xref: /openbmc/linux/include/linux/ceph/libceph.h (revision ebd09753)
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 
39 #define CEPH_OPT_DEFAULT   (CEPH_OPT_TCP_NODELAY)
40 
41 #define ceph_set_opt(client, opt) \
42 	(client)->options->flags |= CEPH_OPT_##opt;
43 #define ceph_test_opt(client, opt) \
44 	(!!((client)->options->flags & CEPH_OPT_##opt))
45 
46 struct ceph_options {
47 	int flags;
48 	struct ceph_fsid fsid;
49 	struct ceph_entity_addr my_addr;
50 	unsigned long mount_timeout;		/* jiffies */
51 	unsigned long osd_idle_ttl;		/* jiffies */
52 	unsigned long osd_keepalive_timeout;	/* jiffies */
53 	unsigned long osd_request_timeout;	/* jiffies */
54 
55 	/*
56 	 * any type that can't be simply compared or doesn't need need
57 	 * to be compared should go beyond this point,
58 	 * ceph_compare_options() should be updated accordingly
59 	 */
60 
61 	struct ceph_entity_addr *mon_addr; /* should be the first
62 					      pointer type of args */
63 	int num_mon;
64 	char *name;
65 	struct ceph_crypto_key *key;
66 };
67 
68 /*
69  * defaults
70  */
71 #define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
72 #define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
73 #define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
74 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
75 
76 #define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
77 #define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
78 #define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
79 #define CEPH_MONC_HUNT_BACKOFF		2
80 #define CEPH_MONC_HUNT_MAX_MULT		10
81 
82 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
83 #define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
84 
85 /*
86  * Handle the largest possible rbd object in one message.
87  * There is no limit on the size of cephfs objects, but it has to obey
88  * rsize and wsize mount options anyway.
89  */
90 #define CEPH_MSG_MAX_DATA_LEN	(32*1024*1024)
91 
92 #define CEPH_AUTH_NAME_DEFAULT   "guest"
93 
94 /* mount state */
95 enum {
96 	CEPH_MOUNT_MOUNTING,
97 	CEPH_MOUNT_MOUNTED,
98 	CEPH_MOUNT_UNMOUNTING,
99 	CEPH_MOUNT_UNMOUNTED,
100 	CEPH_MOUNT_SHUTDOWN,
101 };
102 
103 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
104 {
105 	return timeout ?: MAX_SCHEDULE_TIMEOUT;
106 }
107 
108 struct ceph_mds_client;
109 
110 /*
111  * per client state
112  *
113  * possibly shared by multiple mount points, if they are
114  * mounting the same ceph filesystem/cluster.
115  */
116 struct ceph_client {
117 	struct ceph_fsid fsid;
118 	bool have_fsid;
119 
120 	void *private;
121 
122 	struct ceph_options *options;
123 
124 	struct mutex mount_mutex;      /* serialize mount attempts */
125 	wait_queue_head_t auth_wq;
126 	int auth_err;
127 
128 	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
129 
130 	u64 supported_features;
131 	u64 required_features;
132 
133 	struct ceph_messenger msgr;   /* messenger instance */
134 	struct ceph_mon_client monc;
135 	struct ceph_osd_client osdc;
136 
137 #ifdef CONFIG_DEBUG_FS
138 	struct dentry *debugfs_dir;
139 	struct dentry *debugfs_monmap;
140 	struct dentry *debugfs_osdmap;
141 	struct dentry *debugfs_options;
142 #endif
143 };
144 
145 #define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
146 
147 
148 /*
149  * snapshots
150  */
151 
152 /*
153  * A "snap context" is the set of existing snapshots when we
154  * write data.  It is used by the OSD to guide its COW behavior.
155  *
156  * The ceph_snap_context is refcounted, and attached to each dirty
157  * page, indicating which context the dirty data belonged when it was
158  * dirtied.
159  */
160 struct ceph_snap_context {
161 	refcount_t nref;
162 	u64 seq;
163 	u32 num_snaps;
164 	u64 snaps[];
165 };
166 
167 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
168 					gfp_t gfp_flags);
169 extern struct ceph_snap_context *ceph_get_snap_context(
170 					struct ceph_snap_context *sc);
171 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
172 
173 /*
174  * calculate the number of pages a given length and offset map onto,
175  * if we align the data.
