1Documentation for /proc/sys/fs/*	kernel version 2.2.10
2	(c) 1998, 1999,  Rik van Riel <riel@nl.linux.org>
3	(c) 2009,        Shen Feng<shen@cn.fujitsu.com>
4
5For general info and legal blurb, please look in README.
6
7==============================================================
8
9This file contains documentation for the sysctl files in
10/proc/sys/fs/ and is valid for Linux kernel version 2.2.
11
12The files in this directory can be used to tune and monitor
13miscellaneous and general things in the operation of the Linux
14kernel. Since some of the files _can_ be used to screw up your
15system, it is advisable to read both documentation and source
16before actually making adjustments.
17
181. /proc/sys/fs
19----------------------------------------------------------
20
21Currently, these files are in /proc/sys/fs:
22- aio-max-nr
23- aio-nr
24- dentry-state
25- dquot-max
26- dquot-nr
27- file-max
28- file-nr
29- inode-max
30- inode-nr
31- inode-state
32- nr_open
33- overflowuid
34- overflowgid
35- protected_hardlinks
36- protected_symlinks
37- suid_dumpable
38- super-max
39- super-nr
40
41==============================================================
42
43aio-nr & aio-max-nr:
44
45aio-nr is the running total of the number of events specified on the
46io_setup system call for all currently active aio contexts.  If aio-nr
47reaches aio-max-nr then io_setup will fail with EAGAIN.  Note that
48raising aio-max-nr does not result in the pre-allocation or re-sizing
49of any kernel data structures.
50
51==============================================================
52
53dentry-state:
54
55From linux/fs/dentry.c:
56--------------------------------------------------------------
57struct {
58        int nr_dentry;
59        int nr_unused;
60        int age_limit;         /* age in seconds */
61        int want_pages;        /* pages requested by system */
62        int dummy[2];
63} dentry_stat = {0, 0, 45, 0,};
64--------------------------------------------------------------
65
66Dentries are dynamically allocated and deallocated, and
67nr_dentry seems to be 0 all the time. Hence it's safe to
68assume that only nr_unused, age_limit and want_pages are
69used. Nr_unused seems to be exactly what its name says.
70Age_limit is the age in seconds after which dcache entries
71can be reclaimed when memory is short and want_pages is
72nonzero when shrink_dcache_pages() has been called and the
73dcache isn't pruned yet.
74
75==============================================================
76
77dquot-max & dquot-nr:
78
79The file dquot-max shows the maximum number of cached disk
80quota entries.
81
82The file dquot-nr shows the number of allocated disk quota
83entries and the number of free disk quota entries.
84
85If the number of free cached disk quotas is very low and
86you have some awesome number of simultaneous system users,
87you might want to raise the limit.
88
89==============================================================
90
91file-max & file-nr:
92
93The value in file-max denotes the maximum number of file-
94handles that the Linux kernel will allocate. When you get lots
95of error messages about running out of file handles, you might
96want to increase this limit.
97
98Historically,the kernel was able to allocate file handles
99dynamically, but not to free them again. The three values in
100file-nr denote the number of allocated file handles, the number
101of allocated but unused file handles, and the maximum number of
102file handles. Linux 2.6 always reports 0 as the number of free
103file handles -- this is not an error, it just means that the
104number of allocated file handles exactly matches the number of
105used file handles.
106
107Attempts to allocate more file descriptors than file-max are
108reported with printk, look for "VFS: file-max limit <number>
109reached".
110==============================================================
111
112nr_open:
113
114This denotes the maximum number of file-handles a process can
115allocate. Default value is 1024*1024 (1048576) which should be
116enough for most machines. Actual limit depends on RLIMIT_NOFILE
117resource limit.
118
119==============================================================
120
121inode-max, inode-nr & inode-state:
122
123As with file handles, the kernel allocates the inode structures
124dynamically, but can't free them yet.
125
126The value in inode-max denotes the maximum number of inode
127handlers. This value should be 3-4 times larger than the value
128in file-max, since stdin, stdout and network sockets also
129need an inode struct to handle them. When you regularly run
130out of inodes, you need to increase this value.
131
132The file inode-nr contains the first two items from
133inode-state, so we'll skip to that file...
134
135Inode-state contains three actual numbers and four dummies.
136The actual numbers are, in order of appearance, nr_inodes,
137nr_free_inodes and preshrink.
138
139Nr_inodes stands for the number of inodes the system has
140allocated, this can be slightly more than inode-max because
141Linux allocates them one pageful at a time.
142
143Nr_free_inodes represents the number of free inodes (?) and
144preshrink is nonzero when the nr_inodes > inode-max and the
145system needs to prune the inode list instead of allocating
146more.
147
148==============================================================
149
150overflowgid & overflowuid:
151
152Some filesystems only support 16-bit UIDs and GIDs, although in Linux
153UIDs and GIDs are 32 bits. When one of these filesystems is mounted
154with writes enabled, any UID or GID that would exceed 65535 is translated
155to a fixed value before being written to disk.
156
157These sysctls allow you to change the value of the fixed UID and GID.
158The default is 65534.
