0001 ===============================
0002 Documentation for /proc/sys/fs/
0003 ===============================
0004
0005 kernel version 2.2.10
0006
0007 Copyright (c) 1998, 1999, Rik van Riel <riel@nl.linux.org>
0008
0009 Copyright (c) 2009, Shen Feng<shen@cn.fujitsu.com>
0010
0011 For general info and legal blurb, please look in intro.rst.
0012
0013 ------------------------------------------------------------------------------
0014
0015 This file contains documentation for the sysctl files in
0016 /proc/sys/fs/ and is valid for Linux kernel version 2.2.
0017
0018 The files in this directory can be used to tune and monitor
0019 miscellaneous and general things in the operation of the Linux
0020 kernel. Since some of the files _can_ be used to screw up your
0021 system, it is advisable to read both documentation and source
0022 before actually making adjustments.
0023
0024 1. /proc/sys/fs
0025 ===============
0026
0027 Currently, these files are in /proc/sys/fs:
0028
0029 - aio-max-nr
0030 - aio-nr
0031 - dentry-state
0032 - dquot-max
0033 - dquot-nr
0034 - file-max
0035 - file-nr
0036 - inode-max
0037 - inode-nr
0038 - inode-state
0039 - nr_open
0040 - overflowuid
0041 - overflowgid
0042 - pipe-user-pages-hard
0043 - pipe-user-pages-soft
0044 - protected_fifos
0045 - protected_hardlinks
0046 - protected_regular
0047 - protected_symlinks
0048 - suid_dumpable
0049 - super-max
0050 - super-nr
0051
0052
0053 aio-nr & aio-max-nr
0054 -------------------
0055
0056 aio-nr is the running total of the number of events specified on the
0057 io_setup system call for all currently active aio contexts. If aio-nr
0058 reaches aio-max-nr then io_setup will fail with EAGAIN. Note that
0059 raising aio-max-nr does not result in the pre-allocation or re-sizing
0060 of any kernel data structures.
0061
0062
0063 dentry-state
0064 ------------
0065
0066 From linux/include/linux/dcache.h::
0067
0068 struct dentry_stat_t dentry_stat {
0069 int nr_dentry;
0070 int nr_unused;
0071 int age_limit; /* age in seconds */
0072 int want_pages; /* pages requested by system */
0073 int nr_negative; /* # of unused negative dentries */
0074 int dummy; /* Reserved for future use */
0075 };
0076
0077 Dentries are dynamically allocated and deallocated.
0078
0079 nr_dentry shows the total number of dentries allocated (active
0080 + unused). nr_unused shows the number of dentries that are not
0081 actively used, but are saved in the LRU list for future reuse.
0082
0083 Age_limit is the age in seconds after which dcache entries
0084 can be reclaimed when memory is short and want_pages is
0085 nonzero when shrink_dcache_pages() has been called and the
0086 dcache isn't pruned yet.
0087
0088 nr_negative shows the number of unused dentries that are also
0089 negative dentries which do not map to any files. Instead,
0090 they help speeding up rejection of non-existing files provided
0091 by the users.
0092
0093
0094 dquot-max & dquot-nr
0095 --------------------
0096
0097 The file dquot-max shows the maximum number of cached disk
0098 quota entries.
0099
0100 The file dquot-nr shows the number of allocated disk quota
0101 entries and the number of free disk quota entries.
0102
0103 If the number of free cached disk quotas is very low and
0104 you have some awesome number of simultaneous system users,
0105 you might want to raise the limit.
0106
0107
0108 file-max & file-nr
0109 ------------------
0110
0111 The value in file-max denotes the maximum number of file-
0112 handles that the Linux kernel will allocate. When you get lots
0113 of error messages about running out of file handles, you might
0114 want to increase this limit.
0115
0116 Historically,the kernel was able to allocate file handles
0117 dynamically, but not to free them again. The three values in
0118 file-nr denote the number of allocated file handles, the number
0119 of allocated but unused file handles, and the maximum number of
0120 file handles. Linux 2.6 always reports 0 as the number of free
0121 file handles -- this is not an error, it just means that the
0122 number of allocated file handles exactly matches the number of
0123 used file handles.
0124
0125 Attempts to allocate more file descriptors than file-max are
0126 reported with printk, look for "VFS: file-max limit <number>
0127 reached".
0128
0129
0130 nr_open
0131 -------
0132
0133 This denotes the maximum number of file-handles a process can
0134 allocate. Default value is 1024*1024 (1048576) which should be
0135 enough for most machines. Actual limit depends on RLIMIT_NOFILE
0136 resource limit.
0137
0138
0139 inode-max, inode-nr & inode-state
0140 ---------------------------------
0141
0142 As with file handles, the kernel allocates the inode structures
0143 dynamically, but can't free them yet.
