1
2FANOTIFY_INIT(2) Linux Programmer's Manual FANOTIFY_INIT(2)
3
4
5
7 fanotify_init - create and initialize fanotify group
8
10 #include <fcntl.h>
11 #include <sys/fanotify.h>
12
13 int fanotify_init(unsigned int flags, unsigned int event_f_flags);
14
16 For an overview of the fanotify API, see fanotify(7).
17
18 fanotify_init() initializes a new fanotify group and returns a file de‐
19 scriptor for the event queue associated with the group.
20
21 The file descriptor is used in calls to fanotify_mark(2) to specify the
22 files, directories, mounts or filesystems for which fanotify events
23 shall be created. These events are received by reading from the file
24 descriptor. Some events are only informative, indicating that a file
25 has been accessed. Other events can be used to determine whether an‐
26 other application is permitted to access a file or directory. Permis‐
27 sion to access filesystem objects is granted by writing to the file de‐
28 scriptor.
29
30 Multiple programs may be using the fanotify interface at the same time
31 to monitor the same files.
32
33 In the current implementation, the number of fanotify groups per user
34 is limited to 128. This limit cannot be overridden.
35
36 Calling fanotify_init() requires the CAP_SYS_ADMIN capability. This
37 constraint might be relaxed in future versions of the API. Therefore,
38 certain additional capability checks have been implemented as indicated
39 below.
40
41 The flags argument contains a multi-bit field defining the notification
42 class of the listening application and further single bit fields speci‐
43 fying the behavior of the file descriptor.
44
45 If multiple listeners for permission events exist, the notification
46 class is used to establish the sequence in which the listeners receive
47 the events.
48
49 Only one of the following notification classes may be specified in
50 flags:
51
52 FAN_CLASS_PRE_CONTENT
53 This value allows the receipt of events notifying that a file
54 has been accessed and events for permission decisions if a file
55 may be accessed. It is intended for event listeners that need
56 to access files before they contain their final data. This no‐
57 tification class might be used by hierarchical storage managers,
58 for example.
59
60 FAN_CLASS_CONTENT
61 This value allows the receipt of events notifying that a file
62 has been accessed and events for permission decisions if a file
63 may be accessed. It is intended for event listeners that need
64 to access files when they already contain their final content.
65 This notification class might be used by malware detection pro‐
66 grams, for example.
67
68 FAN_CLASS_NOTIF
69 This is the default value. It does not need to be specified.
70 This value only allows the receipt of events notifying that a
71 file has been accessed. Permission decisions before the file is
72 accessed are not possible.
73
74 Listeners with different notification classes will receive events in
75 the order FAN_CLASS_PRE_CONTENT, FAN_CLASS_CONTENT, FAN_CLASS_NOTIF.
76 The order of notification for listeners in the same notification class
77 is undefined.
78
79 The following bits can additionally be set in flags:
80
81 FAN_CLOEXEC
82 Set the close-on-exec flag (FD_CLOEXEC) on the new file descrip‐
83 tor. See the description of the O_CLOEXEC flag in open(2).
84
85 FAN_NONBLOCK
86 Enable the nonblocking flag (O_NONBLOCK) for the file descrip‐
87 tor. Reading from the file descriptor will not block. Instead,
88 if no data is available, read(2) fails with the error EAGAIN.
89
90 FAN_UNLIMITED_QUEUE
91 Remove the limit of 16384 events for the event queue. Use of
92 this flag requires the CAP_SYS_ADMIN capability.
93
94 FAN_UNLIMITED_MARKS
95 Remove the limit of 8192 marks. Use of this flag requires the
96 CAP_SYS_ADMIN capability.
97
98 FAN_REPORT_TID (since Linux 4.20)
99 Report thread ID (TID) instead of process ID (PID) in the pid
100 field of the struct fanotify_event_metadata supplied to read(2)
101 (see fanotify(7)).
102
103 FAN_REPORT_FID (since Linux 5.1)
104 This value allows the receipt of events which contain additional
105 information about the underlying filesystem object correlated to
106 an event. An additional record of type FAN_EVENT_INFO_TYPE_FID
107 encapsulates the information about the object and is included
108 alongside the generic event metadata structure. The file de‐
109 scriptor that is used to represent the object correlated to an
110 event is instead substituted with a file handle. It is intended
111 for applications that may find the use of a file handle to iden‐
112 tify an object more suitable than a file descriptor. Addition‐
113 ally, it may be used for applications monitoring a directory or
114 a filesystem that are interested in the directory entry modifi‐
115 cation events FAN_CREATE, FAN_DELETE, and FAN_MOVE, or in events
116 such as FAN_ATTRIB, FAN_DELETE_SELF, and FAN_MOVE_SELF. All the
117 events above require an fanotify group that identifies filesys‐
118 tem objects by file handles. Note that for the directory entry
119 modification events the reported file handle identifies the mod‐
120 ified directory and not the created/deleted/moved child object.
121 The use of FAN_CLASS_CONTENT or FAN_CLASS_PRE_CONTENT is not
122 permitted with this flag and will result in the error EINVAL.
123 See fanotify(7) for additional details.
124
125 FAN_REPORT_DIR_FID (since Linux 5.9)
126 Events for fanotify groups initialized with this flag will con‐
127 tain (see exceptions below) additional information about a di‐
128 rectory object correlated to an event. An additional record of
129 type FAN_EVENT_INFO_TYPE_DFID encapsulates the information about
130 the directory object and is included alongside the generic event
131 metadata structure. For events that occur on a non-directory
132 object, the additional structure includes a file handle that
133 identifies the parent directory filesystem object. Note that
134 there is no guarantee that the directory filesystem object will
135 be found at the location described by the file handle informa‐
136 tion at the time the event is received. When combined with the
137 flag FAN_REPORT_FID, two records may be reported with events
138 that occur on a non-directory object, one to identify the non-
139 directory object itself and one to identify the parent directory
140 object. Note that in some cases, a filesystem object does not
141 have a parent, for example, when an event occurs on an unlinked
142 but open file. In that case, with the FAN_REPORT_FID flag, the
143 event will be reported with only one record to identify the non-
144 directory object itself, because there is no directory associ‐
145 ated with the event. Without the FAN_REPORT_FID flag, no event
146 will be reported. See fanotify(7) for additional details.
