Documentation/filesystems/inotify.txt

   1                                    inotify
   2             a powerful yet simple file change notification system
   3
   4
   5
   6 Document started 15 Mar 2005 by Robert Love <rml@novell.com>
   7
   8
   9 (i) User Interface
  10
  11 Inotify is controlled by a set of three system calls and normal file I/O on a
  12 returned file descriptor.
  13
  14 First step in using inotify is to initialise an inotify instance:
  15
  16         int fd = inotify_init ();
  17
  18 Each instance is associated with a unique, ordered queue.
  19
  20 Change events are managed by "watches".  A watch is an (object,mask) pair where
  21 the object is a file or directory and the mask is a bit mask of one or more
  22 inotify events that the application wishes to receive.  See <linux/inotify.h>
  23 for valid events.  A watch is referenced by a watch descriptor, or wd.
  24
  25 Watches are added via a path to the file.
  26
  27 Watches on a directory will return events on any files inside of the directory.
  28
  29 Adding a watch is simple:
  30
  31         int wd = inotify_add_watch (fd, path, mask);
  32
  33 Where "fd" is the return value from inotify_init(), path is the path to the
  34 object to watch, and mask is the watch mask (see <linux/inotify.h>).
  35
  36 You can update an existing watch in the same manner, by passing in a new mask.
  37
  38 An existing watch is removed via
  39
  40         int ret = inotify_rm_watch (fd, wd);
  41
  42 Events are provided in the form of an inotify_event structure that is read(2)
  43 from a given inotify instance.  The filename is of dynamic length and follows
  44 the struct. It is of size len.  The filename is padded with null bytes to
  45 ensure proper alignment.  This padding is reflected in len.
  46
  47 You can slurp multiple events by passing a large buffer, for example
  48
  49         size_t len = read (fd, buf, BUF_LEN);
  50
  51 Where "buf" is a pointer to an array of "inotify_event" structures at least
  52 BUF_LEN bytes in size.  The above example will return as many events as are
  53 available and fit in BUF_LEN.
  54
  55 Each inotify instance fd is also select()- and poll()-able.
  56
  57 You can find the size of the current event queue via the standard FIONREAD
  58 ioctl on the fd returned by inotify_init().
  59
  60 All watches are destroyed and cleaned up on close.
  61
  62
  63 (ii)
  64
  65 Prototypes:
  66
  67         int inotify_init (void);
  68         int inotify_add_watch (int fd, const char *path, __u32 mask);
  69         int inotify_rm_watch (int fd, __u32 mask);
  70
  71
  72 (iii) Kernel Interface
  73
  74 Inotify's kernel API consists a set of functions for managing watches and an
  75 event callback.
  76
  77 To use the kernel API, you must first initialize an inotify instance with a set
  78 of inotify_operations.  You are given an opaque inotify_handle, which you use
  79 for any further calls to inotify.
  80
  81     struct inotify_handle *ih = inotify_init(my_event_handler);
  82
  83 You must provide a function for processing events and a function for destroying
  84 the inotify watch.
  85
  86     void handle_event(struct inotify_watch *watch, u32 wd, u32 mask,
  87                       u32 cookie, const char *name, struct inode *inode)
  88
  89         watch - the pointer to the inotify_watch that triggered this call
  90         wd - the watch descriptor
  91         mask - describes the event that occurred
  92         cookie - an identifier for synchronizing events
  93         name - the dentry name for affected files in a directory-based event
  94         inode - the affected inode in a directory-based event
  95
  96     void destroy_watch(struct inotify_watch *watch)
  97
  98 You may add watches by providing a pre-allocated and initialized inotify_watch
  99 structure and specifying the inode to watch along with an inotify event mask.
 100 You must pin the inode during the call.  You will likely wish to embed the
 101 inotify_watch structure in a structure of your own which contains other
 102 information about the watch.  Once you add an inotify watch, it is immediately
 103 subject to removal depending on filesystem events.  You must grab a reference if
 104 you depend on the watch hanging around after the call.
