate_seq);
251 extern void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
252 					unsigned int which, u64 length);
253 
254 extern struct ceph_osd_data *osd_req_op_extent_osd_data(
255 					struct ceph_osd_request *osd_req,
256 					unsigned int which);
257 extern struct ceph_osd_data *osd_req_op_cls_response_data(
258 					struct ceph_osd_request *osd_req,
259 					unsigned int which);
260 
261 extern void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *,
262 					unsigned int which,
263 					struct page **pages, u64 length,
264 					u32 alignment, bool pages_from_pool,
265 					bool own_pages);
266 extern void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *,
267 					unsigned int which,
268 					struct ceph_pagelist *pagelist);
269 #ifdef CONFIG_BLOCK
270 extern void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *,
271 					unsigned int which,
272 					struct bio *bio, size_t bio_length);
273 #endif /* CONFIG_BLOCK */
274 
275 extern void osd_req_op_cls_request_data_pagelist(struct ceph_osd_request *,
276 					unsigned int which,
277 					struct ceph_pagelist *pagelist);
278 extern void osd_req_op_cls_request_data_pages(struct ceph_osd_request *,
279 					unsigned int which,
280 					struct page **pages, u64 length,
281 					u32 alignment, bool pages_from_pool,
282 					bool own_pages);
283 extern void osd_req_op_cls_response_data_pages(struct ceph_osd_request *,
284 					unsigned int which,
285 					struct page **pages, u64 length,
286 					u32 alignment, bool pages_from_pool,
287 					bool own_pages);
288 
289 extern void osd_req_op_cls_init(struct ceph_osd_request *osd_req,
290 					unsigned int which, u16 opcode,
291 					const char *class, const char *method);
292 extern void osd_req_op_watch_init(struct ceph_osd_request *osd_req,
293 					unsigned int which, u16 opcode,
294 					u64 cookie, u64 version, int flag);
295 
296 extern struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
297 					       struct ceph_snap_context *snapc,
298 					       unsigned int num_ops,
299 					       bool use_mempool,
300 					       gfp_t gfp_flags);
301 
302 extern void ceph_osdc_build_request(struct ceph_osd_request *req, u64 off,
303 				    struct ceph_snap_context *snapc,
304 				    u64 snap_id,
305 				    struct timespec *mtime);
306 
307 extern struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *,
308 				      struct ceph_file_layout *layout,
309 				      struct ceph_vino vino,
310 				      u64 offset, u64 *len,
311 				      int num_ops, int opcode, int flags,
312 				      struct ceph_snap_context *snapc,
313 				      u32 truncate_seq, u64 truncate_size,
314 				      bool use_mempool);
315 
316 extern void ceph_osdc_set_request_linger(struct ceph_osd_client *osdc,
317 					 struct ceph_osd_request *req);
318 extern void ceph_osdc_unregister_linger_request(struct ceph_osd_client *osdc,
319 						struct ceph_osd_request *req);
320 
321 static inline void ceph_osdc_get_request(struct ceph_osd_request *req)
322 {
323 	kref_get(&req->r_kref);
324 }
325 extern void ceph_osdc_release_request(struct kref *kref);
326 static inline void ceph_osdc_put_request(struct ceph_osd_request *req)
327 {
328 	kref_put(&req->r_kref, ceph_osdc_release_request);
329 }
330 
331 extern int ceph_osdc_start_request(struct ceph_osd_client *osdc,
332 				   struct ceph_osd_request *req,
333 				   bool nofail);
334 extern int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
335 				  struct ceph_osd_request *req);
336 extern void ceph_osdc_sync(struct ceph_osd_client *osdc);
337 
338 extern int ceph_osdc_readpages(struct ceph_osd_client *osdc,
339 			       struct ceph_vino vino,
340 			       struct ceph_file_layout *layout,
341 			       u64 off, u64 *plen,
342 			       u32 truncate_seq, u64 truncate_size,
343 			       struct page **pages, int nr_pages,
344 			       int page_align);
345 
346 extern int ceph_osdc_writepages(struct ceph_osd_client *osdc,
347 				struct ceph_vino vino,
348 				struct ceph_file_layout *layout,
349 				struct ceph_snap_context *sc,
350 				u64 off, u64 len,
351 				u32 truncate_seq, u64 truncate_size,
352 				struct timespec *mtime,
353 				struct page **pages, int nr_pages);
354 
355 /* watch/notify events */
356 extern int ceph_osdc_create_event(struct ceph_osd_client *osdc,
357 				  void (*event_cb)(u64, u64, u8, void *),
358 				  void *data, struct ceph_osd_event **pevent);
359 extern void ceph_osdc_cancel_event(struct ceph_osd_event *event);
360 extern void ceph_osdc_put_event(struct ceph_osd_event *event);
361 #endif
362 
363