xref: /openbmc/linux/include/linux/ceph/libceph.h (revision fbb6b31a)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_LIBCEPH_H
3 #define _FS_CEPH_LIBCEPH_H
4 
5 #include <linux/ceph/ceph_debug.h>
6 
7 #include <asm/unaligned.h>
8 #include <linux/backing-dev.h>
9 #include <linux/completion.h>
10 #include <linux/exportfs.h>
11 #include <linux/bug.h>
12 #include <linux/fs.h>
13 #include <linux/mempool.h>
14 #include <linux/pagemap.h>
15 #include <linux/wait.h>
16 #include <linux/writeback.h>
17 #include <linux/slab.h>
18 #include <linux/refcount.h>
19 
20 #include <linux/ceph/types.h>
21 #include <linux/ceph/messenger.h>
22 #include <linux/ceph/msgpool.h>
23 #include <linux/ceph/mon_client.h>
24 #include <linux/ceph/osd_client.h>
25 #include <linux/ceph/ceph_fs.h>
26 #include <linux/ceph/string_table.h>
27 
28 /*
29  * mount options
30  */
31 #define CEPH_OPT_FSID             (1<<0)
32 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
33 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
34 #define CEPH_OPT_NOCRC            (1<<3) /* no data crc on writes (msgr1) */
35 #define CEPH_OPT_TCP_NODELAY      (1<<4) /* TCP_NODELAY on TCP sockets */
36 #define CEPH_OPT_NOMSGSIGN        (1<<5) /* don't sign msgs (msgr1) */
37 #define CEPH_OPT_ABORT_ON_FULL    (1<<6) /* abort w/ ENOSPC when full */
38 #define CEPH_OPT_RXBOUNCE         (1<<7) /* double-buffer read data */
39 
40 #define CEPH_OPT_DEFAULT   (CEPH_OPT_TCP_NODELAY)
41 
42 #define ceph_set_opt(client, opt) \
43 	(client)->options->flags |= CEPH_OPT_##opt;
44 #define ceph_test_opt(client, opt) \
45 	(!!((client)->options->flags & CEPH_OPT_##opt))
46 
47 struct ceph_options {
48 	int flags;
49 	struct ceph_fsid fsid;
50 	struct ceph_entity_addr my_addr;
51 	unsigned long mount_timeout;		/* jiffies */
52 	unsigned long osd_idle_ttl;		/* jiffies */
53 	unsigned long osd_keepalive_timeout;	/* jiffies */
54 	unsigned long osd_request_timeout;	/* jiffies */
55 	u32 read_from_replica;  /* CEPH_OSD_FLAG_BALANCE/LOCALIZE_READS */
56 	int con_modes[2];  /* CEPH_CON_MODE_* */
57 
58 	/*
59 	 * any type that can't be simply compared or doesn't need
60 	 * to be compared should go beyond this point,
61 	 * ceph_compare_options() should be updated accordingly
62 	 */
63 
64 	struct ceph_entity_addr *mon_addr; /* should be the first
65 					      pointer type of args */
66 	int num_mon;
67 	char *name;
68 	struct ceph_crypto_key *key;
69 	struct rb_root crush_locs;
70 };
71 
72 /*
73  * defaults
74  */
75 #define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
76 #define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
77 #define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
78 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
79 #define CEPH_READ_FROM_REPLICA_DEFAULT	0  /* read from primary */
80 
81 #define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
82 #define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
83 #define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
84 #define CEPH_MONC_HUNT_BACKOFF		2
85 #define CEPH_MONC_HUNT_MAX_MULT		10
86 
87 #define CEPH_MSG_MAX_CONTROL_LEN (16*1024*1024)
88 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
89 #define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
90 
91 /*
92  * The largest possible rbd data object is 32M.
93  * The largest possible rbd object map object is 64M.
94  *
95  * There is no limit on the size of cephfs objects, but it has to obey
96  * rsize and wsize mount options anyway.
