1 Documentation 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>
5 For general info and legal blurb, please look in README.
7 ==============================================================
9 This file contains documentation for the sysctl files in
10 /proc/sys/fs/ and is valid for Linux kernel version 2.2.
12 The files in this directory can be used to tune and monitor
13 miscellaneous and general things in the operation of the Linux
14 kernel. Since some of the files _can_ be used to screw up your
15 system, it is advisable to read both documentation and source
16 before actually making adjustments.
19 ----------------------------------------------------------
21 Currently, these files are in /proc/sys/fs:
35 - pipe-user-pages-hard
36 - pipe-user-pages-soft
45 ==============================================================
49 aio-nr is the running total of the number of events specified on the
50 io_setup system call for all currently active aio contexts. If aio-nr
51 reaches aio-max-nr then io_setup will fail with EAGAIN. Note that
52 raising aio-max-nr does not result in the pre-allocation or re-sizing
53 of any kernel data structures.
55 ==============================================================
59 From linux/fs/dentry.c:
60 --------------------------------------------------------------
64 int age_limit; /* age in seconds */
65 int want_pages; /* pages requested by system */
67 } dentry_stat = {0, 0, 45, 0,};
68 --------------------------------------------------------------
70 Dentries are dynamically allocated and deallocated, and
71 nr_dentry seems to be 0 all the time. Hence it's safe to
72 assume that only nr_unused, age_limit and want_pages are
73 used. Nr_unused seems to be exactly what its name says.
74 Age_limit is the age in seconds after which dcache entries
75 can be reclaimed when memory is short and want_pages is
76 nonzero when shrink_dcache_pages() has been called and the
77 dcache isn't pruned yet.
79 ==============================================================
83 The file dquot-max shows the maximum number of cached disk
86 The file dquot-nr shows the number of allocated disk quota
87 entries and the number of free disk quota entries.
89 If the number of free cached disk quotas is very low and
90 you have some awesome number of simultaneous system users,
91 you might want to raise the limit.
93 ==============================================================
97 The value in file-max denotes the maximum number of file-
98 handles that the Linux kernel will allocate. When you get lots
99 of error messages about running out of file handles, you might
100 want to increase this limit.
102 Historically,the kernel was able to allocate file handles
103 dynamically, but not to free them again. The three values in
104 file-nr denote the number of allocated file handles, the number
105 of allocated but unused file handles, and the maximum number of
106 file handles. Linux 2.6 always reports 0 as the number of free
107 file handles -- this is not an error, it just means that the
108 number of allocated file handles exactly matches the number of
111 Attempts to allocate more file descriptors than file-max are
112 reported with printk, look for "VFS: file-max limit <number>
114 ==============================================================
118 This denotes the maximum number of file-handles a process can
119 allocate. Default value is 1024*1024 (1048576) which should be
120 enough for most machines. Actual limit depends on RLIMIT_NOFILE
123 ==============================================================
125 inode-max, inode-nr & inode-state:
127 As with file handles, the kernel allocates the inode structures
128 dynamically, but can't free them yet.
130 The value in inode-max denotes the maximum number of inode
131 handlers. This value should be 3-4 times larger than the value
132 in file-max, since stdin, stdout and network sockets also
133 need an inode struct to handle them. When you regularly run
134 out of inodes, you need to increase this value.
136 The file inode-nr contains the first two items from
137 inode-state, so we'll skip to that file...
139 Inode-state contains three actual numbers and four dummies.
140 The actual numbers are, in order of appearance, nr_inodes,
141 nr_free_inodes and preshrink.
143 Nr_inodes stands for the number of inodes the system has
144 allocated, this can be slightly more than inode-max because
145 Linux allocates them one pageful at a time.
147 Nr_free_inodes represents the number of free inodes (?) and
148 preshrink is nonzero when the nr_inodes > inode-max and the
149 system needs to prune the inode list instead of allocating
152 ==============================================================
154 overflowgid & overflowuid:
156 Some filesystems only support 16-bit UIDs and GIDs, although in Linux
157 UIDs and GIDs are 32 bits. When one of these filesystems is mounted
158 with writes enabled, any UID or GID that would exceed 65535 is translated
159 to a fixed value before being written to disk.
161 These sysctls allow you to change the value of the fixed UID and GID.
162 The default is 65534.
164 ==============================================================
166 pipe-user-pages-hard:
168 Maximum total number of pages a non-privileged user may allocate for pipes.
169 Once this limit is reached, no new pipes may be allocated until usage goes
170 below the limit again. When set to 0, no limit is applied, which is the default
173 ==============================================================
175 pipe-user-pages-soft:
177 Maximum total number of pages a non-privileged user may allocate for pipes
178 before the pipe size gets limited to a single page. Once this limit is reached,
179 new pipes will be limited to a single page in size for this user in order to
180 limit total memory usage, and trying to increase them using fcntl() will be
181 denied until usage goes below the limit again. The default value allows to
182 allocate up to 1024 pipes at their default size. When set to 0, no limit is
185 ==============================================================
189 The intent of this protection is to avoid unintentional writes to
190 an attacker-controlled FIFO, where a program expected to create a regular
193 When set to "0", writing to FIFOs is unrestricted.
195 When set to "1" don't allow O_CREAT open on FIFOs that we don't own
196 in world writable sticky directories, unless they are owned by the
197 owner of the directory.
