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