97  */
98 #define CEPH_MSG_MAX_DATA_LEN	(64*1024*1024)
99 
100 #define CEPH_AUTH_NAME_DEFAULT   "guest"
101 
102 /* mount state */
103 enum {
104 	CEPH_MOUNT_MOUNTING,
105 	CEPH_MOUNT_MOUNTED,
106 	CEPH_MOUNT_UNMOUNTING,
107 	CEPH_MOUNT_UNMOUNTED,
108 	CEPH_MOUNT_SHUTDOWN,
109 	CEPH_MOUNT_RECOVER,
110 };
111 
112 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
113 {
114 	return timeout ?: MAX_SCHEDULE_TIMEOUT;
115 }
116 
117 struct ceph_mds_client;
118 
119 /*
120  * per client state
121  *
122  * possibly shared by multiple mount points, if they are
123  * mounting the same ceph filesystem/cluster.
124  */
125 struct ceph_client {
126 	struct ceph_fsid fsid;
127 	bool have_fsid;
128 
129 	void *private;
130 
131 	struct ceph_options *options;
132 
133 	struct mutex mount_mutex;      /* serialize mount attempts */
134 	wait_queue_head_t auth_wq;
135 	int auth_err;
136 
137 	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
138 
139 	u64 supported_features;
140 	u64 required_features;
141 
142 	struct ceph_messenger msgr;   /* messenger instance */
143 	struct ceph_mon_client monc;
144 	struct ceph_osd_client osdc;
145 
146 #ifdef CONFIG_DEBUG_FS
147 	struct dentry *debugfs_dir;
148 	struct dentry *debugfs_monmap;
149 	struct dentry *debugfs_osdmap;
150 	struct dentry *debugfs_options;
151 #endif
152 };
153 
154 #define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
155 
156 static inline bool ceph_msgr2(struct ceph_client *client)
157 {
158 	return client->options->con_modes[0] != CEPH_CON_MODE_UNKNOWN;
159 }
160 
161 /*
162  * snapshots
163  */
164 
165 /*
166  * A "snap context" is the set of existing snapshots when we
167  * write data.  It is used by the OSD to guide its COW behavior.
168  *
169  * The ceph_snap_context is refcounted, and attached to each dirty
170  * page, indicating which context the dirty data belonged when it was
171  * dirtied.
172  */
173 struct ceph_snap_context {
174 	refcount_t nref;
175 	u64 seq;
176 	u32 num_snaps;
177 	u64 snaps[];
178 };
179 
180 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
181 					gfp_t gfp_flags);
182 extern struct ceph_snap_context *ceph_get_snap_context(
183 					struct ceph_snap_context *sc);
184 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
185 
186 /*
187  * calculate the number of pages a given length and offset map onto,
188  * if we align the data.
189  */
190 static inline int calc_pages_for(u64 off, u64 len)
191 {
192 	return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
193 		(off >> PAGE_SHIFT);
194 }
195 
196 #define RB_BYVAL(a)      (a)
197 #define RB_BYPTR(a)      (&(a))
198 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
199 
200 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
201 static bool __insert_##name(struct rb_root *root, type *t)		\
202 {									\
203 	struct rb_node **n = &root->rb_node;				\
204 	struct rb_node *parent = NULL;					\
205 									\
206 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
207 									\
208 	while (*n) {							\
209 		type *cur = rb_entry(*n, type, nodefld);		\
210 		int cmp;						\
211 									\
212 		parent = *n;						\
213 		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
214 		if (cmp < 0)						\
215 			n = &(*n)->rb_left;				\
216 		else if (cmp > 0)					\
217 			n = &(*n)->rb_right;				\
218 		else							\
219 			return false;					\
220 	}								\
221 									\
222 	rb_link_node(&t->nodefld, parent, n);				\
223 	rb_insert_color(&t->nodefld, root);				\
224 	return true;							\
225 }									\
226 static void __maybe_unused insert_##name(struct rb_root *root, type *t)	\
227 {									\
228 	if (!__insert_##name(root, t))					\
229 		BUG();							\
230 }									\
231 static void erase_##name(struct rb_root *root, type *t)			\
232 {									\
233 	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
234 	rb_erase(&t->nodefld, root);					\
235 	RB_CLEAR_NODE(&t->nodefld);					\
236 }
237 
238 /*
239  * @lookup_param_type is a parameter and not constructed from (@type,
240  * @keyfld) with typeof() because adding const is too unwieldy.