 105
 106     inotify_init_watch(&my_watch->iwatch);
 107     inotify_get_watch(&my_watch->iwatch);       // optional
 108     s32 wd = inotify_add_watch(ih, &my_watch->iwatch, inode, mask);
 109     inotify_put_watch(&my_watch->iwatch);       // optional
 110
 111 You may use the watch descriptor (wd) or the address of the inotify_watch for
 112 other inotify operations.  You must not directly read or manipulate data in the
 113 inotify_watch.  Additionally, you must not call inotify_add_watch() more than
 114 once for a given inotify_watch structure, unless you have first called either
 115 inotify_rm_watch() or inotify_rm_wd().
 116
 117 To determine if you have already registered a watch for a given inode, you may
 118 call inotify_find_watch(), which gives you both the wd and the watch pointer for
 119 the inotify_watch, or an error if the watch does not exist.
 120
 121     wd = inotify_find_watch(ih, inode, &watchp);
 122
 123 You may use container_of() on the watch pointer to access your own data
 124 associated with a given watch.  When an existing watch is found,
 125 inotify_find_watch() bumps the refcount before releasing its locks.  You must
 126 put that reference with:
 127
 128     put_inotify_watch(watchp);
 129
 130 Call inotify_find_update_watch() to update the event mask for an existing watch.
 131 inotify_find_update_watch() returns the wd of the updated watch, or an error if
 132 the watch does not exist.
 133
 134     wd = inotify_find_update_watch(ih, inode, mask);
 135
 136 An existing watch may be removed by calling either inotify_rm_watch() or
 137 inotify_rm_wd().
 138
 139     int ret = inotify_rm_watch(ih, &my_watch->iwatch);
 140     int ret = inotify_rm_wd(ih, wd);
 141
 142 A watch may be removed while executing your event handler with the following:
 143
 144     inotify_remove_watch_locked(ih, iwatch);
 145
 146 Call inotify_destroy() to remove all watches from your inotify instance and
 147 release it.  If there are no outstanding references, inotify_destroy() will call
 148 your destroy_watch op for each watch.
 149
 150     inotify_destroy(ih);
 151
 152 When inotify removes a watch, it sends an IN_IGNORED event to your callback.
 153 You may use this event as an indication to free the watch memory.  Note that
 154 inotify may remove a watch due to filesystem events, as well as by your request.
 155 If you use IN_ONESHOT, inotify will remove the watch after the first event, at
 156 which point you may call the final inotify_put_watch.
 157
 158 (iv) Kernel Interface Prototypes
 159
 160         struct inotify_handle *inotify_init(struct inotify_operations *ops);
 161
 162         inotify_init_watch(struct inotify_watch *watch);
 163
 164         s32 inotify_add_watch(struct inotify_handle *ih,
 165                               struct inotify_watch *watch,
 166                               struct inode *inode, u32 mask);
 167
 168         s32 inotify_find_watch(struct inotify_handle *ih, struct inode *inode,
 169                                struct inotify_watch **watchp);
 170
 171         s32 inotify_find_update_watch(struct inotify_handle *ih,
 172                                       struct inode *inode, u32 mask);
 173
 174         int inotify_rm_wd(struct inotify_handle *ih, u32 wd);
 175
 176         int inotify_rm_watch(struct inotify_handle *ih,
 177                              struct inotify_watch *watch);
 178
 179         void inotify_remove_watch_locked(struct inotify_handle *ih,
 180                                          struct inotify_watch *watch);
 181
 182         void inotify_destroy(struct inotify_handle *ih);
 183
 184         void get_inotify_watch(struct inotify_watch *watch);
 185         void put_inotify_watch(struct inotify_watch *watch);
 186
 187
 188 (v) Internal Kernel Implementation
 189
 190 Each inotify instance is represented by an inotify_handle structure.
 191 Inotify's userspace consumers also have an inotify_device which is
 192 associated with the inotify_handle, and on which events are queued.
