libceph.h

  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 <linux/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#define CEPH_OPT_RXBOUNCE         (1<<7) /* double-buffer read data */
 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
 47struct 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	int con_modes[2];  /* CEPH_CON_MODE_* */
 57
 58	/*
 59	 * any type that can't be simply compared or doesn't need
 60	 * to be compared should go beyond this point,
 61	 * ceph_compare_options() should be updated accordingly
 62	 */
 63
 64	struct ceph_entity_addr *mon_addr; /* should be the first
 65					      pointer type of args */
 66	int num_mon;
 67	char *name;
 68	struct ceph_crypto_key *key;
 69	struct rb_root crush_locs;
 70};
 71
 72/*
 73 * defaults
 74 */
 75#define CEPH_MOUNT_TIMEOUT_DEFAULT	msecs_to_jiffies(60 * 1000)
 76#define CEPH_OSD_KEEPALIVE_DEFAULT	msecs_to_jiffies(5 * 1000)
 77#define CEPH_OSD_IDLE_TTL_DEFAULT	msecs_to_jiffies(60 * 1000)
 78#define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0  /* no timeout */
 79#define CEPH_READ_FROM_REPLICA_DEFAULT	0  /* read from primary */
 80
 81#define CEPH_MONC_HUNT_INTERVAL		msecs_to_jiffies(3 * 1000)
 82#define CEPH_MONC_PING_INTERVAL		msecs_to_jiffies(10 * 1000)
 83#define CEPH_MONC_PING_TIMEOUT		msecs_to_jiffies(30 * 1000)
 84#define CEPH_MONC_HUNT_BACKOFF		2
 85#define CEPH_MONC_HUNT_MAX_MULT		10
 86
 87#define CEPH_MSG_MAX_CONTROL_LEN (16*1024*1024)
 88#define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
 89#define CEPH_MSG_MAX_MIDDLE_LEN	(16*1024*1024)
 90
 91/*
 92 * The largest possible rbd data object is 32M.
 93 * The largest possible rbd object map object is 64M.
 94 *
 95 * There is no limit on the size of cephfs objects, but it has to obey
 96 * rsize and wsize mount options anyway.
 97 */
 98#define CEPH_MSG_MAX_DATA_LEN	(64*1024*1024)
 99
100#define CEPH_AUTH_NAME_DEFAULT   "guest"
101
102static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
103{
104	return timeout ?: MAX_SCHEDULE_TIMEOUT;
105}
106
107struct ceph_mds_client;
108
109/*
110 * per client state
111 *
112 * possibly shared by multiple mount points, if they are
113 * mounting the same ceph filesystem/cluster.
114 */
115struct ceph_client {
116	struct ceph_fsid fsid;
117	bool have_fsid;
118
119	void *private;
120
121	struct ceph_options *options;
122
123	struct mutex mount_mutex;      /* serialize mount attempts */
124	wait_queue_head_t auth_wq;
125	int auth_err;
126
127	int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
128
129	u64 supported_features;
130	u64 required_features;
131
132	struct ceph_messenger msgr;   /* messenger instance */
133	struct ceph_mon_client monc;
134	struct ceph_osd_client osdc;
135
136#ifdef CONFIG_DEBUG_FS
137	struct dentry *debugfs_dir;
138	struct dentry *debugfs_monmap;
139	struct dentry *debugfs_osdmap;
140	struct dentry *debugfs_options;
141#endif
142};
143
144#define from_msgr(ms)	container_of(ms, struct ceph_client, msgr)
145
146static inline bool ceph_msgr2(struct ceph_client *client)
147{
148	return client->options->con_modes[0] != CEPH_CON_MODE_UNKNOWN;
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 */
163struct ceph_snap_context {
164	refcount_t nref;
165	u64 seq;
166	u32 num_snaps;
167	u64 snaps[];
168};
169
170extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
171					gfp_t gfp_flags);
172extern struct ceph_snap_context *ceph_get_snap_context(
173					struct ceph_snap_context *sc);
174extern 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 */
180static 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) \
191static bool __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			return false;					\
210	}								\
211									\
212	rb_link_node(&t->nodefld, parent, n);				\
213	rb_insert_color(&t->nodefld, root);				\
214	return true;							\
215}									\
216static void __maybe_unused insert_##name(struct rb_root *root, type *t)	\
217{									\
218	if (!__insert_##name(root, t))					\
219		BUG();							\
220}									\
221static void erase_##name(struct rb_root *root, type *t)			\
222{									\
223	BUG_ON(RB_EMPTY_NODE(&t->nodefld));				\
224	rb_erase(&t->nodefld, root);					\
225	RB_CLEAR_NODE(&t->nodefld);					\
226}
227
228/*
229 * @lookup_param_type is a parameter and not constructed from (@type,
230 * @keyfld) with typeof() because adding const is too unwieldy.
