Merge tag 'ceph-for-4.9-rc1' of git://github.com/ceph/ceph-client
[cascardo/linux.git] / include / linux / ceph / libceph.h
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
52         /*
53          * any type that can't be simply compared or doesn't need need
54          * to be compared should go beyond this point,
55          * ceph_compare_options() should be updated accordingly
56          */
57
58         struct ceph_entity_addr *mon_addr; /* should be the first
59                                               pointer type of args */
60         int num_mon;
61         char *name;
62         struct ceph_crypto_key *key;
63 };
64
65 /*
66  * defaults
67  */
68 #define CEPH_MOUNT_TIMEOUT_DEFAULT      msecs_to_jiffies(60 * 1000)
69 #define CEPH_OSD_KEEPALIVE_DEFAULT      msecs_to_jiffies(5 * 1000)
70 #define CEPH_OSD_IDLE_TTL_DEFAULT       msecs_to_jiffies(60 * 1000)
71
72 #define CEPH_MONC_HUNT_INTERVAL         msecs_to_jiffies(3 * 1000)
73 #define CEPH_MONC_PING_INTERVAL         msecs_to_jiffies(10 * 1000)
74 #define CEPH_MONC_PING_TIMEOUT          msecs_to_jiffies(30 * 1000)
75 #define CEPH_MONC_HUNT_BACKOFF          2
76 #define CEPH_MONC_HUNT_MAX_MULT         10
77
78 #define CEPH_MSG_MAX_FRONT_LEN  (16*1024*1024)
79 #define CEPH_MSG_MAX_MIDDLE_LEN (16*1024*1024)
80 #define CEPH_MSG_MAX_DATA_LEN   (16*1024*1024)
81
82 #define CEPH_AUTH_NAME_DEFAULT   "guest"
83
84 /*
85  * Delay telling the MDS we no longer want caps, in case we reopen
86  * the file.  Delay a minimum amount of time, even if we send a cap
87  * message for some other reason.  Otherwise, take the oppotunity to
88  * update the mds to avoid sending another message later.
89  */
90 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
91 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
92
93 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT        (CEPH_CAPS_PER_RELEASE * 4)
94
95 /* mount state */
96 enum {
97         CEPH_MOUNT_MOUNTING,
98         CEPH_MOUNT_MOUNTED,
99         CEPH_MOUNT_UNMOUNTING,
100         CEPH_MOUNT_UNMOUNTED,
101         CEPH_MOUNT_SHUTDOWN,
102 };
103
104 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
105 {
106         return timeout ?: MAX_SCHEDULE_TIMEOUT;
107 }
108
109 struct ceph_mds_client;
110
111 /*
112  * per client state
113  *
114  * possibly shared by multiple mount points, if they are
115  * mounting the same ceph filesystem/cluster.
116  */
117 struct ceph_client {
118         struct ceph_fsid fsid;
119         bool have_fsid;
120
121         void *private;
122
123         struct ceph_options *options;
124
125         struct mutex mount_mutex;      /* serialize mount attempts */
126         wait_queue_head_t auth_wq;
127         int auth_err;
128
129         int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
130
131         u64 supported_features;
132         u64 required_features;
133
134         struct ceph_messenger msgr;   /* messenger instance */
135         struct ceph_mon_client monc;
136         struct ceph_osd_client osdc;
137
138 #ifdef CONFIG_DEBUG_FS
139         struct dentry *debugfs_dir;
140         struct dentry *debugfs_monmap;
141         struct dentry *debugfs_osdmap;
142         struct dentry *debugfs_options;
143 #endif
144 };
145
146 #define from_msgr(ms)   container_of(ms, struct ceph_client, msgr)
147
148
149 /*
150  * snapshots
151  */
152
153 /*
154  * A "snap context" is the set of existing snapshots when we
155  * write data.  It is used by the OSD to guide its COW behavior.
156  *
157  * The ceph_snap_context is refcounted, and attached to each dirty
158  * page, indicating which context the dirty data belonged when it was
159  * dirtied.
160  */
161 struct ceph_snap_context {
162         atomic_t nref;
163         u64 seq;
164         u32 num_snaps;
165         u64 snaps[];
166 };
167
168 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
169                                         gfp_t gfp_flags);
170 extern struct ceph_snap_context *ceph_get_snap_context(
171                                         struct ceph_snap_context *sc);
172 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
173
174 /*
175  * calculate the number of pages a given length and offset map onto,
176  * if we align the data.
