xref: /openbmc/linux/include/linux/ceph/libceph.h (revision 08f3e087)
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 (msgr1) */
35 #define CEPH_OPT_TCP_NODELAY      (1<<4) /* TCP_NODELAY on TCP sockets */
36 #define CEPH_OPT_NOMSGSIGN        (1<<5) /* don't sign msgs (msgr1) */
37 #define CEPH_OPT_ABORT_ON_FULL    (1<<6) /* abort w/ ENOSPC when full */
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 	u32 read_from_replica;  /* CEPH_OSD_FLAG_BALANCE/LOCALIZE_READS */
55 	int con_modes[2];  /* CEPH_CON_MODE_* */
56 
57 	/*
58 	 * any type that can't be simply compared or doesn't need
59 	 * to be compared should go beyond this point,
60 	 * ceph_compare_options() should be updated accordingly
61 	 */
62 
63 	struct ceph_entity_addr *mon_addr; /* should be the first
64 					      pointer type of args */
65 	int num_mon;
66 	char *name;
67 	struct ceph_crypto_key *key;
68 	struct rb_root crush_locs;
69 };
70 
71 /*
72  * defaults
73  */
74 #define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
75 #define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
76 #define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
77 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
78 #define CEPH_READ_FROM_REPLICA_DEFAULT	0  /* read from primary */
79 
80 #define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
81 #define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
82 #define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
83 #define CEPH_MONC_HUNT_BACKOFF		2
84 #define CEPH_MONC_HUNT_MAX_MULT		10
85 
86 #define CEPH_MSG_MAX_CONTROL_LEN (16*1024*1024)
87 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
88 #define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
89 
90 /*
91  * The largest possible rbd data object is 32M.
92  * The largest possible rbd object map object is 64M.
93  *
94  * There is no limit on the size of cephfs objects, but it has to obey
95  * rsize and wsize mount options anyway.
96  */
97 #define CEPH_MSG_MAX_DATA_LEN	(64*1024*1024)
98 
99 #define CEPH_AUTH_NAME_DEFAULT   "guest"
100 
101 /* mount state */
102 enum {
103 	CEPH_MOUNT_MOUNTING,
104 	CEPH_MOUNT_MOUNTED,
105 	CEPH_MOUNT_UNMOUNTING,
106 	CEPH_MOUNT_UNMOUNTED,
107 	CEPH_MOUNT_SHUTDOWN,
108 	CEPH_MOUNT_RECOVER,
109 };
110 
111 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
112 {
113 	return timeout ?: MAX_SCHEDULE_TIMEOUT;
114 }
115 
116 struct ceph_mds_client;
117 
118 /*
119  * per client state
120  *
121  * possibly shared by multiple mount points, if they are
122  * mounting the same ceph filesystem/cluster.
123  */
124 struct ceph_client {
125 	struct ceph_fsid fsid;
126 	bool have_fsid;
127 
128 	void *private;
129 
130 	struct ceph_options *options;
131 
132 	struct mutex mount_mutex;      /* serialize mount attempts */
133 	wait_queue_head_t auth_wq;
134 	int auth_err;
135 
136 	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
137 
138 	u64 supported_features;
139 	u64 required_features;
140 
141 	struct ceph_messenger msgr;   /* messenger instance */
142 	struct ceph_mon_client monc;
143 	struct ceph_osd_client osdc;
144 
145 #ifdef CONFIG_DEBUG_FS
146 	struct dentry *debugfs_dir;
147 	struct dentry *debugfs_monmap;
148 	struct dentry *debugfs_osdmap;
149 	struct dentry *debugfs_options;
150 #endif
151 };
152 
153 #define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
154 
155 static inline bool ceph_msgr2(struct ceph_client *client)
156 {
157 	return client->options->con_modes[0] != CEPH_CON_MODE_UNKNOWN;
158 }
159 
160 /*
161  * snapshots
162  */
163 
164 /*
165  * A "snap context" is the set of existing snapshots when we
166  * write data.  It is used by the OSD to guide its COW behavior.
167  *
168  * The ceph_snap_context is refcounted, and attached to each dirty
169  * page, indicating which context the dirty data belonged when it was
170  * dirtied.
171  */
172 struct ceph_snap_context {
173 	refcount_t nref;
174 	u64 seq;
175 	u32 num_snaps;
176 	u64 snaps[];
177 };
178 
179 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
180 					gfp_t gfp_flags);
181 extern struct ceph_snap_context *ceph_get_snap_context(
182 					struct ceph_snap_context *sc);
183 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
184 
185 /*
186  * calculate the number of pages a given length and offset map onto,
187  * if we align the data.