176  */
177 static inline int calc_pages_for(u64 off, u64 len)
178 {
179 	return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
180 		(off >> PAGE_SHIFT);
181 }
182 
183 #define RB_BYVAL(a)      (a)
184 #define RB_BYPTR(a)      (&(a))
185 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
186 
187 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
188 static void insert_##name(struct rb_root *root, type *t)		\
189 {									\
190 	struct rb_node **n = &root->rb_node;				\
191 	struct rb_node *parent = NULL;					\
192 									\
193 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
194 									\
195 	while (*n) {							\
196 		type *cur = rb_entry(*n, type, nodefld);		\
197 		int cmp;						\
198 									\
199 		parent = *n;						\
200 		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
201 		if (cmp < 0)						\
202 			n = &(*n)->rb_left;				\
203 		else if (cmp > 0)					\
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 /*
220  * @lookup_param_type is a parameter and not constructed from (@type,
221  * @keyfld) with typeof() because adding const is too unwieldy.
222  */
223 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
224 			       lookup_param_type, nodefld)		\
225 static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
226 {									\
227 	struct rb_node *n = root->rb_node;				\
228 									\
229 	while (n) {							\
230 		type *cur = rb_entry(n, type, nodefld);			\
231 		int cmp;						\
232 									\
233 		cmp = cmpexp(key, keyexp(cur->keyfld));			\
234 		if (cmp < 0)						\
235 			n = n->rb_left;					\
236 		else if (cmp > 0)					\
237 			n = n->rb_right;				\
238 		else							\
239 			return cur;					\
240 	}								\
241 									\
242 	return NULL;							\
243 }
244 
245 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
246 			 lookup_param_type, nodefld)			\
247 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
248 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
249 		       lookup_param_type, nodefld)
250 
251 /*
252  * Shorthands for integer keys.
253  */
254 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
255 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
256 
257 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
258 extern type __lookup_##name##_key;					\
259 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
260 		       typeof(__lookup_##name##_key.keyfld), nodefld)
261 
262 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
263 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
264 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
265 
266 extern struct kmem_cache *ceph_inode_cachep;
267 extern struct kmem_cache *ceph_cap_cachep;
268 extern struct kmem_cache *ceph_cap_flush_cachep;
269 extern struct kmem_cache *ceph_dentry_cachep;
270 extern struct kmem_cache *ceph_file_cachep;
271 extern struct kmem_cache *ceph_dir_file_cachep;
272 
273 /* ceph_common.c */
274 extern bool libceph_compatible(void *data);
275 
276 extern const char *ceph_msg_type_name(int type);
277 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
278 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
279 
280 extern struct ceph_options *ceph_parse_options(char *options,
281 			      const char *dev_name, const char *dev_name_end,
282 			      int (*parse_extra_token)(char *c, void *private),
283 			      void *private);
284 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client);
285 extern void ceph_destroy_options(struct ceph_options *opt);
286 extern int ceph_compare_options(struct ceph_options *new_opt,
287 				struct ceph_client *client);
288 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
289 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
290 u64 ceph_client_gid(struct ceph_client *client);
291 extern void ceph_destroy_client(struct ceph_client *client);
292 extern int __ceph_open_session(struct ceph_client *client,
293 			       unsigned long started);
294 extern int ceph_open_session(struct ceph_client *client);
295 
296 /* pagevec.c */
297 extern void ceph_release_page_vector(struct page **pages, int num_pages);
298 
299 extern struct page **ceph_get_direct_page_vector(const void __user *data,
300 						 int num_pages,
301 						 bool write_page);
302 extern void ceph_put_page_vector(struct page **pages, int num_pages,
303 				 bool dirty);
304 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
305 extern int ceph_copy_user_to_page_vector(struct page **pages,
306 					 const void __user *data,
307 					 loff_t off, size_t len);
308 extern void ceph_copy_to_page_vector(struct page **pages,
309 				    const void *data,
310 				    loff_t off, size_t len);
311 extern void ceph_copy_from_page_vector(struct page **pages,
312 				    void *data,
313 				    loff_t off, size_t len);
314 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
315 
316 
317 #endif /* _FS_CEPH_SUPER_H */
318