xref: /linux/include/linux/ceph/libceph.h (revision 3bdab16c55f57a24245c97d707241dd9b48d1a91)
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 */
35 #define CEPH_OPT_NOMSGAUTH	  (1<<4) /* don't require msg signing feat */
36 #define CEPH_OPT_TCP_NODELAY	  (1<<5) /* TCP_NODELAY on TCP sockets */
37 #define CEPH_OPT_NOMSGSIGN	  (1<<6) /* don't sign msgs */
38 #define CEPH_OPT_ABORT_ON_FULL	  (1<<7) /* abort w/ ENOSPC when full */
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 
56 	/*
57 	 * any type that can't be simply compared or doesn't need
58 	 * to be compared should go beyond this point,
59 	 * ceph_compare_options() should be updated accordingly
60 	 */
61 
62 	struct ceph_entity_addr *mon_addr; /* should be the first
63 					      pointer type of args */
64 	int num_mon;
65 	char *name;
66 	struct ceph_crypto_key *key;
67 };
68 
69 /*
70  * defaults
71  */
72 #define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
73 #define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
74 #define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
75 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
76 
77 #define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
78 #define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
79 #define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
80 #define CEPH_MONC_HUNT_BACKOFF		2
81 #define CEPH_MONC_HUNT_MAX_MULT		10
82 
83 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
84 #define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
85 
86 /*
87  * Handle the largest possible rbd object in one message.
88  * There is no limit on the size of cephfs objects, but it has to obey
89  * rsize and wsize mount options anyway.
90  */
91 #define CEPH_MSG_MAX_DATA_LEN	(32*1024*1024)
92 
93 #define CEPH_AUTH_NAME_DEFAULT   "guest"
94 
95 /* mount state */
96 enum {
97 	CEPH_MOUNT_MOUNTING,
98 	CEPH_MOUNT_MOUNTED,
99 	CEPH_MOUNT_UNMOUNTING,
100 	CEPH_MOUNT_UNMOUNTED,
101 	CEPH_MOUNT_SHUTDOWN,
102 };
103 
104 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
105 {
106 	return timeout ?: MAX_SCHEDULE_TIMEOUT;
107 }
108 
109 struct ceph_mds_client;
110 
111 /*
112  * per client state
113  *
114  * possibly shared by multiple mount points, if they are
115  * mounting the same ceph filesystem/cluster.
116  */
117 struct ceph_client {
118 	struct ceph_fsid fsid;
119 	bool have_fsid;
120 
121 	void *private;
122 
123 	struct ceph_options *options;
124 
125 	struct mutex mount_mutex;      /* serialize mount attempts */
126 	wait_queue_head_t auth_wq;
127 	int auth_err;
128 
129 	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
130 
131 	u64 supported_features;
132 	u64 required_features;
133 
134 	struct ceph_messenger msgr;   /* messenger instance */
135 	struct ceph_mon_client monc;
136 	struct ceph_osd_client osdc;
137 
138 #ifdef CONFIG_DEBUG_FS
139 	struct dentry *debugfs_dir;
140 	struct dentry *debugfs_monmap;
141 	struct dentry *debugfs_osdmap;
142 	struct dentry *debugfs_options;
143 #endif
144 };
145 
146 #define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
147 
148 
149 /*
150  * snapshots
151  */
152 
153 /*
154  * A "snap context" is the set of existing snapshots when we
155  * write data.  It is used by the OSD to guide its COW behavior.
156  *
157  * The ceph_snap_context is refcounted, and attached to each dirty
158  * page, indicating which context the dirty data belonged when it was
159  * dirtied.
160  */
161 struct ceph_snap_context {
162 	refcount_t nref;
163 	u64 seq;
164 	u32 num_snaps;
165 	u64 snaps[];
166 };
167 
168 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
169 					gfp_t gfp_flags);
170 extern struct ceph_snap_context *ceph_get_snap_context(
171 					struct ceph_snap_context *sc);
172 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
173 
174 /*
175  * calculate the number of pages a given length and offset map onto,
176  * if we align the data.
