ceph/src/client/Inode.h

   1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
   2 // vim: ts=8 sw=2 smarttab
   3
   4 #ifndef CEPH_CLIENT_INODE_H
   5 #define CEPH_CLIENT_INODE_H
   6
   7 #include "include/types.h"
   8 #include "include/xlist.h"
   9
  10 #include "mds/mdstypes.h" // hrm
  11
  12 #include "osdc/ObjectCacher.h"
  13 #include "include/assert.h"
  14
  15 #include "InodeRef.h"
  16 #include "UserPerm.h"
  17
  18 class Client;
  19 struct MetaSession;
  20 class Dentry;
  21 class Dir;
  22 struct SnapRealm;
  23 struct Inode;
  24 class ceph_lock_state_t;
  25 class MetaRequest;
  26 class filepath;
  27 class Fh;
  28
  29 struct Cap {
  30   MetaSession *session;
  31   Inode *inode;
  32   xlist<Cap*>::item cap_item;
  33
  34   uint64_t cap_id;
  35   unsigned issued;
  36   unsigned implemented;
  37   unsigned wanted;   // as known to mds.
  38   uint64_t seq, issue_seq;
  39   __u32 mseq;  // migration seq
  40   __u32 gen;
  41   UserPerm latest_perms;
  42
  43   Cap() : session(NULL), inode(NULL), cap_item(this), cap_id(0), issued(0),
  44           implemented(0), wanted(0), seq(0), issue_seq(0), mseq(0), gen(0),
  45           latest_perms()  {}
  46
  47   void dump(Formatter *f) const;
  48 };
  49
  50 struct CapSnap {
  51   //snapid_t follows;  // map key
  52   InodeRef in;
  53   SnapContext context;
  54   int issued, dirty;
  55
  56   uint64_t size;
  57   utime_t ctime, btime, mtime, atime;
  58   version_t time_warp_seq;
  59   uint64_t change_attr;
  60   uint32_t   mode;
  61   uid_t      uid;
  62   gid_t      gid;
  63   map<string,bufferptr> xattrs;
  64   version_t xattr_version;
  65
  66   bufferlist inline_data;
  67   version_t inline_version;
  68
  69   bool writing, dirty_data;
  70   uint64_t flush_tid;
  71
  72   explicit CapSnap(Inode *i)
  73     : in(i), issued(0), dirty(0), size(0), time_warp_seq(0), change_attr(0),
  74       mode(0), uid(0), gid(0), xattr_version(0), inline_version(0),
  75       writing(false), dirty_data(false), flush_tid(0)
  76   {}
  77
  78   void dump(Formatter *f) const;
  79 };
  80
  81 // inode flags
  82 #define I_COMPLETE      1
  83 #define I_DIR_ORDERED   2
  84 #define I_CAP_DROPPED   4
  85 #define I_SNAPDIR_OPEN  8
  86
  87 struct Inode {
  88   Client *client;
  89
  90   // -- the actual inode --
  91   inodeno_t ino; // ORDER DEPENDENCY: oset
  92   snapid_t  snapid;
  93   ino_t faked_ino;
  94
  95   uint32_t   rdev;    // if special file
  96
  97   // affected by any inode change...
  98   utime_t    ctime;   // inode change time
  99   utime_t    btime;   // birth time
 100
 101   // perm (namespace permissions)
 102   uint32_t   mode;
 103   uid_t      uid;
 104   gid_t      gid;
 105
 106   // nlink
 107   int32_t    nlink;
 108
 109   // file (data access)
 110   ceph_dir_layout dir_layout;
 111   file_layout_t layout;
 112   uint64_t   size;        // on directory, # dentries
 113   uint32_t   truncate_seq;
 114   uint64_t   truncate_size;
 115   utime_t    mtime;   // file data modify time.
 116   utime_t    atime;   // file data access time.
 117   uint32_t   time_warp_seq;  // count of (potential) mtime/atime timewarps (i.e., utimes())
 118   uint64_t   change_attr;
 119
 120   uint64_t max_size;  // max size we can write to
 121
 122   // dirfrag, recursive accountin
 123   frag_info_t dirstat;
 124   nest_info_t rstat;
 125
 126   // special stuff
 127   version_t version;           // auth only
 128   version_t xattr_version;
 129
 130   // inline data
 131   version_t  inline_version;
 132   bufferlist inline_data;
 133
 134   bool is_root()    const { return ino == MDS_INO_ROOT; }
 135   bool is_symlink() const { return (mode & S_IFMT) == S_IFLNK; }
 136   bool is_dir()     const { return (mode & S_IFMT) == S_IFDIR; }
 137   bool is_file()    const { return (mode & S_IFMT) == S_IFREG; }
 138
 139   bool has_dir_layout() const {
 140     return layout != file_layout_t();
 141   }
 142
 143   __u32 hash_dentry_name(const string &dn) {
 144     int which = dir_layout.dl_dir_hash;
 145     if (!which)
 146       which = CEPH_STR_HASH_LINUX;
 147     assert(ceph_str_hash_valid(which));
 148     return ceph_str_hash(which, dn.data(), dn.length());
 149   }
 150
 151   unsigned flags;
 152
 153   quota_info_t quota;
 154
 155   bool is_complete_and_ordered() {
 156     static const unsigned wants = I_COMPLETE | I_DIR_ORDERED;
 157     return (flags & wants) == wants;
 158   }
 159
 160   // about the dir (if this is one!)