0144
0145 The value in inode-max denotes the maximum number of inode
0146 handlers. This value should be 3-4 times larger than the value
0147 in file-max, since stdin, stdout and network sockets also
0148 need an inode struct to handle them. When you regularly run
0149 out of inodes, you need to increase this value.
0150
0151 The file inode-nr contains the first two items from
0152 inode-state, so we'll skip to that file...
0153
0154 Inode-state contains three actual numbers and four dummies.
0155 The actual numbers are, in order of appearance, nr_inodes,
0156 nr_free_inodes and preshrink.
0157
0158 Nr_inodes stands for the number of inodes the system has
0159 allocated, this can be slightly more than inode-max because
0160 Linux allocates them one pageful at a time.
0161
0162 Nr_free_inodes represents the number of free inodes (?) and
0163 preshrink is nonzero when the nr_inodes > inode-max and the
0164 system needs to prune the inode list instead of allocating
0165 more.
0166
0167
0168 overflowgid & overflowuid
0169 -------------------------
0170
0171 Some filesystems only support 16-bit UIDs and GIDs, although in Linux
0172 UIDs and GIDs are 32 bits. When one of these filesystems is mounted
0173 with writes enabled, any UID or GID that would exceed 65535 is translated
0174 to a fixed value before being written to disk.
0175
0176 These sysctls allow you to change the value of the fixed UID and GID.
0177 The default is 65534.
0178
0179
0180 pipe-user-pages-hard
0181 --------------------
0182
0183 Maximum total number of pages a non-privileged user may allocate for pipes.
0184 Once this limit is reached, no new pipes may be allocated until usage goes
0185 below the limit again. When set to 0, no limit is applied, which is the default
0186 setting.
0187
0188
0189 pipe-user-pages-soft
0190 --------------------
0191
0192 Maximum total number of pages a non-privileged user may allocate for pipes
0193 before the pipe size gets limited to a single page. Once this limit is reached,
0194 new pipes will be limited to a single page in size for this user in order to
0195 limit total memory usage, and trying to increase them using fcntl() will be
0196 denied until usage goes below the limit again. The default value allows to
0197 allocate up to 1024 pipes at their default size. When set to 0, no limit is
0198 applied.
0199
0200
0201 protected_fifos
0202 ---------------
0203
0204 The intent of this protection is to avoid unintentional writes to
0205 an attacker-controlled FIFO, where a program expected to create a regular
0206 file.
0207
0208 When set to "0", writing to FIFOs is unrestricted.
0209
0210 When set to "1" don't allow O_CREAT open on FIFOs that we don't own
0211 in world writable sticky directories, unless they are owned by the
0212 owner of the directory.
0213
0214 When set to "2" it also applies to group writable sticky directories.
0215
0216 This protection is based on the restrictions in Openwall.
0217
0218
0219 protected_hardlinks
0220 --------------------
0221
0222 A long-standing class of security issues is the hardlink-based
0223 time-of-check-time-of-use race, most commonly seen in world-writable
0224 directories like /tmp. The common method of exploitation of this flaw
0225 is to cross privilege boundaries when following a given hardlink (i.e. a
0226 root process follows a hardlink created by another user). Additionally,
0227 on systems without separated partitions, this stops unauthorized users
0228 from "pinning" vulnerable setuid/setgid files against being upgraded by
0229 the administrator, or linking to special files.
0230
0231 When set to "0", hardlink creation behavior is unrestricted.
0232
0233 When set to "1" hardlinks cannot be created by users if they do not
0234 already own the source file, or do not have read/write access to it.
0235
0236 This protection is based on the restrictions in Openwall and grsecurity.
0237
0238
0239 protected_regular
0240 -----------------
0241
0242 This protection is similar to protected_fifos, but it
0243 avoids writes to an attacker-controlled regular file, where a program
0244 expected to create one.
0245
0246 When set to "0", writing to regular files is unrestricted.
0247
0248 When set to "1" don't allow O_CREAT open on regular files that we
0249 don't own in world writable sticky directories, unless they are
0250 owned by the owner of the directory.
0251
0252 When set to "2" it also applies to group writable sticky directories.
0253
0254
0255 protected_symlinks
0256 ------------------
0257
0258 A long-standing class of security issues is the symlink-based
0259 time-of-check-time-of-use race, most commonly seen in world-writable
0260 directories like /tmp. The common method of exploitation of this flaw
0261 is to cross privilege boundaries when following a given symlink (i.e. a
0262 root process follows a symlink belonging to another user). For a likely
0263 incomplete list of hundreds of examples across the years, please see:
0264 https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=/tmp
0265
0266 When set to "0", symlink following behavior is unrestricted.
0267
0268 When set to "1" symlinks are permitted to be followed only when outside
0269 a sticky world-writable directory, or when the uid of the symlink and
0270 follower match, or when the directory owner matches the symlink's owner.