241  */
242 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
243 			       lookup_param_type, nodefld)		\
244 static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
245 {									\
246 	struct rb_node *n = root->rb_node;				\
247 									\
248 	while (n) {							\
249 		type *cur = rb_entry(n, type, nodefld);			\
250 		int cmp;						\
251 									\
252 		cmp = cmpexp(key, keyexp(cur->keyfld));			\
253 		if (cmp < 0)						\
254 			n = n->rb_left;					\
255 		else if (cmp > 0)					\
256 			n = n->rb_right;				\
257 		else							\
258 			return cur;					\
259 	}								\
260 									\
261 	return NULL;							\
262 }
263 
264 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
265 			 lookup_param_type, nodefld)			\
266 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
267 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
268 		       lookup_param_type, nodefld)
269 
270 /*
271  * Shorthands for integer keys.
272  */
273 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
274 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
275 
276 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
277 extern type __lookup_##name##_key;					\
278 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
279 		       typeof(__lookup_##name##_key.keyfld), nodefld)
280 
281 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
282 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
283 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
284 
285 extern struct kmem_cache *ceph_inode_cachep;
286 extern struct kmem_cache *ceph_cap_cachep;
287 extern struct kmem_cache *ceph_cap_snap_cachep;
288 extern struct kmem_cache *ceph_cap_flush_cachep;
289 extern struct kmem_cache *ceph_dentry_cachep;
290 extern struct kmem_cache *ceph_file_cachep;
291 extern struct kmem_cache *ceph_dir_file_cachep;
292 extern struct kmem_cache *ceph_mds_request_cachep;
293 extern mempool_t *ceph_wb_pagevec_pool;
294 
295 /* ceph_common.c */
296 extern bool libceph_compatible(void *data);
297 
298 extern const char *ceph_msg_type_name(int type);
299 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
300 extern int ceph_parse_fsid(const char *str, struct ceph_fsid *fsid);
301 
302 struct fs_parameter;
303 struct fc_log;
304 struct ceph_options *ceph_alloc_options(void);
305 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
306 		       struct fc_log *l, char delim);
307 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
308 		     struct fc_log *l);
309 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
310 			      bool show_all);
311 extern void ceph_destroy_options(struct ceph_options *opt);
312 extern int ceph_compare_options(struct ceph_options *new_opt,
313 				struct ceph_client *client);
314 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
315 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
316 u64 ceph_client_gid(struct ceph_client *client);
317 extern void ceph_destroy_client(struct ceph_client *client);
318 extern void ceph_reset_client_addr(struct ceph_client *client);
319 extern int __ceph_open_session(struct ceph_client *client,
320 			       unsigned long started);
321 extern int ceph_open_session(struct ceph_client *client);
322 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
323 				unsigned long timeout);
324 
325 /* pagevec.c */
326 extern void ceph_release_page_vector(struct page **pages, int num_pages);
327 extern void ceph_put_page_vector(struct page **pages, int num_pages,
328 				 bool dirty);
329 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
330 extern int ceph_copy_user_to_page_vector(struct page **pages,
331 					 const void __user *data,
332 					 loff_t off, size_t len);
333 extern void ceph_copy_to_page_vector(struct page **pages,
334 				    const void *data,
335 				    loff_t off, size_t len);
336 extern void ceph_copy_from_page_vector(struct page **pages,
337 				    void *data,
338 				    loff_t off, size_t len);
339 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
340 
341 
342 #endif /* _FS_CEPH_SUPER_H */
343