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
  81 /*
  82  * Handle the largest possible rbd object in one message.
  83  * There is no limit on the size of cephfs objects, but it has to obey
  84  * rsize and wsize mount options anyway.
  85  */
  86 #define CEPH_MSG_MAX_DATA_LEN   (32*1024*1024)
  87
  88 #define CEPH_AUTH_NAME_DEFAULT   "guest"
  89
  90 /*
  91  * Delay telling the MDS we no longer want caps, in case we reopen
  92  * the file.  Delay a minimum amount of time, even if we send a cap
  93  * message for some other reason.  Otherwise, take the oppotunity to
  94  * update the mds to avoid sending another message later.
  95  */
  96 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
  97 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
  98
  99 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT        (CEPH_CAPS_PER_RELEASE * 4)
 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 };
 109
 110 static inline unsigned long ceph_timeout_jiffies(unsigned long timeout)
 111 {
 112         return timeout ?: MAX_SCHEDULE_TIMEOUT;
 113 }
 114
 115 struct ceph_mds_client;
 116
 117 /*
 118  * per client state
 119  *
 120  * possibly shared by multiple mount points, if they are
 121  * mounting the same ceph filesystem/cluster.
 122  */
 123 struct ceph_client {
 124         struct ceph_fsid fsid;
 125         bool have_fsid;
 126
 127         void *private;
 128
 129         struct ceph_options *options;
 130
 131         struct mutex mount_mutex;      /* serialize mount attempts */
 132         wait_queue_head_t auth_wq;
 133         int auth_err;
 134
 135         int (*extra_mon_dispatch)(struct ceph_client *, struct ceph_msg *);
 136
 137         u64 supported_features;
 138         u64 required_features;
 139
 140         struct ceph_messenger msgr;   /* messenger instance */
 141         struct ceph_mon_client monc;
 142         struct ceph_osd_client osdc;
 143
 144 #ifdef CONFIG_DEBUG_FS
 145         struct dentry *debugfs_dir;
 146         struct dentry *debugfs_monmap;
 147         struct dentry *debugfs_osdmap;
 148         struct dentry *debugfs_options;
 149 #endif
 150 };
 151
 152 #define from_msgr(ms)   container_of(ms, struct ceph_client, msgr)
 153
 154
 155 /*
 156  * snapshots
 157  */
 158
 159 /*
 160  * A "snap context" is the set of existing snapshots when we
 161  * write data.  It is used by the OSD to guide its COW behavior.
 162  *
 163  * The ceph_snap_context is refcounted, and attached to each dirty
 164  * page, indicating which context the dirty data belonged when it was
 165  * dirtied.
 166  */
 167 struct ceph_snap_context {
 168         atomic_t nref;
 169         u64 seq;
 170         u32 num_snaps;
 171         u64 snaps[];
 172 };
 173
 174 extern struct ceph_snap_context *ceph_create_snap_context(u32 snap_count,
 175                                         gfp_t gfp_flags);
 176 extern struct ceph_snap_context *ceph_get_snap_context(
 177                                         struct ceph_snap_context *sc);
 178 extern void ceph_put_snap_context(struct ceph_snap_context *sc);
 179
 180 /*
 181  * calculate the number of pages a given length and offset map onto,
 182  * if we align the data.
 183  */
 184 static inline int calc_pages_for(u64 off, u64 len)
 185 {
 186         return ((off+len+PAGE_SIZE-1) >> PAGE_SHIFT) -
 187                 (off >> PAGE_SHIFT);
 188 }
 189
 190 /*
 191  * These are not meant to be generic - an integer key is assumed.