147
148 FAN_REPORT_NAME (since Linux 5.9)
149 Events for fanotify groups initialized with this flag will con‐
150 tain additional information about the name of the directory en‐
151 try correlated to an event. This flag must be provided in con‐
152 junction with the flag FAN_REPORT_DIR_FID. Providing this flag
153 value without FAN_REPORT_DIR_FID will result in the error EIN‐
154 VAL. This flag may be combined with the flag FAN_REPORT_FID.
155 An additional record of type FAN_EVENT_INFO_TYPE_DFID_NAME,
156 which encapsulates the information about the directory entry, is
157 included alongside the generic event metadata structure and sub‐
158 stitutes the additional information record of type
159 FAN_EVENT_INFO_TYPE_DFID. The additional record includes a file
160 handle that identifies a directory filesystem object followed by
161 a name that identifies an entry in that directory. For the di‐
162 rectory entry modification events FAN_CREATE, FAN_DELETE, and
163 FAN_MOVE, the reported name is that of the created/deleted/moved
164 directory entry. For other events that occur on a directory ob‐
165 ject, the reported file handle is that of the directory object
166 itself and the reported name is '.'. For other events that oc‐
167 cur on a non-directory object, the reported file handle is that
168 of the parent directory object and the reported name is the name
169 of a directory entry where the object was located at the time of
170 the event. The rationale behind this logic is that the reported
171 directory file handle can be passed to open_by_handle_at(2) to
172 get an open directory file descriptor and that file descriptor
173 along with the reported name can be used to call fstatat(2).
174 The same rule that applies to record type
175 FAN_EVENT_INFO_TYPE_DFID also applies to record type
176 FAN_EVENT_INFO_TYPE_DFID_NAME: if a non-directory object has no
177 parent, either the event will not be reported or it will be re‐
178 ported without the directory entry information. Note that there
179 is no guarantee that the filesystem object will be found at the
180 location described by the directory entry information at the
181 time the event is received. See fanotify(7) for additional de‐
182 tails.
183
184 FAN_REPORT_DFID_NAME
185 This is a synonym for (FAN_REPORT_DIR_FID|FAN_REPORT_NAME).
186
187 The event_f_flags argument defines the file status flags that will be
188 set on the open file descriptions that are created for fanotify events.
189 For details of these flags, see the description of the flags values in
190 open(2). event_f_flags includes a multi-bit field for the access mode.
191 This field can take the following values:
192
193 O_RDONLY
194 This value allows only read access.
195
196 O_WRONLY
197 This value allows only write access.
198
199 O_RDWR This value allows read and write access.
200
201 Additional bits can be set in event_f_flags. The most useful values
202 are:
203
204 O_LARGEFILE
205 Enable support for files exceeding 2 GB. Failing to set this
206 flag will result in an EOVERFLOW error when trying to open a
207 large file which is monitored by an fanotify group on a 32-bit
208 system.
209
210 O_CLOEXEC (since Linux 3.18)
211 Enable the close-on-exec flag for the file descriptor. See the
212 description of the O_CLOEXEC flag in open(2) for reasons why
213 this may be useful.
214
215 The following are also allowable: O_APPEND, O_DSYNC, O_NOATIME, O_NON‐
216 BLOCK, and O_SYNC. Specifying any other flag in event_f_flags yields
217 the error EINVAL (but see BUGS).
218
220 On success, fanotify_init() returns a new file descriptor. On error,
221 -1 is returned, and errno is set to indicate the error.
222
224 EINVAL An invalid value was passed in flags or event_f_flags.
225 FAN_ALL_INIT_FLAGS (deprecated since Linux kernel version 4.20)
226 defines all allowable bits for flags.
227
228 EMFILE The number of fanotify groups for this user exceeds 128.
229
230 EMFILE The per-process limit on the number of open file descriptors has
231 been reached.
232
233 ENOMEM The allocation of memory for the notification group failed.
234
235 ENOSYS This kernel does not implement fanotify_init(). The fanotify
236 API is available only if the kernel was configured with CON‐
237 FIG_FANOTIFY.
238
239 EPERM The operation is not permitted because the caller lacks the
240 CAP_SYS_ADMIN capability.
241
243 fanotify_init() was introduced in version 2.6.36 of the Linux kernel
244 and enabled in version 2.6.37.
245
247 This system call is Linux-specific.
248
250 The following bug was present in Linux kernels before version 3.18:
251
252 * The O_CLOEXEC is ignored when passed in event_f_flags.
253
254 The following bug was present in Linux kernels before version 3.14:
255
256 * The event_f_flags argument is not checked for invalid flags. Flags
257 that are intended only for internal use, such as FMODE_EXEC, can be
258 set, and will consequently be set for the file descriptors returned
259 when reading from the fanotify file descriptor.
260
262 fanotify_mark(2), fanotify(7)
263
265 This page is part of release 5.10 of the Linux man-pages project. A
266 description of the project, information about reporting bugs, and the
267 latest version of this page, can be found at
268 https://www.kernel.org/doc/man-pages/.
269
270
271
272Linux 2020-11-01 FANOTIFY_INIT(2)