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 56 /* 57 * any type that can't be simply compared or doesn't need 58 * to be compared should go beyond this point, 59 * ceph_compare_options() should be updated accordingly 60 */ 61 62 struct ceph_entity_addr *mon_addr; /* should be the first 63 pointer type of args */ 64 int num_mon; 65 char *name; 66 struct ceph_crypto_key *key; 67 struct rb_root crush_locs; 68 }; 69 70 /* 71 * defaults 72 */ 73 #define CEPH_MOUNT_TIMEOUT_DEFAULT msecs_to_jiffies(60 * 1000) 74 #define CEPH_OSD_KEEPALIVE_DEFAULT msecs_to_jiffies(5 * 1000) 75 #define CEPH_OSD_IDLE_TTL_DEFAULT msecs_to_jiffies(60 * 1000) 76 #define CEPH_OSD_REQUEST_TIMEOUT_DEFAULT 0 /* no timeout */ 77 78 #define CEPH_MONC_HUNT_INTERVAL msecs_to_jiffies(3 * 1000) 79 #define CEPH_MONC_PING_INTERVAL msecs_to_jiffies(10 * 1000) 80 #define CEPH_MONC_PING_TIMEOUT msecs_to_jiffies(30 * 1000) 81 #define CEPH_MONC_HUNT_BACKOFF 2 82 #define CEPH_MONC_HUNT_MAX_MULT 10 83 84 #define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024) 85 #define CEPH_MSG_MAX_MIDDLE_LEN (16*1024*1024) 86 87 /* 88 * The largest possible rbd data object is 32M. 89 * The largest possible rbd object map object is 64M. 90 * 91 * There is no limit on the size of cephfs objects, but it has to obey 92 * rsize and wsize mount options anyway. 93 */ 94 #define CEPH_MSG_MAX_DATA_LEN (64*1024*1024) 95 96 #define CEPH_AUTH_NAME_DEFAULT "guest" 97 98 /* mount state */ 99 enum { 100 CEPH_MOUNT_MOUNTING, 101 CEPH_MOUNT_MOUNTED, 102 CEPH_MOUNT_UNMOUNTING, 103 CEPH_MOUNT_UNMOUNTED, 104 CEPH_MOUNT_SHUTDOWN, 105 }; 106 107 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout) 108 { 109 return timeout ?: MAX_SCHEDULE_TIMEOUT; 110 } 111 112 struct 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 */ 120 struct 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 */ 164 struct ceph_snap_context { 165 refcount_t nref; 166 u64 seq; 167 u32 num_snaps; 168 u64 snaps[]; 169 }; 170 171 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count, 172 gfp_t gfp_flags); 173 extern struct ceph_snap_context *ceph_get_snap_context( 174 struct ceph_snap_context *sc); 175 extern 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 */ 181 static 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) \ 192 static bool __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 return false; \ 211 } \ 212 \ 213 rb_link_node(&t->nodefld, parent, n); \ 214 rb_insert_color(&t->nodefld, root); \ 215 return true; \ 216 } \ 217 static void __maybe_unused insert_##name(struct rb_root *root, type *t) \ 218 { \ 219 if (!__insert_##name(root, t)) \ 220 BUG(); \ 221 } \ 222 static void erase_##name(struct rb_root *root, type *t) \ 223 { \ 224 BUG_ON(RB_EMPTY_NODE(&t->nodefld)); \ 225 rb_erase(&t->nodefld, root); \ 226 RB_CLEAR_NODE(&t->nodefld); \ 227 } 228 229 /* 230 * @lookup_param_type is a parameter and not constructed from (@type, 231 * @keyfld) with typeof() because adding const is too unwieldy. 232 */ 233 #define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp, \ 234 lookup_param_type, nodefld) \ 235 static type *lookup_##name(struct rb_root *root, lookup_param_type key) \ 236 { \ 237 struct rb_node *n = root->rb_node; \ 238 \ 239 while (n) { \ 240 type *cur = rb_entry(n, type, nodefld); \ 241 int cmp; \ 242 \ 243 cmp = cmpexp(key, keyexp(cur->keyfld)); \ 244 if (cmp < 0) \ 245 n = n->rb_left; \ 246 else if (cmp > 0) \ 247 n = n->rb_right; \ 248 else \ 249 return cur; \ 250 } \ 251 \ 252 return NULL; \ 253 } 254 255 #define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp, \ 256 lookup_param_type, nodefld) \ 257 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \ 258 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp, \ 259 lookup_param_type, nodefld) 260 261 /* 262 * Shorthands for integer keys. 263 */ 264 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld) \ 265 DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld) 266 267 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld) \ 268 extern type __lookup_##name##_key; \ 269 DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, \ 270 typeof(__lookup_##name##_key.keyfld), nodefld) 271 272 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld) \ 273 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld) \ 274 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld) 275 276 extern struct kmem_cache *ceph_inode_cachep; 277 extern struct kmem_cache *ceph_cap_cachep; 278 extern struct kmem_cache *ceph_cap_flush_cachep; 279 extern struct kmem_cache *ceph_dentry_cachep; 280 extern struct kmem_cache *ceph_file_cachep; 281 extern struct kmem_cache *ceph_dir_file_cachep; 282 extern struct kmem_cache *ceph_mds_request_cachep; 283 284 /* ceph_common.c */ 285 extern bool libceph_compatible(void *data); 286 287 extern const char *ceph_msg_type_name(int type); 288 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid); 289 extern void *ceph_kvmalloc(size_t size, gfp_t flags); 290 291 struct fs_parameter; 292 struct fc_log; 293 struct ceph_options *ceph_alloc_options(void); 294 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt, 295 struct fc_log *l); 296 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt, 297 struct fc_log *l); 298 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client, 299 bool show_all); 300 extern void ceph_destroy_options(struct ceph_options *opt); 301 extern int ceph_compare_options(struct ceph_options *new_opt, 302 struct ceph_client *client); 303 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private); 304 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client); 305 u64 ceph_client_gid(struct ceph_client *client); 306 extern void ceph_destroy_client(struct ceph_client *client); 307 extern void ceph_reset_client_addr(struct ceph_client *client); 308 extern int __ceph_open_session(struct ceph_client *client, 309 unsigned long started); 310 extern int ceph_open_session(struct ceph_client *client); 311 int ceph_wait_for_latest_osdmap(struct ceph_client *client, 312 unsigned long timeout); 313 314 /* pagevec.c */ 315 extern void ceph_release_page_vector(struct page **pages, int num_pages); 316 extern void ceph_put_page_vector(struct page **pages, int num_pages, 317 bool dirty); 318 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags); 319 extern int ceph_copy_user_to_page_vector(struct page **pages, 320 const void __user *data, 321 loff_t off, size_t len); 322 extern void ceph_copy_to_page_vector(struct page **pages, 323 const void *data, 324 loff_t off, size_t len); 325 extern void ceph_copy_from_page_vector(struct page **pages, 326 void *data, 327 loff_t off, size_t len); 328 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages); 329 330 331 #endif /* _FS_CEPH_SUPER_H */ 332