xref: /openbmc/linux/include/linux/ceph/libceph.h (revision 8ee90c5c)
1 #ifndef _FS_CEPH_LIBCEPH_H
2 #define _FS_CEPH_LIBCEPH_H
3 
4 #include <linux/ceph/ceph_debug.h>
5 
6 #include <asm/unaligned.h>
7 #include <linux/backing-dev.h>
8 #include <linux/completion.h>
9 #include <linux/exportfs.h>
10 #include <linux/bug.h>
11 #include <linux/fs.h>
12 #include <linux/mempool.h>
13 #include <linux/pagemap.h>
14 #include <linux/wait.h>
15 #include <linux/writeback.h>
16 #include <linux/slab.h>
17 #include <linux/refcount.h>
18 
19 #include <linux/ceph/types.h>
20 #include <linux/ceph/messenger.h>
21 #include <linux/ceph/msgpool.h>
22 #include <linux/ceph/mon_client.h>
23 #include <linux/ceph/osd_client.h>
24 #include <linux/ceph/ceph_fs.h>
25 #include <linux/ceph/string_table.h>
26 
27 /*
28  * mount options
29  */
30 #define CEPH_OPT_FSID             (1<<0)
31 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
32 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
33 #define CEPH_OPT_NOCRC            (1<<3) /* no data crc on writes */
34 #define CEPH_OPT_NOMSGAUTH	  (1<<4) /* don't require msg signing feat */
35 #define CEPH_OPT_TCP_NODELAY	  (1<<5) /* TCP_NODELAY on TCP sockets */
36 #define CEPH_OPT_NOMSGSIGN	  (1<<6) /* don't sign msgs */
37 
38 #define CEPH_OPT_DEFAULT   (CEPH_OPT_TCP_NODELAY)
39 
40 #define ceph_set_opt(client, opt) \
41 	(client)->options->flags |= CEPH_OPT_##opt;
42 #define ceph_test_opt(client, opt) \
43 	(!!((client)->options->flags & CEPH_OPT_##opt))
44 
45 struct ceph_options {
46 	int flags;
47 	struct ceph_fsid fsid;
48 	struct ceph_entity_addr my_addr;
49 	unsigned long mount_timeout;		/* jiffies */
50 	unsigned long osd_idle_ttl;		/* jiffies */
51 	unsigned long osd_keepalive_timeout;	/* jiffies */
52 	unsigned long osd_request_timeout;	/* jiffies */
53 
54 	/*
55 	 * any type that can't be simply compared or doesn't need need
56 	 * to be compared should go beyond this point,
57 	 * ceph_compare_options() should be updated accordingly
58 	 */
59 
60 	struct ceph_entity_addr *mon_addr; /* should be the first
61 					      pointer type of args */
62 	int num_mon;
63 	char *name;
64 	struct ceph_crypto_key *key;
65 };
66 
67 /*
68  * defaults
69  */
70 #define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
71 #define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
72 #define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
73 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
74 
75 #define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
76 #define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
77 #define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
78 #define CEPH_MONC_HUNT_BACKOFF		2
79 #define CEPH_MONC_HUNT_MAX_MULT		10
80 
81 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
82 #define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
83 #define CEPH_MSG_MAX_DATA_LEN	(16*1024*1024)
84 
85 #define CEPH_AUTH_NAME_DEFAULT   "guest"
86 
87 /* mount state */
88 enum {
89 	CEPH_MOUNT_MOUNTING,
90 	CEPH_MOUNT_MOUNTED,
91 	CEPH_MOUNT_UNMOUNTING,
92 	CEPH_MOUNT_UNMOUNTED,
93 	CEPH_MOUNT_SHUTDOWN,
94 };
95 
96 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
97 {
98 	return timeout ?: MAX_SCHEDULE_TIMEOUT;
99 }
100 
101 struct ceph_mds_client;
102 
103 /*
104  * per client state
105  *
106  * possibly shared by multiple mount points, if they are
107  * mounting the same ceph filesystem/cluster.
108  */
109 struct ceph_client {
110 	struct ceph_fsid fsid;
111 	bool have_fsid;
112 
113 	void *private;
114 
115 	struct ceph_options *options;
116 
117 	struct mutex mount_mutex;      /* serialize mount attempts */
118 	wait_queue_head_t auth_wq;
119 	int auth_err;
120 
121 	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
122 
123 	u64 supported_features;
124 	u64 required_features;
125 
126 	struct ceph_messenger msgr;   /* messenger instance */
127 	struct ceph_mon_client monc;
128 	struct ceph_osd_client osdc;
129 
130 #ifdef CONFIG_DEBUG_FS
131 	struct dentry *debugfs_dir;
132 	struct dentry *debugfs_monmap;
133 	struct dentry *debugfs_osdmap;
134 	struct dentry *debugfs_options;
135 #endif
136 };
137 
138 #define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
139 
140 
141 /*
142  * snapshots
143  */
144 
145 /*
146  * A "snap context" is the set of existing snapshots when we
147  * write data.  It is used by the OSD to guide its COW behavior.
