1 ===============================
2 Documentation for /proc/sys/fs/
3 ===============================
7 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
9 Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>
11 For general info and legal blurb, please look in intro.rst.
13 ------------------------------------------------------------------------------
15 This file contains documentation for the sysctl files in
16 /proc/sys/fs/ and is valid for Linux kernel version 2.2.
18 The files in this directory can be used to tune and monitor
19 miscellaneous and general things in the operation of the Linux
20 kernel. Since some of the files _can_ be used to screw up your
21 system, it is advisable to read both documentation and source
22 before actually making adjustments.
27 Currently, these files are in /proc/sys/fs:
42 - pipe-user-pages-hard
43 - pipe-user-pages-soft
56 aio-nr is the running total of the number of events specified on the
57 io_setup system call for all currently active aio contexts. If aio-nr
58 reaches aio-max-nr then io_setup will fail with EAGAIN. Note that
59 raising aio-max-nr does not result in the pre-allocation or re-sizing
60 of any kernel data structures.
66 From linux/include/linux/dcache.h::
68 struct dentry_stat_t dentry_stat {
71 int age_limit; /* age in seconds */
72 int want_pages; /* pages requested by system */
73 int nr_negative; /* # of unused negative dentries */
74 int dummy; /* Reserved for future use */
77 Dentries are dynamically allocated and deallocated.
79 nr_dentry shows the total number of dentries allocated (active
80 + unused). nr_unused shows the number of dentries that are not
81 actively used, but are saved in the LRU list for future reuse.
83 Age_limit is the age in seconds after which dcache entries
84 can be reclaimed when memory is short and want_pages is
85 nonzero when shrink_dcache_pages() has been called and the
86 dcache isn't pruned yet.
88 nr_negative shows the number of unused dentries that are also
89 negative dentries which do not map to any files. Instead,
90 they help speeding up rejection of non-existing files provided
97 The file dquot-max shows the maximum number of cached disk
100 The file dquot-nr shows the number of allocated disk quota
101 entries and the number of free disk quota entries.
103 If the number of free cached disk quotas is very low and
104 you have some awesome number of simultaneous system users,
105 you might want to raise the limit.
111 The value in file-max denotes the maximum number of file-
112 handles that the Linux kernel will allocate. When you get lots
113 of error messages about running out of file handles, you might
114 want to increase this limit.
116 Historically,the kernel was able to allocate file handles
117 dynamically, but not to free them again. The three values in
118 file-nr denote the number of allocated file handles, the number
119 of allocated but unused file handles, and the maximum number of
120 file handles. Linux 2.6 always reports 0 as the number of free
121 file handles -- this is not an error, it just means that the
122 number of allocated file handles exactly matches the number of
125 Attempts to allocate more file descriptors than file-max are
126 reported with printk, look for "VFS: file-max limit <number>
133 This denotes the maximum number of file-handles a process can
134 allocate. Default value is 1024*1024 (1048576) which should be
135 enough for most machines. Actual limit depends on RLIMIT_NOFILE
139 inode-max, inode-nr & inode-state
140 ---------------------------------
142 As with file handles, the kernel allocates the inode structures
143 dynamically, but can't free them yet.
145 The value in inode-max denotes the maximum number of inode
146 handlers. This value should be 3-4 times larger than the value
147 in file-max, since stdin, stdout and network sockets also
148 need an inode struct to handle them. When you regularly run
149 out of inodes, you need to increase this value.
151 The file inode-nr contains the first two items from
152 inode-state, so we'll skip to that file...
154 Inode-state contains three actual numbers and four dummies.
155 The actual numbers are, in order of appearance, nr_inodes,
156 nr_free_inodes and preshrink.
158 Nr_inodes stands for the number of inodes the system has
159 allocated, this can be slightly more than inode-max because
160 Linux allocates them one pageful at a time.
162 Nr_free_inodes represents the number of free inodes (?) and
163 preshrink is nonzero when the nr_inodes > inode-max and the
164 system needs to prune the inode list instead of allocating
168 overflowgid & overflowuid
169 -------------------------
171 Some filesystems only support 16-bit UIDs and GIDs, although in Linux
172 UIDs and GIDs are 32 bits. When one of these filesystems is mounted
173 with writes enabled, any UID or GID that would exceed 65535 is translated
174 to a fixed value before being written to disk.
176 These sysctls allow you to change the value of the fixed UID and GID.
177 The default is 65534.
183 Maximum total number of pages a non-privileged user may allocate for pipes.
184 Once this limit is reached, no new pipes may be allocated until usage goes
185 below the limit again. When set to 0, no limit is applied, which is the default
192 Maximum total number of pages a non-privileged user may allocate for pipes
193 before the pipe size gets limited to a single page. Once this limit is reached,
194 new pipes will be limited to a single page in size for this user in order to
195 limit total memory usage, and trying to increase them using fcntl() will be
196 denied until usage goes below the limit again. The default value allows to
197 allocate up to 1024 pipes at their default size. When set to 0, no limit is
204 The intent of this protection is to avoid unintentional writes to
205 an attacker-controlled FIFO, where a program expected to create a regular
208 When set to "0", writing to FIFOs is unrestricted.
210 When set to "1" don't allow O_CREAT open on FIFOs that we don't own
211 in world writable sticky directories, unless they are owned by the
212 owner of the directory.
