1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 2
3 ============================================== 3 ===============================================================
4 Inotify - A Powerful yet Simple File Change No 4 Inotify - A Powerful yet Simple File Change Notification System
5 ============================================== 5 ===============================================================
6 6
7 7
8 8
9 Document started 15 Mar 2005 by Robert Love <rm 9 Document started 15 Mar 2005 by Robert Love <rml@novell.com>
10 10
11 Document updated 4 Jan 2015 by Zhang Zhen <zhen 11 Document updated 4 Jan 2015 by Zhang Zhen <zhenzhang.zhang@huawei.com>
12 12
13 - Deleted obsoleted interface, just re 13 - Deleted obsoleted interface, just refer to manpages for user interface.
14 14
15 (i) Rationale 15 (i) Rationale
16 16
17 Q: 17 Q:
18 What is the design decision behind not tyin 18 What is the design decision behind not tying the watch to the open fd of
19 the watched object? 19 the watched object?
20 20
21 A: 21 A:
22 Watches are associated with an open inotify 22 Watches are associated with an open inotify device, not an open file.
23 This solves the primary problem with dnotif 23 This solves the primary problem with dnotify: keeping the file open pins
24 the file and thus, worse, pins the mount. 24 the file and thus, worse, pins the mount. Dnotify is therefore infeasible
25 for use on a desktop system with removable 25 for use on a desktop system with removable media as the media cannot be
26 unmounted. Watching a file should not requ 26 unmounted. Watching a file should not require that it be open.
27 27
28 Q: 28 Q:
29 What is the design decision behind using an 29 What is the design decision behind using an-fd-per-instance as opposed to
30 an fd-per-watch? 30 an fd-per-watch?
31 31
32 A: 32 A:
33 An fd-per-watch quickly consumes more file 33 An fd-per-watch quickly consumes more file descriptors than are allowed,
34 more fd's than are feasible to manage, and 34 more fd's than are feasible to manage, and more fd's than are optimally
35 select()-able. Yes, root can bump the per- 35 select()-able. Yes, root can bump the per-process fd limit and yes, users
36 can use epoll, but requiring both is a sill 36 can use epoll, but requiring both is a silly and extraneous requirement.
37 A watch consumes less memory than an open f 37 A watch consumes less memory than an open file, separating the number
38 spaces is thus sensible. The current desig 38 spaces is thus sensible. The current design is what user-space developers
39 want: Users initialize inotify, once, and a 39 want: Users initialize inotify, once, and add n watches, requiring but one
40 fd and no twiddling with fd limits. Initia 40 fd and no twiddling with fd limits. Initializing an inotify instance two
41 thousand times is silly. If we can impleme 41 thousand times is silly. If we can implement user-space's preferences
42 cleanly--and we can, the idr layer makes st 42 cleanly--and we can, the idr layer makes stuff like this trivial--then we
43 should. 43 should.
44 44
45 There are other good arguments. With a sin 45 There are other good arguments. With a single fd, there is a single
46 item to block on, which is mapped to a sing 46 item to block on, which is mapped to a single queue of events. The single
47 fd returns all watch events and also any po 47 fd returns all watch events and also any potential out-of-band data. If
48 every fd was a separate watch, 48 every fd was a separate watch,
49 49
50 - There would be no way to get event orderi 50 - There would be no way to get event ordering. Events on file foo and
51 file bar would pop poll() on both fd's, b 51 file bar would pop poll() on both fd's, but there would be no way to tell
52 which happened first. A single queue tri 52 which happened first. A single queue trivially gives you ordering. Such
53 ordering is crucial to existing applicati 53 ordering is crucial to existing applications such as Beagle. Imagine
54 "mv a b ; mv b a" events without ordering 54 "mv a b ; mv b a" events without ordering.
55 55
56 - We'd have to maintain n fd's and n intern 56 - We'd have to maintain n fd's and n internal queues with state,
57 versus just one. It is a lot messier in 57 versus just one. It is a lot messier in the kernel. A single, linear
58 queue is the data structure that makes se 58 queue is the data structure that makes sense.
59 59
60 - User-space developers prefer the current 60 - User-space developers prefer the current API. The Beagle guys, for
61 example, love it. Trust me, I asked. It 61 example, love it. Trust me, I asked. It is not a surprise: Who'd want
62 to manage and block on 1000 fd's via sele 62 to manage and block on 1000 fd's via select?
63 63
64 - No way to get out of band data. 64 - No way to get out of band data.
65 65
66 - 1024 is still too low. ;-) 66 - 1024 is still too low. ;-)
67 67
68 When you talk about designing a file change 68 When you talk about designing a file change notification system that
69 scales to 1000s of directories, juggling 10 69 scales to 1000s of directories, juggling 1000s of fd's just does not seem
70 the right interface. It is too heavy. 70 the right interface. It is too heavy.
71 71
72 Additionally, it _is_ possible to more tha 72 Additionally, it _is_ possible to more than one instance and
73 juggle more than one queue and thus more th 73 juggle more than one queue and thus more than one associated fd. There
74 need not be a one-fd-per-process mapping; i 74 need not be a one-fd-per-process mapping; it is one-fd-per-queue and a
75 process can easily want more than one queue 75 process can easily want more than one queue.
76 76
77 Q: 77 Q:
78 Why the system call approach? 78 Why the system call approach?
79 79
80 A: 80 A:
81 The poor user-space interface is the second 81 The poor user-space interface is the second biggest problem with dnotify.
82 Signals are a terrible, terrible interface 82 Signals are a terrible, terrible interface for file notification. Or for
83 anything, for that matter. The ideal solut 83 anything, for that matter. The ideal solution, from all perspectives, is a
84 file descriptor-based one that allows basic 84 file descriptor-based one that allows basic file I/O and poll/select.
85 Obtaining the fd and managing the watches c 85 Obtaining the fd and managing the watches could have been done either via a
86 device file or a family of new system calls 86 device file or a family of new system calls. We decided to implement a
87 family of system calls because that is the 87 family of system calls because that is the preferred approach for new kernel
88 interfaces. The only real difference was w 88 interfaces. The only real difference was whether we wanted to use open(2)
89 and ioctl(2) or a couple of new system call 89 and ioctl(2) or a couple of new system calls. System calls beat ioctls.
90 90
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.