xref: /openbmc/linux/include/linux/ceph/libceph.h (revision 1c2dd16a)
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 
18 #include <linux/ceph/types.h>
19 #include <linux/ceph/messenger.h>
20 #include <linux/ceph/msgpool.h>
21 #include <linux/ceph/mon_client.h>
22 #include <linux/ceph/osd_client.h>
23 #include <linux/ceph/ceph_fs.h>
24 #include <linux/ceph/string_table.h>
25 
26 /*
27  * mount options
28  */
29 #define CEPH_OPT_FSID             (1<<0)
30 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
31 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
32 #define CEPH_OPT_NOCRC            (1<<3) /* no data crc on writes */
33 #define CEPH_OPT_NOMSGAUTH	  (1<<4) /* don't require msg signing feat */
34 #define CEPH_OPT_TCP_NODELAY	  (1<<5) /* TCP_NODELAY on TCP sockets */
35 #define CEPH_OPT_NOMSGSIGN	  (1<<6) /* don't sign msgs */
36 
37 #define CEPH_OPT_DEFAULT   (CEPH_OPT_TCP_NODELAY)
38 
39 #define ceph_set_opt(client, opt) \
40 	(client)->options->flags |= CEPH_OPT_##opt;
41 #define ceph_test_opt(client, opt) \
42 	(!!((client)->options->flags & CEPH_OPT_##opt))
43 
44 struct ceph_options {
45 	int flags;
46 	struct ceph_fsid fsid;
47 	struct ceph_entity_addr my_addr;
48 	unsigned long mount_timeout;		/* jiffies */
49 	unsigned long osd_idle_ttl;		/* jiffies */
50 	unsigned long osd_keepalive_timeout;	/* jiffies */
51 	unsigned long osd_request_timeout;	/* jiffies */
52 
53 	/*
54 	 * any type that can't be simply compared or doesn't need need
55 	 * to be compared should go beyond this point,
56 	 * ceph_compare_options() should be updated accordingly
57 	 */
58 
59 	struct ceph_entity_addr *mon_addr; /* should be the first
60 					      pointer type of args */
61 	int num_mon;
62 	char *name;
63 	struct ceph_crypto_key *key;
64 };
65 
66 /*
67  * defaults
68  */
69 #define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
70 #define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
71 #define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
72 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
73 
74 #define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
75 #define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
76 #define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
77 #define CEPH_MONC_HUNT_BACKOFF		2
78 #define CEPH_MONC_HUNT_MAX_MULT		10
79 
80 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
81 #define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
82 #define CEPH_MSG_MAX_DATA_LEN	(16*1024*1024)
83 
84 #define CEPH_AUTH_NAME_DEFAULT   "guest"
85 
86 /*
87  * Delay telling the MDS we no longer want caps, in case we reopen
88  * the file.  Delay a minimum amount of time, even if we send a cap
89  * message for some other reason.  Otherwise, take the oppotunity to
90  * update the mds to avoid sending another message later.
91  */
92 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
93 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
94 
95 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT        (CEPH_CAPS_PER_RELEASE * 4)
96 
97 /* mount state */
98 enum {
99 	CEPH_MOUNT_MOUNTING,
100 	CEPH_MOUNT_MOUNTED,
101 	CEPH_MOUNT_UNMOUNTING,
102 	CEPH_MOUNT_UNMOUNTED,
103 	CEPH_MOUNT_SHUTDOWN,
104 };
105 
106 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
107 {
108 	return timeout ?: MAX_SCHEDULE_TIMEOUT;
109 }
110 
111 struct ceph_mds_client;
112 
113 /*
114  * per client state
115  *
116  * possibly shared by multiple mount points, if they are
117  * mounting the same ceph filesystem/cluster.
118  */
119 struct ceph_client {
120 	struct ceph_fsid fsid;
121 	bool have_fsid;
122 
123 	void *private;
124 
125 	struct ceph_options *options;
126 
127 	struct mutex mount_mutex;      /* serialize mount attempts */
128 	wait_queue_head_t auth_wq;
129 	int auth_err;
130 
131 	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
132 
133 	u64 supported_features;
134 	u64 required_features;
135 
136 	struct ceph_messenger msgr;   /* messenger instance */
137 	struct ceph_mon_client monc;
138 	struct ceph_osd_client osdc;
139 
140 #ifdef CONFIG_DEBUG_FS
141 	struct dentry *debugfs_dir;
142 	struct dentry *debugfs_monmap;
143 	struct dentry *debugfs_osdmap;
144 	struct dentry *debugfs_options;
145 #endif
146 };
147 
148 #define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
149 
150 
151 /*
152  * snapshots
153  */
154 
155 /*
156  * A "snap context" is the set of existing snapshots when we
157  * write data.  It is used by the OSD to guide its COW behavior.