 161   Dir       *dir;     // if i'm a dir.
 162   fragtree_t dirfragtree;
 163   set<int>  dir_contacts;
 164   uint64_t dir_release_count, dir_ordered_count;
 165   bool dir_hashed, dir_replicated;
 166
 167   // per-mds caps
 168   map<mds_rank_t, Cap*> caps;            // mds -> Cap
 169   Cap *auth_cap;
 170   int64_t cap_dirtier_uid;
 171   int64_t cap_dirtier_gid;
 172   unsigned dirty_caps, flushing_caps;
 173   std::map<ceph_tid_t, int> flushing_cap_tids;
 174   int shared_gen, cache_gen;
 175   int snap_caps, snap_cap_refs;
 176   utime_t hold_caps_until;
 177   xlist<Inode*>::item cap_item, flushing_cap_item;
 178
 179   SnapRealm *snaprealm;
 180   xlist<Inode*>::item snaprealm_item;
 181   InodeRef snapdir_parent;  // only if we are a snapdir inode
 182   map<snapid_t,CapSnap> cap_snaps;   // pending flush to mds
 183
 184   //int open_by_mode[CEPH_FILE_MODE_NUM];
 185   map<int,int> open_by_mode;
 186   map<int,int> cap_refs;
 187
 188   ObjectCacher::ObjectSet oset; // ORDER DEPENDENCY: ino
 189
 190   uint64_t     reported_size, wanted_max_size, requested_max_size;
 191
 192   int       _ref;      // ref count. 1 for each dentry, fh that links to me.
 193   int       ll_ref;   // separate ref count for ll client
 194   set<Dentry*> dn_set;      // if i'm linked to a dentry.
 195   string    symlink;  // symlink content, if it's a symlink
 196   map<string,bufferptr> xattrs;
 197   map<frag_t,int> fragmap;  // known frag -> mds mappings
 198
 199   list<Cond*>       waitfor_caps;
 200   list<Cond*>       waitfor_commit;
 201
 202   Dentry *get_first_parent() {
 203     assert(!dn_set.empty());
 204     return *dn_set.begin();
 205   }
 206
 207   void make_long_path(filepath& p);
 208   void make_nosnap_relative_path(filepath& p);
 209
 210   void get();
 211   int _put(int n=1);
 212
 213   int get_num_ref() {
 214     return _ref;
 215   }
 216
 217   void ll_get() {
 218     ll_ref++;
 219   }
 220   void ll_put(int n=1) {
 221     assert(ll_ref >= n);
 222     ll_ref -= n;
 223   }
 224
 225   // file locks
 226   ceph_lock_state_t *fcntl_locks;
 227   ceph_lock_state_t *flock_locks;
 228
 229   xlist<MetaRequest*> unsafe_ops;
 230
 231   std::set<Fh*> fhs;
 232
 233   Inode(Client *c, vinodeno_t vino, file_layout_t *newlayout)
 234     : client(c), ino(vino.ino), snapid(vino.snapid), faked_ino(0),
 235       rdev(0), mode(0), uid(0), gid(0), nlink(0),
 236       size(0), truncate_seq(1), truncate_size(-1),
 237       time_warp_seq(0), change_attr(0), max_size(0), version(0),
 238       xattr_version(0), inline_version(0), flags(0),
 239       dir(0), dir_release_count(1), dir_ordered_count(1),
 240       dir_hashed(false), dir_replicated(false), auth_cap(NULL),
 241       cap_dirtier_uid(-1), cap_dirtier_gid(-1),
 242       dirty_caps(0), flushing_caps(0), shared_gen(0), cache_gen(0),
 243       snap_caps(0), snap_cap_refs(0),
 244       cap_item(this), flushing_cap_item(this),
 245       snaprealm(0), snaprealm_item(this),
 246       oset((void *)this, newlayout->pool_id, this->ino),
 247       reported_size(0), wanted_max_size(0), requested_max_size(0),
 248       _ref(0), ll_ref(0), dn_set(),
 249       fcntl_locks(NULL), flock_locks(NULL)
 250   {
 251     memset(&dir_layout, 0, sizeof(dir_layout));
 252     memset(&quota, 0, sizeof(quota));
 253   }
 254   ~Inode();
 255
 256   vinodeno_t vino() const { return vinodeno_t(ino, snapid); }
 257
 258   struct Compare {
 259     bool operator() (Inode* const & left, Inode* const & right) {
 260       if (left->ino.val < right->ino.val) {
 261         return (left->snapid.val < right->snapid.val);
 262       }
 263       return false;
 264     }
 265   };
 266
 267   bool check_mode(const UserPerm& perms, unsigned want);
 268
 269   // CAPS --------
 270   void get_open_ref(int mode);
 271   bool put_open_ref(int mode);
 272
 273   void get_cap_ref(int cap);
 274   int put_cap_ref(int cap);
 275   bool is_any_caps();
 276   bool cap_is_valid(Cap* cap) const;
 277   int caps_issued(int *implemented = 0) const;
 278   void touch_cap(Cap *cap);
 279   void try_touch_cap(mds_rank_t mds);
 280   bool caps_issued_mask(unsigned mask);
 281   int caps_used();
 282   int caps_file_wanted();
 283   int caps_wanted();
 284   int caps_mds_wanted();
 285   int caps_dirty();
 286   const UserPerm *get_best_perms();
 287
 288   bool have_valid_size();
 289   Dir *open_dir();
 290
 291   void add_fh(Fh *f) {fhs.insert(f);}
 292   void rm_fh(Fh *f) {fhs.erase(f);}
 293   void set_async_err(int r);
 294   void dump(Formatter *f) const;
 295 };
 296
 297 ostream& operator<<(ostream &out, const Inode &in);
 298
 299 #endif