 193
 194 Each watch is associated with an inotify_watch structure.  Watches are chained
 195 off of each associated inotify_handle and each associated inode.
 196
 197 See fs/notify/inotify/inotify_fsnotify.c and fs/notify/inotify/inotify_user.c
 198 for the locking and lifetime rules.
 199
 200
 201 (vi) Rationale
 202
 203 Q: What is the design decision behind not tying the watch to the open fd of
 204    the watched object?
 205
 206 A: Watches are associated with an open inotify device, not an open file.
 207    This solves the primary problem with dnotify: keeping the file open pins
 208    the file and thus, worse, pins the mount.  Dnotify is therefore infeasible
 209    for use on a desktop system with removable media as the media cannot be
 210    unmounted.  Watching a file should not require that it be open.
 211
 212 Q: What is the design decision behind using an-fd-per-instance as opposed to
 213    an fd-per-watch?
 214
 215 A: An fd-per-watch quickly consumes more file descriptors than are allowed,
 216    more fd's than are feasible to manage, and more fd's than are optimally
 217    select()-able.  Yes, root can bump the per-process fd limit and yes, users
 218    can use epoll, but requiring both is a silly and extraneous requirement.
 219    A watch consumes less memory than an open file, separating the number
 220    spaces is thus sensible.  The current design is what user-space developers
 221    want: Users initialize inotify, once, and add n watches, requiring but one
 222    fd and no twiddling with fd limits.  Initializing an inotify instance two
 223    thousand times is silly.  If we can implement user-space's preferences
 224    cleanly--and we can, the idr layer makes stuff like this trivial--then we
 225    should.
 226
 227    There are other good arguments.  With a single fd, there is a single
 228    item to block on, which is mapped to a single queue of events.  The single
 229    fd returns all watch events and also any potential out-of-band data.  If
 230    every fd was a separate watch,
 231
 232    - There would be no way to get event ordering.  Events on file foo and
 233      file bar would pop poll() on both fd's, but there would be no way to tell
 234      which happened first.  A single queue trivially gives you ordering.  Such
 235      ordering is crucial to existing applications such as Beagle.  Imagine
 236      "mv a b ; mv b a" events without ordering.
 237
 238    - We'd have to maintain n fd's and n internal queues with state,
 239      versus just one.  It is a lot messier in the kernel.  A single, linear
 240      queue is the data structure that makes sense.
 241
 242    - User-space developers prefer the current API.  The Beagle guys, for
 243      example, love it.  Trust me, I asked.  It is not a surprise: Who'd want
 244      to manage and block on 1000 fd's via select?
 245
 246    - No way to get out of band data.
 247
 248    - 1024 is still too low.  ;-)
 249
 250    When you talk about designing a file change notification system that
 251    scales to 1000s of directories, juggling 1000s of fd's just does not seem
 252    the right interface.  It is too heavy.
 253
 254    Additionally, it _is_ possible to  more than one instance  and
 255    juggle more than one queue and thus more than one associated fd.  There
 256    need not be a one-fd-per-process mapping; it is one-fd-per-queue and a
 257    process can easily want more than one queue.
 258
 259 Q: Why the system call approach?
 260
 261 A: The poor user-space interface is the second biggest problem with dnotify.
 262    Signals are a terrible, terrible interface for file notification.  Or for
 263    anything, for that matter.  The ideal solution, from all perspectives, is a
 264    file descriptor-based one that allows basic file I/O and poll/select.
 265    Obtaining the fd and managing the watches could have been done either via a
 266    device file or a family of new system calls.  We decided to implement a
 267    family of system calls because that is the preferred approach for new kernel
 268    interfaces.  The only real difference was whether we wanted to use open(2)
 269    and ioctl(2) or a couple of new system calls.  System calls beat ioctls.
 270