177  */
178 static inline int calc_pages_for(u64 off, u64 len)
179 {
180         return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
181                 (off >> PAGE_SHIFT);
182 }
183
184 /*
185  * These are not meant to be generic - an integer key is assumed.
186  */
187 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, 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                                                                         \
198                 parent = *n;                                            \
199                 if (t->keyfld < cur->keyfld)                            \
200                         n = &(*n)->rb_left;                             \
201                 else if (t->keyfld > cur->keyfld)                       \
202                         n = &(*n)->rb_right;                            \
203                 else                                                    \
204                         BUG();                                          \
205         }                                                               \
206                                                                         \
207         rb_link_node(&t->nodefld, parent, n);                           \
208         rb_insert_color(&t->nodefld, root);                             \
209 }                                                                       \
210 static void erase_##name(struct rb_root *root, type *t)                 \
211 {                                                                       \
212         BUG_ON(RB_EMPTY_NODE(&t->nodefld));                             \
213         rb_erase(&t->nodefld, root);                                    \
214         RB_CLEAR_NODE(&t->nodefld);                                     \
215 }
216
217 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)              \
218 extern type __lookup_##name##_key;                                      \
219 static type *lookup_##name(struct rb_root *root,                        \
220                            typeof(__lookup_##name##_key.keyfld) key)    \
221 {                                                                       \
222         struct rb_node *n = root->rb_node;                              \
223                                                                         \
224         while (n) {                                                     \
225                 type *cur = rb_entry(n, type, nodefld);                 \
226                                                                         \
227                 if (key < cur->keyfld)                                  \
228                         n = n->rb_left;                                 \
229                 else if (key > cur->keyfld)                             \
230                         n = n->rb_right;                                \
231                 else                                                    \
232                         return cur;                                     \
233         }                                                               \
234                                                                         \
235         return NULL;                                                    \
236 }
237
238 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)                    \
239 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)                     \
240 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
241
242 extern struct kmem_cache *ceph_inode_cachep;
243 extern struct kmem_cache *ceph_cap_cachep;
244 extern struct kmem_cache *ceph_cap_flush_cachep;
245 extern struct kmem_cache *ceph_dentry_cachep;
246 extern struct kmem_cache *ceph_file_cachep;
247
248 /* ceph_common.c */
249 extern bool libceph_compatible(void *data);
250
251 extern const char *ceph_msg_type_name(int type);
252 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
253 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
254
255 extern struct ceph_options *ceph_parse_options(char *options,
256                               const char *dev_name, const char *dev_name_end,
257                               int (*parse_extra_token)(char *c, void *private),
258                               void *private);
259 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client);
260 extern void ceph_destroy_options(struct ceph_options *opt);
261 extern int ceph_compare_options(struct ceph_options *new_opt,
262                                 struct ceph_client *client);
263 extern struct ceph_client *ceph_create_client(struct ceph_options *opt,
264                                               void *private,
265                                               u64 supported_features,
266                                               u64 required_features);
267 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
268 u64 ceph_client_gid(struct ceph_client *client);
269 extern void ceph_destroy_client(struct ceph_client *client);
270 extern int __ceph_open_session(struct ceph_client *client,
271                                unsigned long started);
272 extern int ceph_open_session(struct ceph_client *client);
273
274 /* pagevec.c */
275 extern void ceph_release_page_vector(struct page **pages, int num_pages);
276
277 extern struct page **ceph_get_direct_page_vector(const void __user *data,
278                                                  int num_pages,
279                                                  bool write_page);
280 extern void ceph_put_page_vector(struct page **pages, int num_pages,
281                                  bool dirty);
282 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
283 extern int ceph_copy_user_to_page_vector(struct page **pages,
284                                          const void __user *data,
285                                          loff_t off, size_t len);
286 extern void ceph_copy_to_page_vector(struct page **pages,
287                                     const void *data,
288                                     loff_t off, size_t len);
289 extern void ceph_copy_from_page_vector(struct page **pages,
290                                     void *data,
291                                     loff_t off, size_t len);
292 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
293
294
295 #endif /* _FS_CEPH_SUPER_H */