231 */
232#define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
233			       lookup_param_type, nodefld)		\
234static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
235{									\
236	struct rb_node *n = root->rb_node;				\
237									\
238	while (n) {							\
239		type *cur = rb_entry(n, type, nodefld);			\
240		int cmp;						\
241									\
242		cmp = cmpexp(key, keyexp(cur->keyfld));			\
243		if (cmp < 0)						\
244			n = n->rb_left;					\
245		else if (cmp > 0)					\
246			n = n->rb_right;				\
247		else							\
248			return cur;					\
249	}								\
250									\
251	return NULL;							\
252}
253
254#define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
255			 lookup_param_type, nodefld)			\
256DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
257DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
258		       lookup_param_type, nodefld)
259
260/*
261 * Shorthands for integer keys.
262 */
263#define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
264DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
265
266#define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
267extern type __lookup_##name##_key;					\
268DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
269		       typeof(__lookup_##name##_key.keyfld), nodefld)
270
271#define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
272DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
273DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
274
275extern struct kmem_cache *ceph_inode_cachep;
276extern struct kmem_cache *ceph_cap_cachep;
277extern struct kmem_cache *ceph_cap_snap_cachep;
278extern struct kmem_cache *ceph_cap_flush_cachep;
279extern struct kmem_cache *ceph_dentry_cachep;
280extern struct kmem_cache *ceph_file_cachep;
281extern struct kmem_cache *ceph_dir_file_cachep;
282extern struct kmem_cache *ceph_mds_request_cachep;
283extern mempool_t *ceph_wb_pagevec_pool;
284
285/* ceph_common.c */
286extern bool libceph_compatible(void *data);
287
288extern const char *ceph_msg_type_name(int type);
289extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
290extern int ceph_parse_fsid(const char *str, struct ceph_fsid *fsid);
291
292struct fs_parameter;
293struct fc_log;
294struct ceph_options *ceph_alloc_options(void);
295int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
296		       struct fc_log *l, char delim);
297int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
298		     struct fc_log *l);
299int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
300			      bool show_all);
301extern void ceph_destroy_options(struct ceph_options *opt);
302extern int ceph_compare_options(struct ceph_options *new_opt,
303				struct ceph_client *client);
304struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private);
305struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
306u64 ceph_client_gid(struct ceph_client *client);
307extern void ceph_destroy_client(struct ceph_client *client);
308extern void ceph_reset_client_addr(struct ceph_client *client);
309extern int __ceph_open_session(struct ceph_client *client);
310extern int ceph_open_session(struct ceph_client *client);
311int ceph_wait_for_latest_osdmap(struct ceph_client *client,
312				unsigned long timeout);
313
314/* pagevec.c */
315extern void ceph_release_page_vector(struct page **pages, int num_pages);
316extern void ceph_put_page_vector(struct page **pages, int num_pages,
317				 bool dirty);
318extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
319extern void ceph_copy_from_page_vector(struct page **pages,
320				    void *data,
321				    loff_t off, size_t len);
322extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
323
324
325#endif /* _FS_CEPH_SUPER_H */