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 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_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) 154 155 156 /* 157 * snapshots 158 */ 159 160 /* 161 * A "snap context" is the set of existing snapshots when we 162 * write data. It is used by the OSD to guide its COW behavior. 163 * 164 * The ceph_snap_context is refcounted, and attached to each dirty 165 * page, indicating which context the dirty data belonged when it was 166 * dirtied. 167 */ 168 struct ceph_snap_context { 169 refcount_t nref; 170 u64 seq; 171 u32 num_snaps; 172 u64 snaps[]; 173 }; 174 175 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count, 176 gfp_t gfp_flags); 177 extern struct ceph_snap_context *ceph_get_snap_context( 178 struct ceph_snap_context *sc); 179 extern void ceph_put_snap_context(struct ceph_snap_context *sc); 180 181 /* 182 * calculate the number of pages a given length and offset map onto, 183 * if we align the data. 184 */ 185 static inline int calc_pages_for(u64 off, u64 len) 186 { 187 return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) - 188 (off >> PAGE_SHIFT); 189 } 190 191 #define RB_BYVAL(a) (a) 192 #define RB_BYPTR(a) (&(a)) 193 #define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b)) 194 195 #define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \ 196 static bool __insert_##name(struct rb_root *root, type *t) \ 197 { \ 198 struct rb_node **n = &root->rb_node; \ 199 struct rb_node *parent = NULL; \ 200 \ 201 BUG_ON(!RB_EMPTY_NODE(&t->nodefld)); \ 202 \ 203 while (*n) { \ 204 type *cur = rb_entry(*n, type, nodefld); \ 205 int cmp; \ 206 \ 207 parent = *n; \ 208 cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld)); \ 209 if (cmp < 0) \ 210 n = &(*n)->rb_left; \ 211 else if (cmp > 0) \ 212 n = &(*n)->rb_right; \ 213 else \ 214 return false; \ 215 } \ 216 \ 217 rb_link_node(&t->nodefld, parent, n); \ 218 rb_insert_color(&t->nodefld, root); \ 219 return true; \ 220 } \ 221 static void __maybe_unused insert_##name(struct rb_root *root, type *t) \ 222 { \ 223 if (!__insert_##name(root, t)) \ 224 BUG(); \ 225 } \ 226 static void erase_##name(struct rb_root *root, type *t) \ 227 { \ 228 BUG_ON(RB_EMPTY_NODE(&t->nodefld)); \ 229 rb_erase(&t->nodefld, root); \ 230 RB_CLEAR_NODE(&t->nodefld); \ 231 } 232 233 /* 234 * @lookup_param_type is a parameter and not constructed from (@type, 235 * @keyfld) with typeof() because adding const is too unwieldy. 236 */ 237 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp, \ 238 lookup_param_type, nodefld) \ 239 static type *lookup_##name(struct rb_root *root, lookup_param_type key) \ 240 { \ 241 struct rb_node *n = root->rb_node; \ 242 \ 243 while (n) { \ 244 type *cur = rb_entry(n, type, nodefld); \ 245 int cmp; \ 246 \ 247 cmp = cmpexp(key, keyexp(cur->keyfld)); \ 248 if (cmp < 0) \ 249 n = n->rb_left; \ 250 else if (cmp > 0) \ 251 n = n->rb_right; \ 252 else \ 253 return cur; \ 254 } \ 255 \ 256 return NULL; \ 257 } 258 259 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp, \ 260 lookup_param_type, nodefld) \ 261 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \ 262 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp, \ 263 lookup_param_type, nodefld) 264 265 /* 266 * Shorthands for integer keys. 267 */ 268 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld) \ 269 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld) 270 271 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld) \ 272 extern type __lookup_##name##_key; \ 273 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, \ 274 typeof(__lookup_##name##_key.keyfld), nodefld) 275 276 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld) \ 277 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld) \ 278 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld) 279 280 extern struct kmem_cache *ceph_inode_cachep; 281 extern struct kmem_cache *ceph_cap_cachep; 282 extern struct kmem_cache *ceph_cap_flush_cachep; 283 extern struct kmem_cache *ceph_dentry_cachep; 284 extern struct kmem_cache *ceph_file_cachep; 285 extern struct kmem_cache *ceph_dir_file_cachep; 286 extern struct kmem_cache *ceph_mds_request_cachep; 287 extern mempool_t *ceph_wb_pagevec_pool; 288 289 /* ceph_common.c */ 290 extern bool libceph_compatible(void *data); 291 292 extern const char *ceph_msg_type_name(int type); 293 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid); 294 extern void *ceph_kvmalloc(size_t size, gfp_t flags); 295 296 struct fs_parameter; 297 struct fc_log; 298 struct ceph_options *ceph_alloc_options(void); 299 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt, 300 struct fc_log *l); 301 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt, 302 struct fc_log *l); 303 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client, 304 bool show_all); 305 extern void ceph_destroy_options(struct ceph_options *opt); 306 extern int ceph_compare_options(struct ceph_options *new_opt, 307 struct ceph_client *client); 308 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private); 309 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client); 310 u64 ceph_client_gid(struct ceph_client *client); 311 extern void ceph_destroy_client(struct ceph_client *client); 312 extern void ceph_reset_client_addr(struct ceph_client *client); 313 extern int __ceph_open_session(struct ceph_client *client, 314 unsigned long started); 315 extern int ceph_open_session(struct ceph_client *client); 316 int ceph_wait_for_latest_osdmap(struct ceph_client *client, 317 unsigned long timeout); 318 319 /* pagevec.c */ 320 extern void ceph_release_page_vector(struct page **pages, int num_pages); 321 extern void ceph_put_page_vector(struct page **pages, int num_pages, 322 bool dirty); 323 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags); 324 extern int ceph_copy_user_to_page_vector(struct page **pages, 325 const void __user *data, 326 loff_t off, size_t len); 327 extern void ceph_copy_to_page_vector(struct page **pages, 328 const void *data, 329 loff_t off, size_t len); 330 extern void ceph_copy_from_page_vector(struct page **pages, 331 void *data, 332 loff_t off, size_t len); 333 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages); 334 335 336 #endif /* _FS_CEPH_SUPER_H */ 337