159
160==============================================================
161
162protected_hardlinks:
163
164A long-standing class of security issues is the hardlink-based
165time-of-check-time-of-use race, most commonly seen in world-writable
166directories like /tmp. The common method of exploitation of this flaw
167is to cross privilege boundaries when following a given hardlink (i.e. a
168root process follows a hardlink created by another user). Additionally,
169on systems without separated partitions, this stops unauthorized users
170from "pinning" vulnerable setuid/setgid files against being upgraded by
171the administrator, or linking to special files.
172
173When set to "0", hardlink creation behavior is unrestricted.
174
175When set to "1" hardlinks cannot be created by users if they do not
176already own the source file, or do not have read/write access to it.
177
178This protection is based on the restrictions in Openwall and grsecurity.
179
180==============================================================
181
182protected_symlinks:
183
184A long-standing class of security issues is the symlink-based
185time-of-check-time-of-use race, most commonly seen in world-writable
186directories like /tmp. The common method of exploitation of this flaw
187is to cross privilege boundaries when following a given symlink (i.e. a
188root process follows a symlink belonging to another user). For a likely
189incomplete list of hundreds of examples across the years, please see:
190http://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=/tmp
191
192When set to "0", symlink following behavior is unrestricted.
193
194When set to "1" symlinks are permitted to be followed only when outside
195a sticky world-writable directory, or when the uid of the symlink and
196follower match, or when the directory owner matches the symlink's owner.
197
198This protection is based on the restrictions in Openwall and grsecurity.
199
200==============================================================
201
202suid_dumpable:
203
204This value can be used to query and set the core dump mode for setuid
205or otherwise protected/tainted binaries. The modes are
206
2070 - (default) - traditional behaviour. Any process which has changed
208	privilege levels or is execute only will not be dumped.
2091 - (debug) - all processes dump core when possible. The core dump is
210	owned by the current user and no security is applied. This is
211	intended for system debugging situations only. Ptrace is unchecked.
212	This is insecure as it allows regular users to examine the memory
213	contents of privileged processes.
2142 - (suidsafe) - any binary which normally would not be dumped is dumped
215	anyway, but only if the "core_pattern" kernel sysctl is set to
216	either a pipe handler or a fully qualified path. (For more details
217	on this limitation, see CVE-2006-2451.) This mode is appropriate
218	when administrators are attempting to debug problems in a normal
219	environment, and either have a core dump pipe handler that knows
220	to treat privileged core dumps with care, or specific directory
221	defined for catching core dumps. If a core dump happens without
222	a pipe handler or fully qualifid path, a message will be emitted
223	to syslog warning about the lack of a correct setting.
224
225==============================================================
226
227super-max & super-nr:
228
229These numbers control the maximum number of superblocks, and
230thus the maximum number of mounted filesystems the kernel
231can have. You only need to increase super-max if you need to
232mount more filesystems than the current value in super-max
233allows you to.
234
235==============================================================
236
237aio-nr & aio-max-nr:
238
239aio-nr shows the current system-wide number of asynchronous io
240requests.  aio-max-nr allows you to change the maximum value
241aio-nr can grow to.
242
243==============================================================
244
245
2462. /proc/sys/fs/binfmt_misc
247----------------------------------------------------------
248
249Documentation for the files in /proc/sys/fs/binfmt_misc is
250in Documentation/binfmt_misc.txt.
251
252
2533. /proc/sys/fs/mqueue - POSIX message queues filesystem
254----------------------------------------------------------
255
256The "mqueue"  filesystem provides  the necessary kernel features to enable the
257creation of a  user space  library that  implements  the  POSIX message queues
258API (as noted by the  MSG tag in the  POSIX 1003.1-2001 version  of the System
259Interfaces specification.)
260
261The "mqueue" filesystem contains values for determining/setting  the amount of
262resources used by the file system.
263
264/proc/sys/fs/mqueue/queues_max is a read/write  file for  setting/getting  the
265maximum number of message queues allowed on the system.
266
267/proc/sys/fs/mqueue/msg_max  is  a  read/write file  for  setting/getting  the
268maximum number of messages in a queue value.  In fact it is the limiting value
269for another (user) limit which is set in mq_open invocation. This attribute of
270a queue must be less or equal then msg_max.
271
272/proc/sys/fs/mqueue/msgsize_max is  a read/write  file for setting/getting the
273maximum  message size value (it is every  message queue's attribute set during
274its creation).
275
276/proc/sys/fs/mqueue/msg_default is  a read/write  file for setting/getting the
277default number of messages in a queue value if attr parameter of mq_open(2) is
278NULL. If it exceed msg_max, the default value is initialized msg_max.
279
280/proc/sys/fs/mqueue/msgsize_default is a read/write file for setting/getting
281the default message size value if attr parameter of mq_open(2) is NULL. If it
282exceed msgsize_max, the default value is initialized msgsize_max.
283
2844. /proc/sys/fs/epoll - Configuration options for the epoll interface
285--------------------------------------------------------
286
287This directory contains configuration options for the epoll(7) interface.
288
289max_user_watches
290----------------
291
292Every epoll file descriptor can store a number of files to be monitored
293for event readiness. Each one of these monitored files constitutes a "watch".
294This configuration option sets the maximum number of "watches" that are
295allowed for each user.
296Each "watch" costs roughly 90 bytes on a 32bit kernel, and roughly 160 bytes
297on a 64bit one.
298The current default value for  max_user_watches  is the 1/32 of the available
299low memory, divided for the "watch" cost in bytes.
300
301