214 When set to "2" it also applies to group writable sticky directories.
216 This protection is based on the restrictions in Openwall.
222 A long-standing class of security issues is the hardlink-based
223 time-of-check-time-of-use race, most commonly seen in world-writable
224 directories like /tmp. The common method of exploitation of this flaw
225 is to cross privilege boundaries when following a given hardlink (i.e. a
226 root process follows a hardlink created by another user). Additionally,
227 on systems without separated partitions, this stops unauthorized users
228 from "pinning" vulnerable setuid/setgid files against being upgraded by
229 the administrator, or linking to special files.
231 When set to "0", hardlink creation behavior is unrestricted.
233 When set to "1" hardlinks cannot be created by users if they do not
234 already own the source file, or do not have read/write access to it.
236 This protection is based on the restrictions in Openwall and grsecurity.
242 This protection is similar to protected_fifos, but it
243 avoids writes to an attacker-controlled regular file, where a program
244 expected to create one.
246 When set to "0", writing to regular files is unrestricted.
248 When set to "1" don't allow O_CREAT open on regular files that we
249 don't own in world writable sticky directories, unless they are
250 owned by the owner of the directory.
252 When set to "2" it also applies to group writable sticky directories.
258 A long-standing class of security issues is the symlink-based
259 time-of-check-time-of-use race, most commonly seen in world-writable
260 directories like /tmp. The common method of exploitation of this flaw
261 is to cross privilege boundaries when following a given symlink (i.e. a
262 root process follows a symlink belonging to another user). For a likely
263 incomplete list of hundreds of examples across the years, please see:
264 http://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=/tmp
266 When set to "0", symlink following behavior is unrestricted.
268 When set to "1" symlinks are permitted to be followed only when outside
269 a sticky world-writable directory, or when the uid of the symlink and
270 follower match, or when the directory owner matches the symlink's owner.
272 This protection is based on the restrictions in Openwall and grsecurity.
278 This value can be used to query and set the core dump mode for setuid
279 or otherwise protected/tainted binaries. The modes are
281 = ========== ===============================================================
282 0 (default) traditional behaviour. Any process which has changed
283 privilege levels or is execute only will not be dumped.
284 1 (debug) all processes dump core when possible. The core dump is
285 owned by the current user and no security is applied. This is
286 intended for system debugging situations only.
288 This is insecure as it allows regular users to examine the
289 memory contents of privileged processes.
290 2 (suidsafe) any binary which normally would not be dumped is dumped
291 anyway, but only if the "core_pattern" kernel sysctl is set to
292 either a pipe handler or a fully qualified path. (For more
293 details on this limitation, see CVE-2006-2451.) This mode is
294 appropriate when administrators are attempting to debug
295 problems in a normal environment, and either have a core dump
296 pipe handler that knows to treat privileged core dumps with
297 care, or specific directory defined for catching core dumps.
298 If a core dump happens without a pipe handler or fully
299 qualified path, a message will be emitted to syslog warning
300 about the lack of a correct setting.
301 = ========== ===============================================================
307 These numbers control the maximum number of superblocks, and
308 thus the maximum number of mounted filesystems the kernel
309 can have. You only need to increase super-max if you need to
310 mount more filesystems than the current value in super-max
317 aio-nr shows the current system-wide number of asynchronous io
318 requests. aio-max-nr allows you to change the maximum value
325 This denotes the maximum number of mounts that may exist
326 in a mount namespace.
330 2. /proc/sys/fs/binfmt_misc
331 ===========================
333 Documentation for the files in /proc/sys/fs/binfmt_misc is
334 in Documentation/admin-guide/binfmt-misc.rst.
337 3. /proc/sys/fs/mqueue - POSIX message queues filesystem
338 ========================================================
341 The "mqueue" filesystem provides the necessary kernel features to enable the
342 creation of a user space library that implements the POSIX message queues
343 API (as noted by the MSG tag in the POSIX 1003.1-2001 version of the System
344 Interfaces specification.)
346 The "mqueue" filesystem contains values for determining/setting the amount of
347 resources used by the file system.
349 /proc/sys/fs/mqueue/queues_max is a read/write file for setting/getting the
350 maximum number of message queues allowed on the system.
352 /proc/sys/fs/mqueue/msg_max is a read/write file for setting/getting the
353 maximum number of messages in a queue value. In fact it is the limiting value
354 for another (user) limit which is set in mq_open invocation. This attribute of
355 a queue must be less or equal then msg_max.
357 /proc/sys/fs/mqueue/msgsize_max is a read/write file for setting/getting the
358 maximum message size value (it is every message queue's attribute set during
361 /proc/sys/fs/mqueue/msg_default is a read/write file for setting/getting the
362 default number of messages in a queue value if attr parameter of mq_open(2) is
363 NULL. If it exceed msg_max, the default value is initialized msg_max.
365 /proc/sys/fs/mqueue/msgsize_default is a read/write file for setting/getting
366 the default message size value if attr parameter of mq_open(2) is NULL. If it
367 exceed msgsize_max, the default value is initialized msgsize_max.
369 4. /proc/sys/fs/epoll - Configuration options for the epoll interface
370 =====================================================================
372 This directory contains configuration options for the epoll(7) interface.
377 Every epoll file descriptor can store a number of files to be monitored
378 for event readiness. Each one of these monitored files constitutes a "watch".
379 This configuration option sets the maximum number of "watches" that are
380 allowed for each user.
381 Each "watch" costs roughly 90 bytes on a 32bit kernel, and roughly 160 bytes
383 The current default value for max_user_watches is the 1/32 of the available
384 low memory, divided for the "watch" cost in bytes.