148  *
149  * The ceph_snap_context is refcounted, and attached to each dirty
150  * page, indicating which context the dirty data belonged when it was
151  * dirtied.
152  */
153 struct ceph_snap_context {
154 	refcount_t nref;
155 	u64 seq;
156 	u32 num_snaps;
157 	u64 snaps[];
158 };
159 
160 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
161 					gfp_t gfp_flags);
162 extern struct ceph_snap_context *ceph_get_snap_context(
163 					struct ceph_snap_context *sc);
164 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
165 
166 /*
167  * calculate the number of pages a given length and offset map onto,
168  * if we align the data.
169  */
170 static inline int calc_pages_for(u64 off, u64 len)
171 {
172 	return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
173 		(off >> PAGE_SHIFT);
174 }
175 
176 #define RB_BYVAL(a)      (a)
177 #define RB_BYPTR(a)      (&(a))
178 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
179 
180 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
181 static void insert_##name(struct rb_root *root, type *t)		\
182 {									\
183 	struct rb_node **n = &root->rb_node;				\
184 	struct rb_node *parent = NULL;					\
185 									\
186 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
187 									\
188 	while (*n) {							\
189 		type *cur = rb_entry(*n, type, nodefld);		\
190 		int cmp;						\
191 									\
192 		parent = *n;						\
193 		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
194 		if (cmp < 0)						\
195 			n = &(*n)->rb_left;				\
196 		else if (cmp > 0)					\
197 			n = &(*n)->rb_right;				\
198 		else							\
199 			BUG();						\
200 	}								\
201 									\
202 	rb_link_node(&t->nodefld, parent, n);				\
203 	rb_insert_color(&t->nodefld, root);				\
204 }									\
205 static void erase_##name(struct rb_root *root, type *t)			\
206 {									\
207 	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
208 	rb_erase(&t->nodefld, root);					\
209 	RB_CLEAR_NODE(&t->nodefld);					\
210 }
211 
212 /*
213  * @lookup_param_type is a parameter and not constructed from (@type,
214  * @keyfld) with typeof() because adding const is too unwieldy.
215  */
216 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
217 			       lookup_param_type, nodefld)		\
218 static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
219 {									\
220 	struct rb_node *n = root->rb_node;				\
221 									\
222 	while (n) {							\
223 		type *cur = rb_entry(n, type, nodefld);			\
224 		int cmp;						\
225 									\
226 		cmp = cmpexp(key, keyexp(cur->keyfld));			\
227 		if (cmp < 0)						\
228 			n = n->rb_left;					\
229 		else if (cmp > 0)					\
230 			n = n->rb_right;				\
231 		else							\
232 			return cur;					\
233 	}								\
234 									\
235 	return NULL;							\
236 }
237 
238 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
239 			 lookup_param_type, nodefld)			\
240 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
241 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
242 		       lookup_param_type, nodefld)
243 
244 /*
245  * Shorthands for integer keys.
246  */
247 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
248 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
249 
250 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
251 extern type __lookup_##name##_key;					\
252 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
253 		       typeof(__lookup_##name##_key.keyfld), nodefld)
254 
255 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
256 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
257 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
258 
259 extern struct kmem_cache *ceph_inode_cachep;
260 extern struct kmem_cache *ceph_cap_cachep;
261 extern struct kmem_cache *ceph_cap_flush_cachep;
262 extern struct kmem_cache *ceph_dentry_cachep;
263 extern struct kmem_cache *ceph_file_cachep;
264 
265 /* ceph_common.c */
266 extern bool libceph_compatible(void *data);
267 
268 extern const char *ceph_msg_type_name(int type);
269 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
270 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
271 
272 extern struct ceph_options *ceph_parse_options(char *options,
273 			      const char *dev_name, const char *dev_name_end,
274 			      int (*parse_extra_token)(char *c, void *private),
275 			      void *private);
276 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client);
277 extern void ceph_destroy_options(struct ceph_options *opt);
278 extern int ceph_compare_options(struct ceph_options *new_opt,
279 				struct ceph_client *client);
280 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
281 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
282 u64 ceph_client_gid(struct ceph_client *client);
283 extern void ceph_destroy_client(struct ceph_client *client);
284 extern int __ceph_open_session(struct ceph_client *client,
285 			       unsigned long started);
286 extern int ceph_open_session(struct ceph_client *client);
287 
288 /* pagevec.c */
289 extern void ceph_release_page_vector(struct page **pages, int num_pages);
290 
291 extern struct page **ceph_get_direct_page_vector(const void __user *data,
292 						 int num_pages,
293 						 bool write_page);
294 extern void ceph_put_page_vector(struct page **pages, int num_pages,
295 				 bool dirty);
296 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
297 extern int ceph_copy_user_to_page_vector(struct page **pages,
298 					 const void __user *data,
299 					 loff_t off, size_t len);
300 extern void ceph_copy_to_page_vector(struct page **pages,
301 				    const void *data,
302 				    loff_t off, size_t len);
303 extern void ceph_copy_from_page_vector(struct page **pages,
304 				    void *data,
305 				    loff_t off, size_t len);
306 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
307 
308 
309 #endif /* _FS_CEPH_SUPER_H */
310