xref: /openbmc/linux/include/linux/ceph/libceph.h (revision 4bb1e4e7)
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  * The largest possible rbd data object is 32M.
88  * The largest possible rbd object map object is 64M.
89  *
90  * There is no limit on the size of cephfs objects, but it has to obey
91  * rsize and wsize mount options anyway.
92  */
93 #define CEPH_MSG_MAX_DATA_LEN	(64*1024*1024)
94 
95 #define CEPH_AUTH_NAME_DEFAULT   "guest"
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 	refcount_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 #define RB_BYVAL(a)      (a)
187 #define RB_BYPTR(a)      (&(a))
188 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
189 
190 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
191 static void insert_##name(struct rb_root *root, type *t)		\
192 {									\
193 	struct rb_node **n = &root->rb_node;				\
194 	struct rb_node *parent = NULL;					\
195 									\
196 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
197 									\
198 	while (*n) {							\
199 		type *cur = rb_entry(*n, type, nodefld);		\
200 		int cmp;						\
201 									\
202 		parent = *n;						\
203 		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
204 		if (cmp < 0)						\
205 			n = &(*n)->rb_left;				\
206 		else if (cmp > 0)					\
207 			n = &(*n)->rb_right;				\
208 		else							\
209 			BUG();						\
210 	}								\
211 									\
212 	rb_link_node(&t->nodefld, parent, n);				\
213 	rb_insert_color(&t->nodefld, root);				\
214 }									\
215 static void erase_##name(struct rb_root *root, type *t)			\
216 {									\
217 	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
218 	rb_erase(&t->nodefld, root);					\
219 	RB_CLEAR_NODE(&t->nodefld);					\
220 }
221 
222 /*
223  * @lookup_param_type is a parameter and not constructed from (@type,
224  * @keyfld) with typeof() because adding const is too unwieldy.
225  */
226 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
227 			       lookup_param_type, nodefld)		\
228 static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
229 {									\
230 	struct rb_node *n = root->rb_node;				\
231 									\
232 	while (n) {							\
233 		type *cur = rb_entry(n, type, nodefld);			\
234 		int cmp;						\
235 									\
236 		cmp = cmpexp(key, keyexp(cur->keyfld));			\
237 		if (cmp < 0)						\
238 			n = n->rb_left;					\
239 		else if (cmp > 0)					\
240 			n = n->rb_right;				\
241 		else							\
242 			return cur;					\
243 	}								\
244 									\
245 	return NULL;							\
246 }
247 
248 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
249 			 lookup_param_type, nodefld)			\
250 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
251 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
252 		       lookup_param_type, nodefld)
253 
254 /*
255  * Shorthands for integer keys.
256  */
257 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
258 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
259 
260 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
261 extern type __lookup_##name##_key;					\
262 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
263 		       typeof(__lookup_##name##_key.keyfld), nodefld)
264 
265 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
266 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
267 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
268 
269 extern struct kmem_cache *ceph_inode_cachep;
270 extern struct kmem_cache *ceph_cap_cachep;
271 extern struct kmem_cache *ceph_cap_flush_cachep;
272 extern struct kmem_cache *ceph_dentry_cachep;
273 extern struct kmem_cache *ceph_file_cachep;
274 extern struct kmem_cache *ceph_dir_file_cachep;
275 
276 /* ceph_common.c */
277 extern bool libceph_compatible(void *data);
278 
279 extern const char *ceph_msg_type_name(int type);
280 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
281 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
282 
283 struct fs_parameter;
284 struct ceph_options *ceph_alloc_options(void);
285 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
286 		       struct fs_context *fc);
287 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
288 		     struct fs_context *fc);
289 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
290 			      bool show_all);
291 extern void ceph_destroy_options(struct ceph_options *opt);
292 extern int ceph_compare_options(struct ceph_options *new_opt,
293 				struct ceph_client *client);
294 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
295 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
296 u64 ceph_client_gid(struct ceph_client *client);
297 extern void ceph_destroy_client(struct ceph_client *client);
298 extern void ceph_reset_client_addr(struct ceph_client *client);
299 extern int __ceph_open_session(struct ceph_client *client,
300 			       unsigned long started);
301 extern int ceph_open_session(struct ceph_client *client);
302 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
303 				unsigned long timeout);
304 
305 /* pagevec.c */
306 extern void ceph_release_page_vector(struct page **pages, int num_pages);
307 extern void ceph_put_page_vector(struct page **pages, int num_pages,
308 				 bool dirty);
309 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
310 extern int ceph_copy_user_to_page_vector(struct page **pages,
311 					 const void __user *data,
312 					 loff_t off, size_t len);
313 extern void ceph_copy_to_page_vector(struct page **pages,
314 				    const void *data,
315 				    loff_t off, size_t len);
316 extern void ceph_copy_from_page_vector(struct page **pages,
317 				    void *data,
318 				    loff_t off, size_t len);
319 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
320 
321 
322 #endif /* _FS_CEPH_SUPER_H */
323