1 ===============================
2 Documentation for /proc/sys/fs/
3 ===============================
5 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
7 Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>
9 For general info and legal blurb, please look in intro.rst.
11 ------------------------------------------------------------------------------
13 This file contains documentation for the sysctl files and directories
16 The files in this directory can be used to tune and monitor
17 miscellaneous and general things in the operation of the Linux
18 kernel. Since some of the files *can* be used to screw up your
19 system, it is advisable to read both documentation and source
20 before actually making adjustments.
25 Currently, these files might (depending on your configuration)
26 show up in ``/proc/sys/fs``:
34 ``aio-nr`` shows the current system-wide number of asynchronous io
35 requests. ``aio-max-nr`` allows you to change the maximum value
36 ``aio-nr`` can grow to. If ``aio-nr`` reaches ``aio-nr-max`` then
37 ``io_setup`` will fail with ``EAGAIN``. Note that raising
38 ``aio-max-nr`` does not result in the
39 pre-allocation or re-sizing of any kernel data structures.
45 This file shows the values in ``struct dentry_stat``, as defined in
46 ``linux/include/linux/dcache.h``::
48 struct dentry_stat_t dentry_stat {
51 int age_limit; /* age in seconds */
52 int want_pages; /* pages requested by system */
53 int nr_negative; /* # of unused negative dentries */
54 int dummy; /* Reserved for future use */
57 Dentries are dynamically allocated and deallocated.
59 ``nr_dentry`` shows the total number of dentries allocated (active
60 + unused). ``nr_unused shows`` the number of dentries that are not
61 actively used, but are saved in the LRU list for future reuse.
63 ``age_limit`` is the age in seconds after which dcache entries
64 can be reclaimed when memory is short and ``want_pages`` is
65 nonzero when ``shrink_dcache_pages()`` has been called and the
66 dcache isn't pruned yet.
68 ``nr_negative`` shows the number of unused dentries that are also
69 negative dentries which do not map to any files. Instead,
70 they help speeding up rejection of non-existing files provided
77 The value in ``file-max`` denotes the maximum number of file-
78 handles that the Linux kernel will allocate. When you get lots
79 of error messages about running out of file handles, you might
80 want to increase this limit.
82 Historically,the kernel was able to allocate file handles
83 dynamically, but not to free them again. The three values in
84 ``file-nr`` denote the number of allocated file handles, the number
85 of allocated but unused file handles, and the maximum number of
86 file handles. Linux 2.6 and later always reports 0 as the number of free
87 file handles -- this is not an error, it just means that the
88 number of allocated file handles exactly matches the number of
91 Attempts to allocate more file descriptors than ``file-max`` are
92 reported with ``printk``, look for::
94 VFS: file-max limit <number> reached
99 inode-nr & inode-state
100 ----------------------
102 As with file handles, the kernel allocates the inode structures
103 dynamically, but can't free them yet.
105 The file ``inode-nr`` contains the first two items from
106 ``inode-state``, so we'll skip to that file...
108 ``inode-state`` contains three actual numbers and four dummies.
109 The actual numbers are, in order of appearance, ``nr_inodes``,
110 ``nr_free_inodes`` and ``preshrink``.
112 ``nr_inodes`` stands for the number of inodes the system has
115 ``nr_free_inodes`` represents the number of free inodes (?) and
116 preshrink is nonzero when the
117 system needs to prune the inode list instead of allocating
124 This denotes the maximum number of mounts that may exist
125 in a mount namespace.
131 This denotes the maximum number of file-handles a process can
132 allocate. Default value is 1024*1024 (1048576) which should be
133 enough for most machines. Actual limit depends on ``RLIMIT_NOFILE``
137 overflowgid & overflowuid
138 -------------------------
140 Some filesystems only support 16-bit UIDs and GIDs, although in Linux
141 UIDs and GIDs are 32 bits. When one of these filesystems is mounted
142 with writes enabled, any UID or GID that would exceed 65535 is translated
143 to a fixed value before being written to disk.
145 These sysctls allow you to change the value of the fixed UID and GID.
146 The default is 65534.
152 Maximum total number of pages a non-privileged user may allocate for pipes.
153 Once this limit is reached, no new pipes may be allocated until usage goes
154 below the limit again. When set to 0, no limit is applied, which is the default
161 Maximum total number of pages a non-privileged user may allocate for pipes
162 before the pipe size gets limited to a single page. Once this limit is reached,
163 new pipes will be limited to a single page in size for this user in order to
164 limit total memory usage, and trying to increase them using ``fcntl()`` will be
165 denied until usage goes below the limit again. The default value allows to
166 allocate up to 1024 pipes at their default size. When set to 0, no limit is
173 The intent of this protection is to avoid unintentional writes to
174 an attacker-controlled FIFO, where a program expected to create a regular
177 When set to "0", writing to FIFOs is unrestricted.
179 When set to "1" don't allow ``O_CREAT`` open on FIFOs that we don't own
180 in world writable sticky directories, unless they are owned by the
181 owner of the directory.
183 When set to "2" it also applies to group writable sticky directories.
185 This protection is based on the restrictions in Openwall.
191 A long-standing class of security issues is the hardlink-based
192 time-of-check-time-of-use race, most commonly seen in world-writable
193 directories like ``/tmp``. The common method of exploitation of this flaw
194 is to cross privilege boundaries when following a given hardlink (i.e. a
195 root process follows a hardlink created by another user). Additionally,
196 on systems without separated partitions, this stops unauthorized users
197 from "pinning" vulnerable setuid/setgid files against being upgraded by
198 the administrator, or linking to special files.
