Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
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#include <linux/refcount.h>
18
19#include <linux/ceph/types.h>
20#include <linux/ceph/messenger.h>
21#include <linux/ceph/msgpool.h>
22#include <linux/ceph/mon_client.h>
23#include <linux/ceph/osd_client.h>
24#include <linux/ceph/ceph_fs.h>
25#include <linux/ceph/string_table.h>
26
27/*
28 * mount options
29 */
30#define CEPH_OPT_FSID (1<<0)
31#define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */
32#define CEPH_OPT_MYIP (1<<2) /* specified my ip */
33#define CEPH_OPT_NOCRC (1<<3) /* no data crc on writes */
34#define CEPH_OPT_NOMSGAUTH (1<<4) /* don't require msg signing feat */
35#define CEPH_OPT_TCP_NODELAY (1<<5) /* TCP_NODELAY on TCP sockets */
36#define CEPH_OPT_NOMSGSIGN (1<<6) /* don't sign msgs */
37
38#define CEPH_OPT_DEFAULT (CEPH_OPT_TCP_NODELAY)
39
40#define ceph_set_opt(client, opt) \
41 (client)->options->flags |= CEPH_OPT_##opt;
42#define ceph_test_opt(client, opt) \
43 (!!((client)->options->flags & CEPH_OPT_##opt))
44
45struct ceph_options {
46 int flags;
47 struct ceph_fsid fsid;
48 struct ceph_entity_addr my_addr;
49 unsigned long mount_timeout; /* jiffies */
50 unsigned long osd_idle_ttl; /* jiffies */
51 unsigned long osd_keepalive_timeout; /* jiffies */
52 unsigned long osd_request_timeout; /* jiffies */
53
54 /*
55 * any type that can't be simply compared or doesn't need need
56 * to be compared should go beyond this point,
57 * ceph_compare_options() should be updated accordingly
58 */
59
60 struct ceph_entity_addr *mon_addr; /* should be the first
61 pointer type of args */
62 int num_mon;
63 char *name;
64 struct ceph_crypto_key *key;
65};
66
67/*
68 * defaults
69 */
70#define CEPH_MOUNT_TIMEOUT_DEFAULT msecs_to_jiffies(60 * 1000)
71#define CEPH_OSD_KEEPALIVE_DEFAULT msecs_to_jiffies(5 * 1000)
72#define CEPH_OSD_IDLE_TTL_DEFAULT msecs_to_jiffies(60 * 1000)
73#define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0 /* no timeout */
74
75#define CEPH_MONC_HUNT_INTERVAL msecs_to_jiffies(3 * 1000)
76#define CEPH_MONC_PING_INTERVAL msecs_to_jiffies(10 * 1000)
77#define CEPH_MONC_PING_TIMEOUT msecs_to_jiffies(30 * 1000)
78#define CEPH_MONC_HUNT_BACKOFF 2
79#define CEPH_MONC_HUNT_MAX_MULT 10
80
81#define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024)
82#define CEPH_MSG_MAX_MIDDLE_LEN (16*1024*1024)
83#define CEPH_MSG_MAX_DATA_LEN (16*1024*1024)
84
85#define CEPH_AUTH_NAME_DEFAULT "guest"
86
87/*
88 * Delay telling the MDS we no longer want caps, in case we reopen
89 * the file. Delay a minimum amount of time, even if we send a cap
90 * message for some other reason. Otherwise, take the oppotunity to
91 * update the mds to avoid sending another message later.
92 */
93#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
94#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
95
96#define CEPH_CAP_RELEASE_SAFETY_DEFAULT (CEPH_CAPS_PER_RELEASE * 4)
97
98/* mount state */
99enum {
100 CEPH_MOUNT_MOUNTING,
101 CEPH_MOUNT_MOUNTED,
102 CEPH_MOUNT_UNMOUNTING,
103 CEPH_MOUNT_UNMOUNTED,
104 CEPH_MOUNT_SHUTDOWN,
105};
106
107static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
108{
109 return timeout ?: MAX_SCHEDULE_TIMEOUT;
110}
111
112struct ceph_mds_client;
113
114/*
115 * per client state
116 *
117 * possibly shared by multiple mount points, if they are
118 * mounting the same ceph filesystem/cluster.
119 */
120struct ceph_client {
121 struct ceph_fsid fsid;
122 bool have_fsid;
123
124 void *private;
125
126 struct ceph_options *options;
127
128 struct mutex mount_mutex; /* serialize mount attempts */
129 wait_queue_head_t auth_wq;
130 int auth_err;
131
132 int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
133
134 u64 supported_features;
135 u64 required_features;
136
137 struct ceph_messenger msgr; /* messenger instance */
138 struct ceph_mon_client monc;
139 struct ceph_osd_client osdc;
140
141#ifdef CONFIG_DEBUG_FS
142 struct dentry *debugfs_dir;
143 struct dentry *debugfs_monmap;
144 struct dentry *debugfs_osdmap;
145 struct dentry *debugfs_options;
146#endif
147};
148
149#define from_msgr(ms) container_of(ms, struct ceph_client, msgr)
150
151
152/*
153 * snapshots
154 */
155
156/*
157 * A "snap context" is the set of existing snapshots when we
158 * write data. It is used by the OSD to guide its COW behavior.
