1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
|
.\" $OpenBSD: mdoc.template,v 1.9 2004/07/02 10:36:57 jmc Exp $
.\"
.Dd July 24, 2007
.Dt pnotify 3
.Os
.Sh NAME
.Nm pnotify
.Nd monitor filesystem events
.Sh SYNOPSIS
.In pnotify.h
.Pp
.Fd int pnotify_init(struct pnotify_cb *ctl);
.Fd int pnotify_add_watch(struct pnotify_cb *ctl, const char *pathname, int mask);
.Fd int pnotify_rm_watch(struct pnotify_cb *ctl, int wd);
.Fd int pnotify_get_event(struct pnotify_event *evt, struct pnotify_cb *ctl);
.Fd int pnotify_free(struct pnotify_cb *ctl);
.Pp
.Bd -literal
struct pnotify_event {
int wd;
int mask;
char name[NAME_MAX + 1];
};
.Ed
.Sh DESCRIPTION
The
.Nm
library provides a portable way to monitor filesystem events. It closely resembles
the inotify(7) kernel subsystem in Linux, but is implemented as a userspace library
to support other kernel mechanisms such as kqueue(4).
.Pp
.Fn pnotify_init
initializes a pnotify control block. This structure contains the current state of events and watches, and must be initialized before it can be used.
.Pp
.Fn pnotify_add_watch
adds a watch for a given file or directory. The mask parameter is composed of one
or more bitflags from the following list of events:
.Bl -column "Flag" "Meaning" -offset indent
.It Sy PN_ACCESS Ta "The access time (atime) has been modified."
.It Sy PN_CREATE Ta "A file was created in the directory."
.It Sy PN_DELETE Ta "The file was deleted."
.It Sy PN_MODIFY Ta "The modification time (mtime) has been modified."
.El
.Pp
When calling
.Fn pnotify_add_watch
the following flags may also be used:
.Bl -column "Flag" "Meaning" -offset indent
.It Sy PN_ALL_EVENTS Ta "Monitor all events."
.It Sy PN_ONESHOT Ta "Delete the watch after an event occurs."
.El
.Fn pnotify_rm_watch
removes a watch from the watchlist.
.Fn pnotify_get_event
blocks until an event occurs, and fills the pnotify_event structure with information
about the event. If an error occurs in the underlying kernel event queue, an
event is returned with the PN_ERROR flag set.
.Fn pnotify_free
frees all of the memory associated with a control block and closes the kernel event
descriptor.
.Sh RETURN VALUES
.Fn pnotify_add_watch
returns a positive integer that represents a unique watch descriptor, or -1 if an error occurs.
.Pp
All other functions return zero if successful, or -1 if an error occurs.
.Sh EXAMPLES
The following example creates a watch on the /tmp directory and prints the
names of files that are created or deleted within the directory.
.Bd -literal
struct pnotify_cb ctl;
struct pnotify_event evt;
pnotify_init(&ctl);
pnotify_add_watch(&ctl, "/tmp", PN_ALL_EVENTS);
while (pnotify_get_event(&evt, &ctl, NULL) == 0) {
if (evt.mask & PN_CREATE)
printf("file created: %s", evt.name);
if (evt.mask & PN_DELETE)
printf("file deleted: %s", evt.name);
if (evt.mask & PN_ERROR) {
printf("an error occurred");
break;
}
}
pnotify_free(&ctl);
.Ed
.Sh THREADSAFETY
All
.Nm
functions are fully reentrant and may be used in multithreaded programs.
A single
.Nm
control block (struct pnotify_cb) should not be shared by multiple threads
unless it is protected by a mutex.
.Sh SEE ALSO
.Xr kqueue 4
.Xr inotify 7
.\" .Sh STANDARDS
.Sh HISTORY
pnotify was first released on July 25th, 2007 with support for inotify and kqueue.
.Sh AUTHORS
Mark Heily <devel@heily.com>
.\" .Sh CAVEATS
.\" .Sh BUGS
.\"kqueue(4) support is buggy and only works properly with PN_CREATE events.
|