0271
0272 This protection is based on the restrictions in Openwall and grsecurity.
0273
0274
0275 suid_dumpable:
0276 --------------
0277
0278 This value can be used to query and set the core dump mode for setuid
0279 or otherwise protected/tainted binaries. The modes are
0280
0281 = ========== ===============================================================
0282 0 (default) traditional behaviour. Any process which has changed
0283 privilege levels or is execute only will not be dumped.
0284 1 (debug) all processes dump core when possible. The core dump is
0285 owned by the current user and no security is applied. This is
0286 intended for system debugging situations only.
0287 Ptrace is unchecked.
0288 This is insecure as it allows regular users to examine the
0289 memory contents of privileged processes.
0290 2 (suidsafe) any binary which normally would not be dumped is dumped
0291 anyway, but only if the "core_pattern" kernel sysctl is set to
0292 either a pipe handler or a fully qualified path. (For more
0293 details on this limitation, see CVE-2006-2451.) This mode is
0294 appropriate when administrators are attempting to debug
0295 problems in a normal environment, and either have a core dump
0296 pipe handler that knows to treat privileged core dumps with
0297 care, or specific directory defined for catching core dumps.
0298 If a core dump happens without a pipe handler or fully
0299 qualified path, a message will be emitted to syslog warning
0300 about the lack of a correct setting.
0301 = ========== ===============================================================
0302
0303
0304 super-max & super-nr
0305 --------------------
0306
0307 These numbers control the maximum number of superblocks, and
0308 thus the maximum number of mounted filesystems the kernel
0309 can have. You only need to increase super-max if you need to
0310 mount more filesystems than the current value in super-max
0311 allows you to.
0312
0313
0314 aio-nr & aio-max-nr
0315 -------------------
0316
0317 aio-nr shows the current system-wide number of asynchronous io
0318 requests. aio-max-nr allows you to change the maximum value
0319 aio-nr can grow to.
0320
0321
0322 mount-max
0323 ---------
0324
0325 This denotes the maximum number of mounts that may exist
0326 in a mount namespace.
0327
0328
0329
0330 2. /proc/sys/fs/binfmt_misc
0331 ===========================
0332
0333 Documentation for the files in /proc/sys/fs/binfmt_misc is
0334 in Documentation/admin-guide/binfmt-misc.rst.
0335
0336
0337 3. /proc/sys/fs/mqueue - POSIX message queues filesystem
0338 ========================================================
0339
0340
0341 The "mqueue" filesystem provides the necessary kernel features to enable the
0342 creation of a user space library that implements the POSIX message queues
0343 API (as noted by the MSG tag in the POSIX 1003.1-2001 version of the System
0344 Interfaces specification.)
0345
0346 The "mqueue" filesystem contains values for determining/setting the amount of
0347 resources used by the file system.
0348
0349 /proc/sys/fs/mqueue/queues_max is a read/write file for setting/getting the
0350 maximum number of message queues allowed on the system.
0351
0352 /proc/sys/fs/mqueue/msg_max is a read/write file for setting/getting the
0353 maximum number of messages in a queue value. In fact it is the limiting value
0354 for another (user) limit which is set in mq_open invocation. This attribute of
0355 a queue must be less or equal then msg_max.
0356
0357 /proc/sys/fs/mqueue/msgsize_max is a read/write file for setting/getting the
0358 maximum message size value (it is every message queue's attribute set during
0359 its creation).
0360
0361 /proc/sys/fs/mqueue/msg_default is a read/write file for setting/getting the
0362 default number of messages in a queue value if attr parameter of mq_open(2) is
0363 NULL. If it exceed msg_max, the default value is initialized msg_max.
0364
0365 /proc/sys/fs/mqueue/msgsize_default is a read/write file for setting/getting
0366 the default message size value if attr parameter of mq_open(2) is NULL. If it
0367 exceed msgsize_max, the default value is initialized msgsize_max.
0368
0369 4. /proc/sys/fs/epoll - Configuration options for the epoll interface
0370 =====================================================================
0371
0372 This directory contains configuration options for the epoll(7) interface.
0373
0374 max_user_watches
0375 ----------------
0376
0377 Every epoll file descriptor can store a number of files to be monitored
0378 for event readiness. Each one of these monitored files constitutes a "watch".
0379 This configuration option sets the maximum number of "watches" that are
0380 allowed for each user.
0381 Each "watch" costs roughly 90 bytes on a 32bit kernel, and roughly 160 bytes
0382 on a 64bit one.
0383 The current default value for max_user_watches is the 1/25 (4%) of the
0384 available low memory, divided for the "watch" cost in bytes.