159 *
160 * The ceph_snap_context is refcounted, and attached to each dirty
161 * page, indicating which context the dirty data belonged when it was
162 * dirtied.
163 */
164struct ceph_snap_context {
165 refcount_t nref;
166 u64 seq;
167 u32 num_snaps;
168 u64 snaps[];
169};
170
171extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
172 gfp_t gfp_flags);
173extern struct ceph_snap_context *ceph_get_snap_context(
174 struct ceph_snap_context *sc);
175extern void ceph_put_snap_context(struct ceph_snap_context *sc);
176
177/*
178 * calculate the number of pages a given length and offset map onto,
179 * if we align the data.
180 */
181static inline int calc_pages_for(u64 off, u64 len)
182{
183 return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
184 (off >> PAGE_SHIFT);
185}
186
187#define RB_BYVAL(a) (a)
188#define RB_BYPTR(a) (&(a))
189#define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
190
191#define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
192static void insert_##name(struct rb_root *root, type *t) \
193{ \
194 struct rb_node **n = &root->rb_node; \
195 struct rb_node *parent = NULL; \
196 \
197 BUG_ON(!RB_EMPTY_NODE(&t->nodefld)); \
198 \
199 while (*n) { \
200 type *cur = rb_entry(*n, type, nodefld); \
201 int cmp; \
202 \
203 parent = *n; \
204 cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld)); \
205 if (cmp < 0) \
206 n = &(*n)->rb_left; \
207 else if (cmp > 0) \
208 n = &(*n)->rb_right; \
209 else \
210 BUG(); \
211 } \
212 \
213 rb_link_node(&t->nodefld, parent, n); \
214 rb_insert_color(&t->nodefld, root); \
215} \
216static void erase_##name(struct rb_root *root, type *t) \
217{ \
218 BUG_ON(RB_EMPTY_NODE(&t->nodefld)); \
219 rb_erase(&t->nodefld, root); \
220 RB_CLEAR_NODE(&t->nodefld); \
221}
222
223/*
224 * @lookup_param_type is a parameter and not constructed from (@type,
225 * @keyfld) with typeof() because adding const is too unwieldy.
226 */
227#define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp, \
228 lookup_param_type, nodefld) \
229static type *lookup_##name(struct rb_root *root, lookup_param_type key) \
230{ \
231 struct rb_node *n = root->rb_node; \
232 \
233 while (n) { \
234 type *cur = rb_entry(n, type, nodefld); \
235 int cmp; \
236 \
237 cmp = cmpexp(key, keyexp(cur->keyfld)); \
238 if (cmp < 0) \
239 n = n->rb_left; \
240 else if (cmp > 0) \
241 n = n->rb_right; \
242 else \
243 return cur; \
244 } \
245 \
246 return NULL; \
247}
248
249#define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp, \
250 lookup_param_type, nodefld) \
251DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
252DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp, \
253 lookup_param_type, nodefld)
254
255/*
256 * Shorthands for integer keys.
257 */
258#define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld) \
259DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
260
261#define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld) \
262extern type __lookup_##name##_key; \
263DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, \
264 typeof(__lookup_##name##_key.keyfld), nodefld)
265
266#define DEFINE_RB_FUNCS(name, type, keyfld, nodefld) \
267DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld) \
268DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
269
270extern struct kmem_cache *ceph_inode_cachep;
271extern struct kmem_cache *ceph_cap_cachep;
272extern struct kmem_cache *ceph_cap_flush_cachep;
273extern struct kmem_cache *ceph_dentry_cachep;
274extern struct kmem_cache *ceph_file_cachep;
275
276/* ceph_common.c */
277extern bool libceph_compatible(void *data);
278
279extern const char *ceph_msg_type_name(int type);
280extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
281extern void *ceph_kvmalloc(size_t size, gfp_t flags);
282
283extern struct ceph_options *ceph_parse_options(char *options,
284 const char *dev_name, const char *dev_name_end,
285 int (*parse_extra_token)(char *c, void *private),
286 void *private);
287int ceph_print_client_options(struct seq_file *m, struct ceph_client *client);
288extern void ceph_destroy_options(struct ceph_options *opt);
289extern int ceph_compare_options(struct ceph_options *new_opt,
290 struct ceph_client *client);
291struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
292struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
293u64 ceph_client_gid(struct ceph_client *client);
294extern void ceph_destroy_client(struct ceph_client *client);
295extern int __ceph_open_session(struct ceph_client *client,
296 unsigned long started);
297extern int ceph_open_session(struct ceph_client *client);
298
299/* pagevec.c */
300extern void ceph_release_page_vector(struct page **pages, int num_pages);
301
302extern struct page **ceph_get_direct_page_vector(const void __user *data,
303 int num_pages,
304 bool write_page);
305extern void ceph_put_page_vector(struct page **pages, int num_pages,
306 bool dirty);
307extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
308extern int ceph_copy_user_to_page_vector(struct page **pages,
309 const void __user *data,
310 loff_t off, size_t len);
311extern void ceph_copy_to_page_vector(struct page **pages,
312 const void *data,
313 loff_t off, size_t len);
314extern void ceph_copy_from_page_vector(struct page **pages,
315 void *data,
316 loff_t off, size_t len);
317extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
318
319
320#endif /* _FS_CEPH_SUPER_H */