200 When set to "0", hardlink creation behavior is unrestricted.
202 When set to "1" hardlinks cannot be created by users if they do not
203 already own the source file, or do not have read/write access to it.
205 This protection is based on the restrictions in Openwall and grsecurity.
211 This protection is similar to `protected_fifos`_, but it
212 avoids writes to an attacker-controlled regular file, where a program
213 expected to create one.
215 When set to "0", writing to regular files is unrestricted.
217 When set to "1" don't allow ``O_CREAT`` open on regular files that we
218 don't own in world writable sticky directories, unless they are
219 owned by the owner of the directory.
221 When set to "2" it also applies to group writable sticky directories.
227 A long-standing class of security issues is the symlink-based
228 time-of-check-time-of-use race, most commonly seen in world-writable
229 directories like ``/tmp``. The common method of exploitation of this flaw
230 is to cross privilege boundaries when following a given symlink (i.e. a
231 root process follows a symlink belonging to another user). For a likely
232 incomplete list of hundreds of examples across the years, please see:
233 https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=/tmp
235 When set to "0", symlink following behavior is unrestricted.
237 When set to "1" symlinks are permitted to be followed only when outside
238 a sticky world-writable directory, or when the uid of the symlink and
239 follower match, or when the directory owner matches the symlink's owner.
241 This protection is based on the restrictions in Openwall and grsecurity.
247 This value can be used to query and set the core dump mode for setuid
248 or otherwise protected/tainted binaries. The modes are
250 = ========== ===============================================================
251 0 (default) Traditional behaviour. Any process which has changed
252 privilege levels or is execute only will not be dumped.
253 1 (debug) All processes dump core when possible. The core dump is
254 owned by the current user and no security is applied. This is
255 intended for system debugging situations only.
257 This is insecure as it allows regular users to examine the
258 memory contents of privileged processes.
259 2 (suidsafe) Any binary which normally would not be dumped is dumped
260 anyway, but only if the ``core_pattern`` kernel sysctl (see
261 :ref:`Documentation/admin-guide/sysctl/kernel.rst <core_pattern>`)
263 either a pipe handler or a fully qualified path. (For more
264 details on this limitation, see CVE-2006-2451.) This mode is
265 appropriate when administrators are attempting to debug
266 problems in a normal environment, and either have a core dump
267 pipe handler that knows to treat privileged core dumps with
268 care, or specific directory defined for catching core dumps.
269 If a core dump happens without a pipe handler or fully
270 qualified path, a message will be emitted to syslog warning
271 about the lack of a correct setting.
272 = ========== ===============================================================
276 2. /proc/sys/fs/binfmt_misc
277 ===========================
279 Documentation for the files in ``/proc/sys/fs/binfmt_misc`` is
280 in Documentation/admin-guide/binfmt-misc.rst.
283 3. /proc/sys/fs/mqueue - POSIX message queues filesystem
284 ========================================================
287 The "mqueue" filesystem provides the necessary kernel features to enable the
288 creation of a user space library that implements the POSIX message queues
289 API (as noted by the MSG tag in the POSIX 1003.1-2001 version of the System
290 Interfaces specification.)
292 The "mqueue" filesystem contains values for determining/setting the
293 amount of resources used by the file system.
295 ``/proc/sys/fs/mqueue/queues_max`` is a read/write file for
296 setting/getting the maximum number of message queues allowed on the
299 ``/proc/sys/fs/mqueue/msg_max`` is a read/write file for
300 setting/getting the maximum number of messages in a queue value. In
301 fact it is the limiting value for another (user) limit which is set in
302 ``mq_open`` invocation. This attribute of a queue must be less than
303 or equal to ``msg_max``.
305 ``/proc/sys/fs/mqueue/msgsize_max`` is a read/write file for
306 setting/getting the maximum message size value (it is an attribute of
307 every message queue, set during its creation).
309 ``/proc/sys/fs/mqueue/msg_default`` is a read/write file for
310 setting/getting the default number of messages in a queue value if the
311 ``attr`` parameter of ``mq_open(2)`` is ``NULL``. If it exceeds
312 ``msg_max``, the default value is initialized to ``msg_max``.
314 ``/proc/sys/fs/mqueue/msgsize_default`` is a read/write file for
315 setting/getting the default message size value if the ``attr``
316 parameter of ``mq_open(2)`` is ``NULL``. If it exceeds
317 ``msgsize_max``, the default value is initialized to ``msgsize_max``.
319 4. /proc/sys/fs/epoll - Configuration options for the epoll interface
320 =====================================================================
322 This directory contains configuration options for the epoll(7) interface.
327 Every epoll file descriptor can store a number of files to be monitored
328 for event readiness. Each one of these monitored files constitutes a "watch".
329 This configuration option sets the maximum number of "watches" that are
330 allowed for each user.
331 Each "watch" costs roughly 90 bytes on a 32-bit kernel, and roughly 160 bytes
333 The current default value for ``max_user_watches`` is 4% of the
334 available low memory, divided by the "watch" cost in bytes.