188  */
189 static inline int calc_pages_for(u64 off, u64 len)
190 {
191 	return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
192 		(off >> PAGE_SHIFT);
193 }
194 
195 #define RB_BYVAL(a)      (a)
196 #define RB_BYPTR(a)      (&(a))
197 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
198 
199 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
200 static bool __insert_##name(struct rb_root *root, type *t)		\
201 {									\
202 	struct rb_node **n = &root->rb_node;				\
203 	struct rb_node *parent = NULL;					\
204 									\
205 	BUG_ON(!RB_EMPTY_NODE(&t->nodefld));				\
206 									\
207 	while (*n) {							\
208 		type *cur = rb_entry(*n, type, nodefld);		\
209 		int cmp;						\
210 									\
211 		parent = *n;						\
212 		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
213 		if (cmp < 0)						\
214 			n = &(*n)->rb_left;				\
215 		else if (cmp > 0)					\
216 			n = &(*n)->rb_right;				\
217 		else							\
218 			return false;					\
219 	}								\
220 									\
221 	rb_link_node(&t->nodefld, parent, n);				\
222 	rb_insert_color(&t->nodefld, root);				\
223 	return true;							\
224 }									\
225 static void __maybe_unused insert_##name(struct rb_root *root, type *t)	\
226 {									\
227 	if (!__insert_##name(root, t))					\
228 		BUG();							\
229 }									\
230 static void erase_##name(struct rb_root *root, type *t)			\
231 {									\
232 	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
233 	rb_erase(&t->nodefld, root);					\
234 	RB_CLEAR_NODE(&t->nodefld);					\
235 }
236 
237 /*
238  * @lookup_param_type is a parameter and not constructed from (@type,
239  * @keyfld) with typeof() because adding const is too unwieldy.
240  */
241 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
242 			       lookup_param_type, nodefld)		\
243 static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
244 {									\
245 	struct rb_node *n = root->rb_node;				\
246 									\
247 	while (n) {							\
248 		type *cur = rb_entry(n, type, nodefld);			\
249 		int cmp;						\
250 									\
251 		cmp = cmpexp(key, keyexp(cur->keyfld));			\
252 		if (cmp < 0)						\
253 			n = n->rb_left;					\
254 		else if (cmp > 0)					\
255 			n = n->rb_right;				\
256 		else							\
257 			return cur;					\
258 	}								\
259 									\
260 	return NULL;							\
261 }
262 
263 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
264 			 lookup_param_type, nodefld)			\
265 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
266 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
267 		       lookup_param_type, nodefld)
268 
269 /*
270  * Shorthands for integer keys.
271  */
272 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
273 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
274 
275 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
276 extern type __lookup_##name##_key;					\
277 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
278 		       typeof(__lookup_##name##_key.keyfld), nodefld)
279 
280 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
281 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
282 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
283 
284 extern struct kmem_cache *ceph_inode_cachep;
285 extern struct kmem_cache *ceph_cap_cachep;
286 extern struct kmem_cache *ceph_cap_flush_cachep;
287 extern struct kmem_cache *ceph_dentry_cachep;
288 extern struct kmem_cache *ceph_file_cachep;
289 extern struct kmem_cache *ceph_dir_file_cachep;
290 extern struct kmem_cache *ceph_mds_request_cachep;
291 extern mempool_t *ceph_wb_pagevec_pool;
292 
293 /* ceph_common.c */
294 extern bool libceph_compatible(void *data);
295 
296 extern const char *ceph_msg_type_name(int type);
297 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
298 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
299 
300 struct fs_parameter;
301 struct fc_log;
302 struct ceph_options *ceph_alloc_options(void);
303 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
304 		       struct fc_log *l);
305 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
306 		     struct fc_log *l);
307 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
308 			      bool show_all);
309 extern void ceph_destroy_options(struct ceph_options *opt);
310 extern int ceph_compare_options(struct ceph_options *new_opt,
311 				struct ceph_client *client);
312 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
313 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
314 u64 ceph_client_gid(struct ceph_client *client);
315 extern void ceph_destroy_client(struct ceph_client *client);
316 extern void ceph_reset_client_addr(struct ceph_client *client);
317 extern int __ceph_open_session(struct ceph_client *client,
318 			       unsigned long started);
319 extern int ceph_open_session(struct ceph_client *client);
320 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
321 				unsigned long timeout);
322 
323 /* pagevec.c */
324 extern void ceph_release_page_vector(struct page **pages, int num_pages);
325 extern void ceph_put_page_vector(struct page **pages, int num_pages,
326 				 bool dirty);
327 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
328 extern int ceph_copy_user_to_page_vector(struct page **pages,
329 					 const void __user *data,
330 					 loff_t off, size_t len);
331 extern void ceph_copy_to_page_vector(struct page **pages,
332 				    const void *data,
333 				    loff_t off, size_t len);
334 extern void ceph_copy_from_page_vector(struct page **pages,
335 				    void *data,
336 				    loff_t off, size_t len);
337 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
338 
339 
340 #endif /* _FS_CEPH_SUPER_H */
341