177  */
178 static inline int calc_pages_for(u64 off, u64 len)
179 {
180 	return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
181 		(off >> PAGE_SHIFT);
182 }
183 
184 #define RB_BYVAL(a)      (a)
185 #define RB_BYPTR(a)      (&(a))
186 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
187 
188 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
189 static void insert_##name(struct rb_root *root, type *t)		\
190 {									\
191 	struct rb_node **n = &root->rb_node;				\
192 	struct rb_node *parent = NULL;					\
193 									\
194 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
195 									\
196 	while (*n) {							\
197 		type *cur = rb_entry(*n, type, nodefld);		\
198 		int cmp;						\
199 									\
200 		parent = *n;						\
201 		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
202 		if (cmp < 0)						\
203 			n = &(*n)->rb_left;				\
204 		else if (cmp > 0)					\
205 			n = &(*n)->rb_right;				\
206 		else							\
207 			BUG();						\
208 	}								\
209 									\
210 	rb_link_node(&t->nodefld, parent, n);				\
211 	rb_insert_color(&t->nodefld, root);				\
212 }									\
213 static void erase_##name(struct rb_root *root, type *t)			\
214 {									\
215 	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
216 	rb_erase(&t->nodefld, root);					\
217 	RB_CLEAR_NODE(&t->nodefld);					\
218 }
219 
220 /*
221  * @lookup_param_type is a parameter and not constructed from (@type,
222  * @keyfld) with typeof() because adding const is too unwieldy.
223  */
224 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
225 			       lookup_param_type, nodefld)		\
226 static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
227 {									\
228 	struct rb_node *n = root->rb_node;				\
229 									\
230 	while (n) {							\
231 		type *cur = rb_entry(n, type, nodefld);			\
232 		int cmp;						\
233 									\
234 		cmp = cmpexp(key, keyexp(cur->keyfld));			\
235 		if (cmp < 0)						\
236 			n = n->rb_left;					\
237 		else if (cmp > 0)					\
238 			n = n->rb_right;				\
239 		else							\
240 			return cur;					\
241 	}								\
242 									\
243 	return NULL;							\
244 }
245 
246 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
247 			 lookup_param_type, nodefld)			\
248 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
249 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
250 		       lookup_param_type, nodefld)
251 
252 /*
253  * Shorthands for integer keys.
254  */
255 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
256 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
257 
258 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
259 extern type __lookup_##name##_key;					\
260 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
261 		       typeof(__lookup_##name##_key.keyfld), nodefld)
262 
263 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
264 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
265 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
266 
267 extern struct kmem_cache *ceph_inode_cachep;
268 extern struct kmem_cache *ceph_cap_cachep;
269 extern struct kmem_cache *ceph_cap_flush_cachep;
270 extern struct kmem_cache *ceph_dentry_cachep;
271 extern struct kmem_cache *ceph_file_cachep;
272 extern struct kmem_cache *ceph_dir_file_cachep;
273 
274 /* ceph_common.c */
275 extern bool libceph_compatible(void *data);
276 
277 extern const char *ceph_msg_type_name(int type);
278 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
279 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
280 
281 extern struct ceph_options *ceph_parse_options(char *options,
282 			      const char *dev_name, const char *dev_name_end,
283 			      int (*parse_extra_token)(char *c, void *private),
284 			      void *private);
285 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
286 			      bool show_all);
287 extern void ceph_destroy_options(struct ceph_options *opt);
288 extern int ceph_compare_options(struct ceph_options *new_opt,
289 				struct ceph_client *client);
290 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
291 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
292 u64 ceph_client_gid(struct ceph_client *client);
293 extern void ceph_destroy_client(struct ceph_client *client);
294 extern int __ceph_open_session(struct ceph_client *client,
295 			       unsigned long started);
296 extern int ceph_open_session(struct ceph_client *client);
297 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
298 				unsigned long timeout);
299 
300 /* pagevec.c */
301 extern void ceph_release_page_vector(struct page **pages, int num_pages);
302 
303 extern struct page **ceph_get_direct_page_vector(const void __user *data,
304 						 int num_pages,
305 						 bool write_page);
306 extern void ceph_put_page_vector(struct page **pages, int num_pages,
307 				 bool dirty);
308 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
309 extern int ceph_copy_user_to_page_vector(struct page **pages,
310 					 const void __user *data,
311 					 loff_t off, size_t len);
312 extern void ceph_copy_to_page_vector(struct page **pages,
313 				    const void *data,
314 				    loff_t off, size_t len);
315 extern void ceph_copy_from_page_vector(struct page **pages,
316 				    void *data,
317 				    loff_t off, size_t len);
318 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
319 
320 
321 #endif /* _FS_CEPH_SUPER_H */
322