 192  */
 193 #define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)             \
 194 static void insert_##name(struct rb_root *root, type *t)                \
 195 {                                                                       \
 196         struct rb_node **n = &root->rb_node;                            \
 197         struct rb_node *parent = NULL;                                  \
 198                                                                         \
 199         BUG_ON(!RB_EMPTY_NODE(&t->nodefld));                            \
 200                                                                         \
 201         while (*n) {                                                    \
 202                 type *cur = rb_entry(*n, type, nodefld);                \
 203                                                                         \
 204                 parent = *n;                                            \
 205                 if (t->keyfld < cur->keyfld)                            \
 206                         n = &(*n)->rb_left;                             \
 207                 else if (t->keyfld > cur->keyfld)                       \
 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 }                                                                       \
 216 static 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 #define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)              \
 224 extern type __lookup_##name##_key;                                      \
 225 static type *lookup_##name(struct rb_root *root,                        \
 226                            typeof(__lookup_##name##_key.keyfld) key)    \
 227 {                                                                       \
 228         struct rb_node *n = root->rb_node;                              \
 229                                                                         \
 230         while (n) {                                                     \
 231                 type *cur = rb_entry(n, type, nodefld);                 \
 232                                                                         \
 233                 if (key < cur->keyfld)                                  \
 234                         n = n->rb_left;                                 \
 235                 else if (key > cur->keyfld)                             \
 236                         n = n->rb_right;                                \
 237                 else                                                    \
 238                         return cur;                                     \
 239         }                                                               \
 240                                                                         \
 241         return NULL;                                                    \
 242 }
 243
 244 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)                    \
 245 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)                     \
 246 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
 247
 248 extern struct kmem_cache *ceph_inode_cachep;
 249 extern struct kmem_cache *ceph_cap_cachep;
 250 extern struct kmem_cache *ceph_cap_flush_cachep;
 251 extern struct kmem_cache *ceph_dentry_cachep;
 252 extern struct kmem_cache *ceph_file_cachep;
 253
 254 /* ceph_common.c */
 255 extern bool libceph_compatible(void *data);
 256
 257 extern const char *ceph_msg_type_name(int type);
 258 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
 259 extern void *ceph_kvmalloc(size_t size, gfp_t flags);
 260
 261 extern struct ceph_options *ceph_parse_options(char *options,
 262                               const char *dev_name, const char *dev_name_end,
 263                               int (*parse_extra_token)(char *c, void *private),
 264                               void *private);
 265 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client);
 266 extern void ceph_destroy_options(struct ceph_options *opt);
 267 extern int ceph_compare_options(struct ceph_options *new_opt,
 268                                 struct ceph_client *client);
 269 extern struct ceph_client *ceph_create_client(struct ceph_options *opt,
 270                                               void *private,
 271                                               u64 supported_features,
 272                                               u64 required_features);
 273 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client);
 274 u64 ceph_client_gid(struct ceph_client *client);
 275 extern void ceph_destroy_client(struct ceph_client *client);
 276 extern int __ceph_open_session(struct ceph_client *client,
 277                                unsigned long started);
 278 extern int ceph_open_session(struct ceph_client *client);
 279 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
 280                                 unsigned long timeout);
 281
 282 /* pagevec.c */
 283 extern void ceph_release_page_vector(struct page **pages, int num_pages);
 284
 285 extern struct page **ceph_get_direct_page_vector(const void __user *data,
 286                                                  int num_pages,
 287                                                  bool write_page);
 288 extern void ceph_put_page_vector(struct page **pages, int num_pages,
 289                                  bool dirty);
 290 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
 291 extern int ceph_copy_user_to_page_vector(struct page **pages,
 292                                          const void __user *data,
 293                                          loff_t off, size_t len);
 294 extern void ceph_copy_to_page_vector(struct page **pages,
 295                                     const void *data,
 296                                     loff_t off, size_t len);
 297 extern void ceph_copy_from_page_vector(struct page **pages,
 298                                     void *data,
 299                                     loff_t off, size_t len);
 300 extern void ceph_zero_page_vector_range(int off, int len, struct page **pages);
 301
 302
 303 #endif /* _FS_CEPH_SUPER_H */