199 When set to "2" it also applies to group writable sticky directories.
201 This protection is based on the restrictions in Openwall.
203 ==============================================================
207 A long-standing class of security issues is the hardlink-based
208 time-of-check-time-of-use race, most commonly seen in world-writable
209 directories like /tmp. The common method of exploitation of this flaw
210 is to cross privilege boundaries when following a given hardlink (i.e. a
211 root process follows a hardlink created by another user). Additionally,
212 on systems without separated partitions, this stops unauthorized users
213 from "pinning" vulnerable setuid/setgid files against being upgraded by
214 the administrator, or linking to special files.
216 When set to "0", hardlink creation behavior is unrestricted.
218 When set to "1" hardlinks cannot be created by users if they do not
219 already own the source file, or do not have read/write access to it.
221 This protection is based on the restrictions in Openwall and grsecurity.
223 ==============================================================
227 This protection is similar to protected_fifos, but it
228 avoids writes to an attacker-controlled regular file, where a program
229 expected to create one.
231 When set to "0", writing to regular files is unrestricted.
233 When set to "1" don't allow O_CREAT open on regular files that we
234 don't own in world writable sticky directories, unless they are
235 owned by the owner of the directory.
237 When set to "2" it also applies to group writable sticky directories.
239 ==============================================================
243 A long-standing class of security issues is the symlink-based
244 time-of-check-time-of-use race, most commonly seen in world-writable
245 directories like /tmp. The common method of exploitation of this flaw
246 is to cross privilege boundaries when following a given symlink (i.e. a
247 root process follows a symlink belonging to another user). For a likely
248 incomplete list of hundreds of examples across the years, please see:
249 http://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=/tmp
251 When set to "0", symlink following behavior is unrestricted.
253 When set to "1" symlinks are permitted to be followed only when outside
254 a sticky world-writable directory, or when the uid of the symlink and
255 follower match, or when the directory owner matches the symlink's owner.
257 This protection is based on the restrictions in Openwall and grsecurity.
259 ==============================================================
263 This value can be used to query and set the core dump mode for setuid
264 or otherwise protected/tainted binaries. The modes are
266 0 - (default) - traditional behaviour. Any process which has changed
267 privilege levels or is execute only will not be dumped.
268 1 - (debug) - all processes dump core when possible. The core dump is
269 owned by the current user and no security is applied. This is
270 intended for system debugging situations only. Ptrace is unchecked.
271 This is insecure as it allows regular users to examine the memory
272 contents of privileged processes.
273 2 - (suidsafe) - any binary which normally would not be dumped is dumped
274 anyway, but only if the "core_pattern" kernel sysctl is set to
275 either a pipe handler or a fully qualified path. (For more details
276 on this limitation, see CVE-2006-2451.) This mode is appropriate
277 when administrators are attempting to debug problems in a normal
278 environment, and either have a core dump pipe handler that knows
279 to treat privileged core dumps with care, or specific directory
280 defined for catching core dumps. If a core dump happens without
281 a pipe handler or fully qualifid path, a message will be emitted
282 to syslog warning about the lack of a correct setting.
284 ==============================================================
286 super-max & super-nr:
288 These numbers control the maximum number of superblocks, and
289 thus the maximum number of mounted filesystems the kernel
290 can have. You only need to increase super-max if you need to
291 mount more filesystems than the current value in super-max
294 ==============================================================
298 aio-nr shows the current system-wide number of asynchronous io
299 requests. aio-max-nr allows you to change the maximum value
302 ==============================================================
306 This denotes the maximum number of mounts that may exist
307 in a mount namespace.
309 ==============================================================
312 2. /proc/sys/fs/binfmt_misc
313 ----------------------------------------------------------
315 Documentation for the files in /proc/sys/fs/binfmt_misc is
316 in Documentation/admin-guide/binfmt-misc.rst.
319 3. /proc/sys/fs/mqueue - POSIX message queues filesystem
320 ----------------------------------------------------------
322 The "mqueue" filesystem provides the necessary kernel features to enable the
323 creation of a user space library that implements the POSIX message queues
324 API (as noted by the MSG tag in the POSIX 1003.1-2001 version of the System
325 Interfaces specification.)
327 The "mqueue" filesystem contains values for determining/setting the amount of
328 resources used by the file system.
330 /proc/sys/fs/mqueue/queues_max is a read/write file for setting/getting the
331 maximum number of message queues allowed on the system.
333 /proc/sys/fs/mqueue/msg_max is a read/write file for setting/getting the
334 maximum number of messages in a queue value. In fact it is the limiting value
335 for another (user) limit which is set in mq_open invocation. This attribute of
336 a queue must be less or equal then msg_max.
338 /proc/sys/fs/mqueue/msgsize_max is a read/write file for setting/getting the
339 maximum message size value (it is every message queue's attribute set during
342 /proc/sys/fs/mqueue/msg_default is a read/write file for setting/getting the
343 default number of messages in a queue value if attr parameter of mq_open(2) is
344 NULL. If it exceed msg_max, the default value is initialized msg_max.
346 /proc/sys/fs/mqueue/msgsize_default is a read/write file for setting/getting
347 the default message size value if attr parameter of mq_open(2) is NULL. If it
348 exceed msgsize_max, the default value is initialized msgsize_max.
350 4. /proc/sys/fs/epoll - Configuration options for the epoll interface
351 --------------------------------------------------------
353 This directory contains configuration options for the epoll(7) interface.
358 Every epoll file descriptor can store a number of files to be monitored
359 for event readiness. Each one of these monitored files constitutes a "watch".
360 This configuration option sets the maximum number of "watches" that are
361 allowed for each user.
362 Each "watch" costs roughly 90 bytes on a 32bit kernel, and roughly 160 bytes
364 The current default value for max_user_watches is the 1/32 of the available
365 low memory, divided for the "watch" cost in bytes.