158  *
159  * The ceph_snap_context is refcounted, and attached to each dirty
160  * page, indicating which context the dirty data belonged when it was
161  * dirtied.
162  */
163 struct ceph_snap_context {
164 	atomic_t nref;
165 	u64 seq;
166 	u32 num_snaps;
167 	u64 snaps[];
168 };
169 
170 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
171 					gfp_t gfp_flags);
172 extern struct ceph_snap_context *ceph_get_snap_context(
173 					struct ceph_snap_context *sc);
174 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
175 
176 /*
177  * calculate the number of pages a given length and offset map onto,
178  * if we align the data.
179  */
180 static inline int calc_pages_for(u64 off, u64 len)
181 {
182 	return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
183 		(off >> PAGE_SHIFT);
184 }
185 
186 /*
187  * These are not meant to be generic - an integer key is assumed.
188  */
189 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
190 static void insert_##name(struct rb_root *root, type *t)		\
191 {									\
192 	struct rb_node **n = &root->rb_node;				\
193 	struct rb_node *parent = NULL;					\
194 									\
195 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
196 									\
197 	while (*n) {							\
198 		type *cur = rb_entry(*n, type, nodefld);		\
199 									\
200 		parent = *n;						\
201 		if (t->keyfld < cur->keyfld)				\
202 			n = &(*n)->rb_left;				\
203 		else if (t->keyfld > cur->keyfld)			\
204 			n = &(*n)->rb_right;				\
205 		else							\
206 			BUG();						\
207 	}								\
208 									\
209 	rb_link_node(&t->nodefld, parent, n);				\
210 	rb_insert_color(&t->nodefld, root);				\
211 }									\
212 static void erase_##name(struct rb_root *root, type *t)			\
213 {									\
214 	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
215 	rb_erase(&t->nodefld, root);					\
216 	RB_CLEAR_NODE(&t->nodefld);					\
217 }
218 
219 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
220 extern type __lookup_##name##_key;					\
221 static type *lookup_##name(struct rb_root *root,			\
222 			   typeof(__lookup_##name##_key.keyfld) key)	\
223 {									\
224 	struct rb_node *n = root->rb_node;				\
225 									\
226 	while (n) {							\
227 		type *cur = rb_entry(n, type, nodefld);			\
228 									\
229 		if (key < cur->keyfld)					\
230 			n = n->rb_left;					\
231 		else if (key > cur->keyfld)				\
232 			n = n->rb_right;				\
233 		else							\
234 			return cur;					\
235 	}								\
236 									\
237 	return NULL;							\
238 }
239 
240 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
241 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
242 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
243 
244 extern struct kmem_cache *ceph_inode_cachep;
245 extern struct kmem_cache *ceph_cap_cachep;
246 extern struct kmem_cache *ceph_cap_flush_cachep;
247 extern struct kmem_cache *ceph_dentry_cachep;
248 extern struct kmem_cache *ceph_file_cachep;
249 
250 /* ceph_common.c */
251 extern bool libceph_compatible(void *data);
252 
253 extern const char *ceph_msg_type_name(int type);
254 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
255 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
256 
257 extern struct ceph_options *ceph_parse_options(char *options,
258 			      const char *dev_name, const char *dev_name_end,
259 			      int (*parse_extra_token)(char *c, void *private),
260 			      void *private);
261 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client);
262 extern void ceph_destroy_options(struct ceph_options *opt);
263 extern int ceph_compare_options(struct ceph_options *new_opt,
264 				struct ceph_client *client);
265 extern struct ceph_client *ceph_create_client(struct ceph_options *opt,
266 					      void *private,
267 					      u64 supported_features,
268 					      u64 required_features);
269 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
270 u64 ceph_client_gid(struct ceph_client *client);
271 extern void ceph_destroy_client(struct ceph_client *client);
272 extern int __ceph_open_session(struct ceph_client *client,
273 			       unsigned long started);
274 extern int ceph_open_session(struct ceph_client *client);
275 
276 /* pagevec.c */
277 extern void ceph_release_page_vector(struct page **pages, int num_pages);
278 
279 extern struct page **ceph_get_direct_page_vector(const void __user *data,
280 						 int num_pages,
281 						 bool write_page);
282 extern void ceph_put_page_vector(struct page **pages, int num_pages,
283 				 bool dirty);
284 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
285 extern int ceph_copy_user_to_page_vector(struct page **pages,
286 					 const void __user *data,
287 					 loff_t off, size_t len);
288 extern void ceph_copy_to_page_vector(struct page **pages,
289 				    const void *data,
290 				    loff_t off, size_t len);
291 extern void ceph_copy_from_page_vector(struct page **pages,
292 				    void *data,
293 				    loff_t off, size_t len);
294 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
295 
296 
297 #endif /* _FS_CEPH_SUPER_H */
298